decoration decoration
Stories

GROKLAW
When you want to know more...
decoration
For layout only
Home
Archives
Site Map
Search
About Groklaw
Awards
Legal Research
Timelines
ApplevSamsung
ApplevSamsung p.2
ArchiveExplorer
Autozone
Bilski
Cases
Cast: Lawyers
Comes v. MS
Contracts/Documents
Courts
DRM
Gordon v MS
GPL
Grokdoc
HTML How To
IPI v RH
IV v. Google
Legal Docs
Lodsys
MS Litigations
MSvB&N
News Picks
Novell v. MS
Novell-MS Deal
ODF/OOXML
OOXML Appeals
OraclevGoogle
Patents
ProjectMonterey
Psystar
Quote Database
Red Hat v SCO
Salus Book
SCEA v Hotz
SCO Appeals
SCO Bankruptcy
SCO Financials
SCO Overview
SCO v IBM
SCO v Novell
SCO:Soup2Nuts
SCOsource
Sean Daly
Software Patents
Switch to Linux
Transcripts
Unix Books
Your contributions keep Groklaw going.
To donate to Groklaw 2.0:

Groklaw Gear

Click here to send an email to the editor of this weblog.


To read comments to this article, go here
Oracle v. Google - Third Damages Report: Arguments and Counter-Motions to Strike
Monday, February 27 2012 @ 02:45 PM EST

It's been months since we have seen the number of filings that were made last Friday, but the squabbling over Oracle's third attempt at an expert damages report from Dr. Cockburn is now fully engaged. Google had its say (Google Moves to Strike Portions of 3rd Oracle Damages Report and Google Takes Exception to Third Damages Report [as text]), and now it is Oracle's turn to respond.

Before heading off through the looking glass, there were a number of other filings that we will address. First, a reminder from the court that the parties are going to be required to disclose the compensation paid to their expert witnesses. (724 [PDF; Text]) It is clear the court wants the jury to be aware these experts have, at a minimum, a financial bias toward their client.

Second, we have the not altogether unexpected order on the issue of patent marking. (725 [PDF; Text]) Since the parties were only able to agree in a single instance on a stipulation with respect to the marking issue, the court is only adopting that single stipulation and throwing all of the so-called contingent stipulations out the window. The parties will have to fight this out at trial. As the court points out, however, this will not result in a longer trial, only a difference in how the time allocated to the parties for the trial will be spent. This will undoubtedly eat into the time available to argue other issues around the patent infringement claims, but then again, given the fact that most of those claims have been found invalid upon reexamination, maybe this will have little impact at all.

In the last of the orders issued by the court on Friday, the court directed the parties to "submit a candid discussion of the impact these rejections will have on the shape of trial," by March 9. The court has also asked whether the trial should be delayed to allow the last of the reexaminations to be completed. This last question may be a moot point as Robert Van Nest, lead counsel for Google, has reminded the court (748 [PDF; Text]) that he will likely be unavailable for trial in the April-June timed frame set aside by the court. That means that the trial will most likely be postponed until the September-December, 2012 timeframe, which should be sufficient time for the USPTO to complete the reexaminations.

In a related matter, Oracle has also asked (732 [PDF; Text]) to be able to file a supplemental expert report with respect to the patent validity issues raised by Google with respect to ’702 and ’205 and information relied upon by the USPTO in its rejection of the claims of these patents during reexamination. (723 [PDF; Text]) Now on to Oracle's response to Google's. Before considering all of the Oracle filings in response to Google's motion to strike portions of the third Oracle damages report, let's review Google's primary objections to the report. Google's objections largely centered on its belief that Dr. Cockburn, Oracle's damages expert, failed to respond to the instructions of the court when the court permitted this third attempt. Google asserts that the primary focus of the third report needed to be the apportionment of valuation between patent claims asserted and those not asserted and between copyright claims asserted and those not asserted. Google believes Dr. Cockburn has failed to address the apportionment issue with his third attempt, and most of its motion to strike (points B-D in the motion) focus on that failure. (See, Google Takes Exception to Third Damages Report [as text]) Apart from that criticism, Google addresses various methodologies which Dr. Cockburn and Dr. Shugan adopted in preparing the third damages report.

Oracle has filed five primary responsive documents, and we include the text of a sixth as exemplary of some of supporting responses provided by Oracle/Sun engineers. The six filings that we provide in text below are:

  • Oracle's response to Google's motion to strike portions of the third Cockburn report (737 [PDF; Text])
  • Cockburn's declaration in support of Oracle's response (739 [PDF; Text])
  • Shugan's declaration in support of Oracle's response (740 [PDF; Text])
  • Wong's declaration in support of Oracle's response (745 [PDF; Text])
  • Oracle's motion to strike portions of (Google expert) Leonard's supplemental report (criticizing the third Cockburn report) (729 [PDF; Text])
  • Oracle's motion to strike portions of (Google expert) Cox's supplemental report (criticizing the third Cockburn report) (734 [PDF; Text])

The two key points of contention are:

  • Valuing the copyrights: - According to Google, Cockburn valued the copyrights asserted by estimating the cost to Sun to independently develop a mobile platform. It is hard to understand the relevance of that number as it would have been the costs avoided by Google through access to that code that would better reflect the value. Oracle and Cockburn believe they have used the proper approach; Google disagrees. Oracle believes, if anything, they have overstated the value in Google's favor. But this is really an argument over precision with Google contending Oracle has not achieved the level of precision required by the court and Oracle contending it has.
  • Valuing the patents: The differences between the parties' positions on the patent value allocation mirrors that on the copyright side. Google contends Oracle has made no attempt to allocate value on a claim by claim basis with respect to the patent asserted and that Oracle has been imprecise in allocating value among all of the patents that would have been licensed to Google in the 2006 deal. Further, Google argues that the Oracle/Sun engineers engaged to assess the technical merits of the Sun patent portfolio was biased because these are the same engineers called upon to identify the best patents to assert against Google. Oracle, on its part, sees no bias and sees no failure.

Most of the remaining arguments are secondary to these. If the court finds that the apportionment methodologies adopted by Cockburn in this third attempt are plausible (they don't have to be optimal, simply plausible), then I would expect the Google motion to largely fail. If the court agrees with Google that the methodologies are not plausible, then Oracle is toast. I would place the odds in Oracle's favor simply because the court doesn't want to be seen as being too stiff-necked in deciding this matter.


Jump To Comments


***************

Docket

02/23/2012 - 724 - ORDER REGARDING EXPERTS' COMPENSATION. Signed by Judge Alsup on February 23, 2012. (whalc1, COURT STAFF) (Filed on 2/23/2012) (Entered: 02/23/2012)

02/24/2012 - 725 - ORDER REGARDING PATENT MARKING JOINT STATEMENT re 721 Statement filed by Google Inc., Oracle America, Inc.. Signed by Judge Alsup on February 24, 2012. (whalc1, COURT STAFF) (Filed on 2/24/2012) (Entered: 02/24/2012)

02/24/2012 - 726 - ORDER REGARDING REEXAMINATIONS re 722 Notice (Other) filed by Google Inc., Oracle America, Inc.. Signed by Judge Alsup on February 24, 2012. (whalc1, COURT STAFF) (Filed on 2/24/2012) (Entered: 02/24/2012)

02/24/2012 - 727 - Administrative Motion to File Under Seal PORTIONS OF MOTION TO STRIKE PORTIONS OF THE SUPPLEMENTAL EXPERT REPORT OF DR. GREGORY K. LEONARD filed by Oracle America, Inc.. (Attachments: # 1 Proposed Order)(Holtzman, Steven) (Filed on 2/24/2012) (Entered: 02/24/2012)

02/24/2012 - 728 - Declaration of ANDREW TEMKIN in Support of 727 Administrative Motion to File Under Seal PORTIONS OF MOTION TO STRIKE PORTIONS OF THE SUPPLEMENTAL EXPERT REPORT OF DR. GREGORY K. LEONARD filed byOracle America, Inc.. (Related document(s) 727 ) (Holtzman, Steven) (Filed on 2/24/2012) (Entered: 02/24/2012)

02/24/2012 - 729 - MOTION to Strike PORTIONS OF GREGORY LEONARDS SUPPLEMENTAL REPORT filed by Oracle America, Inc.. Responses due by 3/2/2012. Replies due by 3/6/2012. (Holtzman, Steven) (Filed on 2/24/2012) (Entered: 02/24/2012)

02/24/2012 - 730 - Declaration of IAIN M. COCKBURN in Support of 729 MOTION to Strike PORTIONS OF GREGORY LEONARDS SUPPLEMENTAL REPORT filed byOracle America, Inc.. (Related document(s) 729 ) (Holtzman, Steven) (Filed on 2/24/2012) (Entered: 02/24/2012)

02/24/2012 - 731 - Declaration of BEKO REBLITZ-RICHARDSON in Support of 729 MOTION to Strike PORTIONS OF GREGORY LEONARDS SUPPLEMENTAL REPORT filed by Oracle America, Inc.. (Attachments: # 1 Exhibit A, # 2 Exhibit B, # 3 Exhibit C, # 4 Exhibit C, # 5 Exhibit E, # 6 Exhibit F, # 7 Exhibit G, # 8 Exhibit H)(Related document(s) 729 ) (Holtzman, Steven) (Filed on 2/24/2012) (Entered: 02/24/2012)

02/24/2012 - 732 - RESPONSE to re 723 Letter regarding '702 and '205 invalidity supplementation by Oracle America, Inc.. (Jacobs, Michael) (Filed on 2/24/2012) (Entered: 02/24/2012)

02/24/2012 - 733 - Administrative Motion to File Under Seal PORTIONS OF ITS MOTION TO STRIK PORTIONS OF THE SUPPLEMENTAL EXPERT REPORT OF DR. ALAN J. COX filed by Oracle America, Inc.. (Attachments: # 1 Proposed Order)(Holtzman, Steven) (Filed on 2/24/2012) (Entered: 02/24/2012)

02/24/2012 - 734 - MOTION to Strike EXCLUDE PORTIONS OF THE SUPPLEMENTAL EXPERT REPORT OF DR. ALAN J. COX filed by Oracle America, Inc.. Responses due by 3/2/2012. Replies due by 3/6/2012. (Holtzman, Steven) (Filed on 2/24/2012) (Entered: 02/24/2012)

02/24/2012 - 735 - Declaration of MEREDITH DEARBORN in Support of 734 MOTION to Strike EXCLUDE PORTIONS OF THE SUPPLEMENTAL EXPERT REPORT OF DR. ALAN J. COX filed byOracle America, Inc.. (Attachments: # 1 Exhibit A, # 2 Exhibit B, # 3 Exhibit C)(Related document(s) 734 ) (Holtzman, Steven) (Filed on 2/24/2012) (Entered: 02/24/2012)

02/24/2012 - 736 - Administrative Motion to File Under Seal PORTIONS OF ITS OPPOSITION TO GOOGLES DAUBERT MOTION filed by Oracle America, Inc.. (Attachments: # 1 Proposed Order)(Holtzman, Steven) (Filed on 2/24/2012) (Entered: 02/24/2012)

02/24/2012 - 737 - RESPONSE (re 718 MOTION to Strike Portions of Third Expert Report by Iain Cockburn and Expert Report by Steven Shugan; Memorandum of Points and Authorities in Support Thereof ) filed by Oracle America, Inc.. (Holtzman, Steven) (Filed on 2/24/2012) (Entered: 02/24/2012)

02/24/2012 - 738 - Declaration of FRED NORTON in Support of 737 Opposition/Response to Motion, filed byOracle America, Inc.. (Attachments: # 1 Exhibit A, # 2 Exhibit B, # 3 Exhibit C, # 4 Exhibit D, # 5 Exhibit E, # 6 Exhibit F, # 7 Exhibit G, # 8 Exhibit H, # 9 Exhibit I, # 10 Exhibit J)(Related document(s) 737 ) (Holtzman, Steven) (Filed on 2/24/2012) (Entered: 02/24/2012)

02/24/2012 - 739 - Declaration of IAIN M. COCKBURN in Support of 737 Opposition/Response to Motion, filed by Oracle America, Inc.. (Attachments: # 1 Exhibit A, # 2 Exhibit B)(Related document(s) 737 ) (Holtzman, Steven) (Filed on 2/24/2012) (Entered: 02/24/2012)

02/24/2012 - 740 - Declaration of STEVEN M. SHUGAN in Support of 737 Opposition/Response to Motion, filed byOracle America, Inc.. (Attachments: # 1 Exhibit A, # 2 Exhibit B)(Related document(s) 737 ) (Holtzman, Steven) (Filed on 2/24/2012) (Entered: 02/24/2012)

02/24/2012 - 741 - Declaration of PETER KESSLER in Support of 737 Opposition/Response to Motion, filed by Oracle America, Inc.. (Related document(s) 737 ) (Holtzman, Steven) (Filed on 2/24/2012) (Entered: 02/24/2012)

02/24/2012 - 742 - Declaration of CHRISTOPHER PLUMMER in Support of 737 Opposition/Response to Motion, filed by Oracle America, Inc.. (Related document(s) 737 ) (Holtzman, Steven) (Filed on 2/24/2012) (Entered: 02/24/2012)

02/24/2012 - 743 - Declaration of MARK REINHOLD in Support of 737 Opposition/Response to Motion, filed by Oracle America, Inc.. (Related document(s) 737 ) (Holtzman, Steven) (Filed on 2/24/2012) (Entered: 02/24/2012)

02/24/2012 - 744 - Declaration of JOHN R. ROSE in Support of 737 Opposition/Response to Motion, filed by Oracle America, Inc.. (Related document(s) 737 ) (Holtzman, Steven) (Filed on 2/24/2012) (Entered: 02/24/2012)

02/24/2012 - 745 - Declaration of HINKMOND WONG in Support of 737 Opposition/Response to Motion, filed by Oracle America, Inc.. (Related document(s) 737 ) (Holtzman, Steven) (Filed on 2/24/2012) (Entered: 02/24/2012)

02/24/2012 - 746 - Declaration of GEORGE SIMION in Support of 737 Opposition/Response to Motion, filed by Oracle America, Inc.. (Related document(s) 737 ) (Holtzman, Steven) (Filed on 2/24/2012) (Entered: 02/24/2012)

02/24/2012 - 747 - ERRATA re 736 Administrative Motion to File Under Seal PORTIONS OF ITS OPPOSITION TO GOOGLES DAUBERT MOTION [CORRECTED] PROPOSED ORDER by Oracle America, Inc.. (Holtzman, Steven) (Filed on 2/24/2012) (Entered: 02/24/2012)

02/24/2012 - 748 - Letter from Robert A. Van Nest . (Van Nest, Robert) (Filed on 2/24/2012) (Entered: 02/24/2012)


*******************

Documents

724

IN THE UNITED STATES DISTRICT COURT
FOR THE NORTHERN DISTRICT OF CALIFORNIA

ORACLE AMERICA, INC.,
Plaintiff,
v.
GOOGLE INC.,
Defendant.

No. C 10-03561 WHA

ORDER REGARDING
EXPERTS’ COMPENSATION

At trial, all retained experts must be prepared to testify to the amounts that they and their firms have received in compensation and expect to receive for work they did up to the end of trial as well as the billing rate for all personnel involved. Counsel shall instruct the experts to be prepared on this subject.

Counsel are reminded that they and witnesses must keep the period from April 16 until late June available for the trial.

IT IS SO ORDERED.

Dated: February 23, 2012.

/s/Williams Alsup
WILLIAM ALSUP
UNITED STATES DISTRICT JUDGE


725

IN THE UNITED STATES DISTRICT COURT
FOR THE NORTHERN DISTRICT OF CALIFORNIA

ORACLE AMERICA, INC.,
Plaintiff,
v.
GOOGLE INC.,
Defendant.

No. C 10-03561 WHA

ORDER REGARDING PATENT
MARKING JOINT STATEMENT

Counsel have filed a supplemental joint statement regarding the Oracle products that practiced the patents in suit. In this statement, they largely offer points on which they disagree. That is, while they came close to an agreement, they could not close the deal and agree on the use that the jury and Court could make of the proposed stipulations.

A stipulation requires full agreement. Until the parties reach a complete agreement on the evidentiary use of the stipulation, then it is not effective. Both sides are entitled to withhold their acceptance until they get a full loaf instead of half a loaf. No one will be held to any conditional stipulation for any purpose until it is fully agreed on as to its use by the Court and jury.

In the Court’s view, the failure to reach this stipulation will require counsel to use their allotted time to cover what, in most cases, would be easily agreed upon by reasonable counsel. The conditional stipulations will have no binding effect on anyone.

The foregoing is not true with respect to the ’702 patent and Java OS 1.1 products. This single unconditional stipulation is hereby binding on the parties.

IT IS SO ORDERED.

Dated: February 24, 2012.

/s/William Alsup
WILLIAM ALSUP
UNITED STATES DISTRICT JUDGE


726

IN THE UNITED STATES DISTRICT COURT
FOR THE NORTHERN DISTRICT OF CALIFORNIA

ORACLE AMERICA, INC.,
Plaintiff,
v.
GOOGLE INC.,
Defendant.

No. C 10-03561 WHA

ORDER REGARDING
REEXAMINATIONS

Based on the recently filed joint statement regarding reexaminations, the parties are requested to do the following. By NOON ON MARCH 9, the parties shall submit a candid discussion of the impact these rejections will have on the shape of trial. Please discuss whether in light of the track record of final rejections, it is best to delay the trial to allow the single remaining non-final reexamination to run its course.

IT IS SO ORDERED.

Dated: February 24, 2012.

/s/William Alsup
WILLIAM ALSUP
UNITED STATES DISTRICT JUDGE


729

UNITED STATES DISTRICT COURT
NORTHERN DISTRICT OF CALIFORNIA
SAN FRANCISCO DIVISION

ORACLE AMERICA, INC.
Plaintiff,
v.
GOOGLE, INC.
Defendant.

Case No. CV 10-03561 WHA

ORACLE AMERICA, INC.’S MOTION TO
STRIKE PORTIONS OF GREGORY
LEONARD’S SUPPLEMENTAL REPORT

Dept.: Courtroom 8, 19th Floor
Judge: Honorable William H. Alsup


TABLE OF CONTENTS

NOTICE OF MOTION AND MOTION TO STRIKE iii
MEMORANDUM OF POINTS AND AUTHORITIES 1
I. INTRODUCTION 1
II. ARGUMENT 2
A. Dr. Leonard’s Forward Citations Analysis Is Profoundly Flawed And
His Results From That Analysis Are Unreliable And Irrelevant
2
B. Dr. Leonard Includes Misleading And Irrelevant Calculations Based On
An Accounting Document Prepared For Oracle In 2010
5
C. Dr. Leonard’s Claim That Google’s Expected Gains Are Determinative
With Respect To The Amount Of Any Reasonable Royalty Awardable
To Oracle Is Wrong And Improper
8
IV. CONCLUSION 10

i


TABLE OF AUTHORITIES

CASES

Georgia-Pacific v. U.S. Plywood Corp.,
318 F.Supp. 1116 (S.D.N.Y. 1970)
8, 9
Golight, Inc. v. Wal-Mart Stores, Inc.,
355 F.3d 1327 (Fed. Cir. 2004)
9
Lucent Techs, Inc. v. Gateway, Inc.,
580 F.3d 1301 (Fed. Cir. 2009)
6
Monsanto Co. v. Ralph,
382 F.3d 1374 (Fed. Cir. 2004)
9
Panduit Corp v. Stahlin Bros. Fibre Works, Inc.,
575 F.2d 1152 (6th Cir. 1978)
9
Rite-Hite Corp. v. Kelley Co.,
56 F.3d 1538 (Fed. Cir. 1995)
9
Ticketmaster Corp. v. Tickets.com, Inc.,
No. CV 99-07654 HLH (VBKx), 2003 WL 25781901 (C.D. Cal. Feb. 10, 2003)
10

ii


NOTICE OF MOTION AND MOTION TO STRIKE

PLEASE TAKE NOTICE that Plaintiff Oracle America, Inc. (“Oracle”) hereby moves to exclude portions of the opinions and testimony of Google, Inc.’s (“Google’s”) patent damages expert Dr. Gregory K. Leonard. This motion is based on the following memorandum of points and authorities, the declarations of Beko Reblitz-Richardson and Prof. Iain Cockburn and accompanying exhibits, the entire record in this matter, and on such evidence as may be presented at any hearing on this Motion, on a date to be determined by the Court, as well as any other ground the Court deems just and proper.

DATED: February 24, 2012

BOIES, SCHILLER & FLEXNER LLP

/s/ Steven C. Holtzman
Steven C. Holtzman
Attorneys for Plaintiff
ORACLE AMERICA, INC

iii


MEMORANDUM OF POINTS AND AUTHORITIES

I. INTRODUCTION

Oracle requests that the Court strike three portions of the February 17, 2012 supplemental report of Google’s patent damages expert, Dr. Gregory Leonard. (See Declaration of Beko Reblitz- Richardson (“Richardson Decl.”) Exh. A (“Supplemental Leonard Report”).) As explained below, these portions contain calculations and assertions that are unreliable, irrelevant, and would serve no purpose other than to mislead and confuse the jury regarding the calculation of damages in this case.

First, Oracle requests that the Court strike Dr. Leonard’s “forward citation” patents analysis. (Leonard Report at 7.) To calculate the relative value of certain patents owned by Sun, Dr. Leonard counts the number of times those patents were cited by later-issued patents. Even assuming that forward citation analysis is a reliable valuation tool, there are two fundamental flaws in Dr. Leonard’s approach, each of which makes his results unreliable: (1) Dr. Leonard fails to account for the fact that certain patents (such as the ’104) were re-issued, and fails to include counts for the predecessor patents; and (2) Dr. Leonard fails to control for the fact that certain patents (such as the ’720) issued years after the other patents, and therefore naturally would not be cited as frequently as the older patents. A correction for the first error alone results in the ’104 patent being ranked 1st based on the total number of citations, not 11th, as Dr. Leonard erroneously claims in his report.

Second, Oracle requests that the Court strike Dr. Leonard’s testimony and calculations based on a 2010 accounting document prepared for Oracle in connection with its acquisition of Sun. (Supplemental Leonard Report at 9.) The document on which Dr. Leonard relies reflects an effort to allocate, purely for bookkeeping purposes, the specific amount that Oracle paid to acquire Sun, not measure the actual worth of any of the Sun assets, different elements of which possessed varying characteristics and investment value. The document – and thus Dr. Leonard’s calculations – bears no relationship to the value of any particular intellectual property to Sun or to Google, let alone the relative value of the patents-in-suit and the full 2006 Bundle. Dr. Leonard’s calculations are both misleading and irrelevant, and permitting him to present those calculations to the jury would be contrary to the Court’s prior rulings in this case, including the Court’s refusal to permit testimony regarding the amount of the settlement in the earlier Sun v. Microsoft litigation.

1


Third, Oracle requests that the Court strike Dr. Leonard’s suggestion that the reasonable royalty should be limited to “the value that Google was expecting to receive” from the infringement. (Supplemental Leonard Report at 2.) It is well-established that a patentee’s expected revenues and losses should be considered in calculating a reasonable royalty, and that the infringer’s expected revenue from the infringement does not cap the amount of a reasonable royalty or lost license fee. Permitting Dr. Leonard to offer contrary testimony would confuse the jury and intrude on the Court’s role in instructing the jury as to the relevant considerations for calculating a reasonable royalty.

Striking these misleading, inaccurate, and irrelevant portions of Dr. Leonard’s supplemental report is especially important given the Court’s ruling that “no deposition shall be taken of Google’s experts, Oracle may not serve a further reply report, and Oracle will not be allowed to present Dr. Cockburn as a rebuttal witness on the new materials.” (Dkt. 702 at 2.) Opportunities to depose an expert prior to cross-examination at trial and have another expert rebut those opinions at trial are important checks on the reliability and relevance of expert testimony. Here, the absence of those checks creates a greater risk of the jury being misled by Dr. Leonard, with prejudice to Oracle.

II. ARGUMENT

A. Dr. Leonard’s Forward Citations Analysis Is Profoundly Flawed And His Results
From That Analysis Are Unreliable And Irrelevant
In response to Prof. Cockburn’s group and value apportionment analysis, Dr. Leonard offers a “forward citations” analysis with respect to certain patents included in the 2006 Bundle. (See Supplemental Leonard Report at p. 7.) The citations analysis consists of counting the number of times each patent is cited by later patents and using the counts as a proxy for the value of each patent. Dr. Leonard did not previously refer to forward citations or citation counts in his previous reports or testimony in this case. Dr. Leonard’s supplemental report contains one paragraph that concerns that new analysis, which states in full:

[REDACTED]

2


[REDACTED]
(Id. (footnote omitted).)

Prof. Cockburn has worked extensively with patent counts and citations since he wrote his Ph.D. thesis in the 1980s, and has since conducted substantial research on citations analysis. (See Richardson Decl. Exh. B (Cockburn Depo. Tr. at 112:10-113:13); Declaration of Iain M. Cockburn in Support of Oracle America, Inc.’s Motion to Strike Portions of Gregory Leonard’s Supplemental Report (“Cockburn Leonard Decl.”) at ¶ 3).) As he testified in detail at his deposition, analysis of forward citations is not a useful way to measure the relative value of the specific patents-in suit in this case. In particular, Prof. Cockburn testified that

If you count those up in circumstances where it’s meaningful to do so and make adjustments where necessary, you will find often that that measure will correlate with other things. . . .

I considered it [citation analysis]. Considered it quite carefully and did not do it for a couple of reasons. One is a purely practical one, which is most of the patents in this portfolio are relatively young and that means that not enough patents have been issued subsequently to be able to have the possibility of citing these patents, especially once you take into account lags between application and issuance and so forth.

So a recent cohort of patents is going to be one which is very difficult to assess using citation analysis just because, if you like, the signal-to-noise ratio buried in the number of citations that a patent attracts is not favorable. So that’s a practical problem.

More generally there’s the question of, are these citation types of measures particularly useful at discriminating at the level of individual patents? Where they have been shown to be useful, generally speaking, is in the context of statistical studies with very large sample sizes where much of the noise, if you like, can come out in the wash or you have sufficient numbers of observations and statistical power to use the kinds of methodologies which will apply appropriate adjustments so that you can meaningfully compare one patent to another.

3


Those are the circumstances under which I think citation analysis stands a chance of working reliably. They don’t apply here.
(Richardson Decl. Exh. B (Cockburn Depo. Tr. at 112:21-114:13).) Dr. Leonard’s report does not state that he has ever conducted a citation count analysis before, nor does it contain any discussion of the methodological challenges specifically identified by Prof. Cockburn. But even if a forward citations analysis were appropriate in this case – and it is not for the reasons identified by Prof. Cockburn and academic studies1 – Dr. Leonard’s attempt at such an analysis in this case is fundamentally flawed and misleading.

Dr. Leonard does not disclose any exhibit or calculations that actually show his forward citation counts or the methods he uses to reach those numbers. However, the backup data that Google produced following submission of the report reveals that Dr. Leonard’s forward citations analysis suffers from at least two fundamental flaws, both of which render Dr. Leonard’s analysis and calculations unreliable, misleading, and inadmissible under the Daubert standard. The backup spreadsheet with his citation counts is attached as Exhibit C to the Richardson Declaration.

First, Dr. Leonard failed to account for the fact that certain patents were re-issued. For example, the predecessor to the ’104 is USRE36204E1 (applied for in November 1996 and issued in April 1999), which has 1 citation. The predecessor to USRE36204E1 is US5367685 (applied for in December 1992 and issued in November 1994), which has 73 citations. (Cockburn Leonard Decl. ¶ 5.) Dr. Leonard failed to attribute the ’204’s one citation to the ’104 and failed to count any of the 73 citations to the ’685 at all. Instead, Dr. Leonard counts only three citations for the ’104 patent. (See Richardson Decl. Exh. C.) In fact, the ’104 patent and its predecessors have been cited 77 times in all. (Cockburn Decl. ¶ 6.) Correcting Dr. Leonard’s elementary but substantial counting errors for re-issued patents changes the rank of the ’104 patent from 11th under Dr. Leonard’s approach to 1st,

___________________________________________

1 See, e.g., Bhaven Sampat & Arvids Ziedonis, “Patent Citations and the Economic Value of Patents,” Handbook of Quantitative Science and Technology Research 2005, Part 2, p. 295 (“Our primary finding is that whilst patent citations are good predictors of whether a university patent is licensed, they are not good predictors of the license revenues earned by technologies conditional upon its licensing.”).

4


with more than double the number of citations than the next patent. (Id.) The significant change in ranking among the patents highlights the significance of Dr. Leonard’s error.

Second, in counting forward citations for the top 22 patents in Dr. Cockburn’s analysis, all of which were issued (or re-issued) between April 27, 1999 and February 1, 2011 (see Richardson Exh. C), Dr. Leonard treats all citations as equal, disregarding the fact that some patents (such as the ’720) were issued many years after other patents in the group.

Citations of course take place over time, so newer patents will tend to have been cited fewer times than older patents, irrespective of value. (See Cockburn Decl. ¶ 7.) In other words, absolute citation counts are partly (and largely) a function of age, not just a function of value. Dr. Leonard makes no effort to control for the obvious effect of time. (See id.) As a result, his analysis systematically understates the value of newer patents. This is particularly relevant to the assessment of the citations for the patents-in-suit, given that the ’720 was issued on September 16, 2008, making it the second most recent patent in the list of 22, and that the ’205 patent was issued on June 16, 2005, also relatively recently. (Id. ¶¶ 9-11.)

The problem with ignoring the issue dates of the patents is obvious in Dr. Leonard’s results: all of the top six patents in Dr. Leonard’s ranking were issued in 2004 or earlier (each receiving over 10 citations), while the three bottom spots in the ranking are occupied by patents issued in 2008 or later (each receiving zero citations). Dr. Leonard includes a second ranking, but that second ranking also fails to deal with this issue. (See id. ¶ 8.)

Dr. Leonard’s failure to account in any way for the more than ten-year span covering the issuance of these 22 patents is a serious methodological flaw that has a substantial impact on the outcome of the ranking. (Id. ¶¶ 7-12.) Permitting Dr. Leonard to testify about this flawed citation analysis, particularly without rebuttal testimony by Prof. Cockburn, would mislead and confuse the jury regarding the relative importance of the patents-in-suit.

B. Dr. Leonard Includes Misleading And Irrelevant Calculations Based On An
Accounting Document Prepared For Oracle In 2010
In his new report, Dr. Leonard improperly relies on a single document, [REDACTED]

5


[REDACTED] It is not a comparable license or a valuation of comparable intellectual property, and cannot be used as an indicator of the value of the patents and copyrights that Google infringed. See Lucent Techs, Inc. v. Gateway, Inc., 580 F.3d 1301, 1325-32 (Fed. Cir. 2009).

Dr. Leonard contends that calculations based on the 2010 accounting document show that Prof. Cockburn’s damages calculations are “unreliable.” (Supplemental Leonard Report at p. 9.) The full text of the relevant paragraph is:

[REDACTED]
(Id. (emphasis in original, footnotes omitted).)

Dr. Leonard’s purported justification for invoking the $505 million to cap Oracle’s damages is mere pretext. Dr. Leonard states that he is “[u]sing Dr. Cockburn’s own methodology” for these calculations. (Id.) But Prof. Cockburn’s methodology does not rely on any accounting valuation of Oracle’s 2010 acquisition of Sun at all.

Dr. Leonard’s calculations are [REDACTED]

6


[REDACTED]

[REDACTED]

Dr. Leonard not only misconstrues the one document he relies upon; he shuts his eyes to the substantial evidence that points in the opposite direction. Sun generated hundreds of millions of

__________________________________

2 The full document is 389 pages long, but Dr. Leonard cited only these three pages in footnote 9 of his report. Oracle has submitted only those three pages with this motion, to try to limit the number of pages submitted to the Court in connection with these Daubert motions, but will of course supply a full copy of the document if the Court requests.

7


dollars in revenue each year by licensing its Java technology to third parties. When Oracle announced that it would acquire Sun for $7.4 billion, Oracle CEO Larry Ellison stated publicly that “Java is the single most important software asset we have ever acquired.” (Richardson Decl. Exh. F.) [REDACTED] Dr. Leonard ignores all of that evidence.

Dr. Leonard has plucked an irrelevant number from an irrelevant document that was prepared for an irrelevant purpose almost five years after the hypothetical negotiation. Without reference to any of the specifics of the hypothetical negotiation or any Georgia Pacific factor, he opines that a small fraction of that irrelevant number is the maximum reasonable royalty in this case. That “analysis” has no basis in either law or fact, is not a relevant comparable or benchmark, and is nothing more than an attempt to prejudice the jury by picking out a low number. Permitting Dr. Leonard to testify to the jury about these calculations would mislead the jury. Consistent with this Court’s previous order regarding the dollar figure of the Sun v. Microsoft settlement, it should be stricken.

C. Dr. Leonard’s Claim That Google’s Expected Gains Are Determinative With
Respect To The Amount Of Any Reasonable Royalty Awardable To Oracle Is
Wrong And Improper
Dr. Leonard opines in his supplemental report:

[REDACTED]
(Supplemental Leonard Report at p. 2 (emphasis added).)

8


The statement that “It is the value that Google was expecting to receive that matters for the reasonable royalty analysis” is an incorrect statement of the law of patent damages. For example, in Rite-Hite Corp. v. Kelley Co., 56 F.3d 1538 (Fed. Cir. 1995), the court ruled:

The district court here conducted the hypothetical negotiation analysis. It determined that Rite-Hite would have been willing to grant a competitor a license to use the ’847 invention only if it received a royalty of no less than one-half of the per unit profits that it was foregoing. In so determining, the court considered that the ’847 patent was a “pioneer” patent with manifest commercial success; that Rite-Hite had consistently followed a policy of exploiting its own patents, rather than licensing to competitors; and that Rite-Hite would have had to forego a large profit by granting a license to Kelley because Kelley was a strong competitor and Rite-Hite anticipated being able to sell a large number of restraints and related products. It was thus not unreasonable for the district court to find that an unwilling patentee would only license for one-half its expected lost profits and that such an amount was a reasonable royalty. The fact that the award was not based on the infringer’s profits did not make it an unreasonable award.
Id. at 1554-55 (emphasis added and citations omitted); see also Golight, Inc. v. Wal-Mart Stores, Inc., 355 F.3d 1327, 1338 (Fed. Cir. 2004) (“There is no rule that a royalty be no higher than the infringer’s net profit margin.”) (citations omitted); Monsanto Co. v. Ralph, 382 F.3d 1374, 1384 (Fed. Cir. 2004) (“[A]lthough an infringer’s anticipated profit from use of the patented invention is among the factors to be considered in determining a reasonable royalty, the law does not require that an infringer be permitted to make a profit.”) (citing Georgia-Pacific v. U.S. Plywood Corp., 318 F.Supp. 1116, 1120 (S.D.N.Y. 1970)) (internal punctuation omitted).

Here, the Court already ruled that “lost convoyed sales also remain relevant to a reasonable royalty even where the patent owner’s proof is insufficient to show lost profit.” (Dkt. 685 at 6.) Other courts have confirmed that expected sales by the licensor should be considered. See also Georgia Pacific, 318 F. Supp. at 1121 (factors include “the anticipated amount of profits that the prospective licensor reasonably thinks he would lose as a result of licensing the patent as compared to the anticipated royalty income”); Panduit Corp v. Stahlin Bros. Fibre Works, Inc., 575 F.2d 1152, 1163 (6th Cir. 1978) (“expectant loss is an element to be considered in retroactively determining a reasonable royalty”). Permitting Dr. Leonard to testify to the contrary would not only confuse the jury but also potentially create a conflict with the Court’s jury instructions. Such testimony is impermissible. See, e.g., Ticketmaster Corp. v. Tickets.com, Inc., No. CV 99-07654 HLH (VBKx),

9


2003 WL 25781901, at *1 (C.D. Cal. Feb. 10, 2003) (“the sole source of the law for the jury is the judge, not the expert witness”).

The Court should strike the paragraph quoted above from Dr. Leonard’s supplemental report and preclude any testimony or argument that only “the value that Google was expecting to receive” matters for the reasonable royalty analysis.

III. CONCLUSION

For these reasons, Oracle respectfully requests that the Court preclude Dr. Leonard from providing any testimony regarding forward citation analysis, the 2010 accounting document, or any purported limitation of a reasonable royalty or actual damages to “the value that Google was expecting to receive.” The corresponding portions of Dr. Leonard’s supplemental report should be stricken.

Dated: February 24, 2012

BOIES, SCHILLER & FLEXNER LLP

By: /s/ Steven C. Holtzman
Steven C. Holtzman

Attorneys for Plaintiff
ORACLE AMERICA, INC.

10



732

[Morrison Foerster letterhead]

February 24, 2012

The Honorable William H. Alsup
United States District Court, Northern District of California
450 Golden Gate Avenue
San Francisco, California 94102

Re: Oracle America, Inc. v. Google Inc., No. 3:10-CV-03561-WHA (N.D. Cal.)

Dear Judge Alsup:

I write in response to Google’s February 22 précis letter requesting leave for the parties to update their expert reports on the ’702 and ’205 patent validity issues to reflect the recent PTO actions.

Although Oracle maintains its objection to introducing evidence from an incomplete reexamination as being more prejudicial than probative under FRE 403, Oracle recognizes that the Court’s decision on Oracle’s Motion in Limine No. 1 is law of the case, and that the Court permitted supplementation with respect to the ’520, ’720, and ’476 patents. (ECF No. 676, at 5.)

If the Court were to grant Google’s current request, Oracle requests leave to respond with its own supplemental expert reports to address any arguments Google newly advances in its supplemental expert reports.

Respectfully submitted,

/s/ Michael A. Jacobs

Michael A. Jacobs


734

UNITED STATES DISTRICT COURT
NORTHERN DISTRICT OF CALIFORNIA
SAN FRANCISCO DIVISION

ORACLE AMERICA, INC.
Plaintiff,
v.
GOOGLE, INC.
Defendant.

Case No. CV 10-03561 WHA

ORACLE AMERICA, INC.’S NOTICE OF
MOTION AND MOTION TO EXCLUDE
PORTIONS OF THE SUPPLEMENTAL
EXPERT REPORT OF DR. ALAN J. COX

Dept.: Courtroom 8, 19th Floor
Judge: Honorable William H. Alsup


TABLE OF CONTENTS

I. INTRODUCTION 1
II. COPYRIGHT DAMAGES STANDARD 2
III. BACKGROUND 3
A. Prof. Cockburn’s Copyright Damages Calculations 3
B. Dr. Cox’s Copyright Damages Calculations 4
IV. ARGUMENT 6
A. Cox’s Changes To His Infringer’s Profits And Lost Profits Calculation
Are Impermissible Under The Court’s Order
6
B. Dr. Cox’s Application of Prof. Cockburn’s Lowest Apportionment
Percentage To Calculate Infringer’s Profits Does Not Rest On A
Reliable Foundation And Should Be Striken
8
C. Cox’s Alternative Calculation Of Infringer’s Profits Based On The
Parties’ Revenue Sharing Discussion Should Be Stricken
10
D. Cox’s Revisions To Lost Profits Do Not Rest On A Reliable
Foundation And Should Be Stricken
12
IV. CONCLUSION 13

i


TABLE OF AUTHORITIES

CASES
On Davis v. The Gap, Inc.,
246 F.3d 152 (2d Cir. 2001)
12
Polar Bear Prods., Inc. v. Timex Corp.,
384 F.3d 700 (9th Cir. 2004)
2, 3, 10, 11
Three Boys Music Corp. v. Bolton,
212 F.3d 477 (9th Cir. 2000)
3
STATUTES
17 U.S.C. § 504(b)17 U.S.C. § 504(b) 2, 3, 8, 12

ii


NOTICE OF MOTION AND MOTION TO STRIKE

PLEASE TAKE NOTICE that Plaintiff Oracle America, Inc. (“Oracle”) hereby moves to exclude portions of the opinions and testimony of Google, Inc.’s (“Google’s”) copyright damages expert Dr. Alan Cox. This motion is based on the following memorandum of points and authorities, the Declaration of Meredith Dearborn and accompanying exhibits, the entire record in this matter, and on such evidence as may be presented at any hearing on this Motion, on a date to be determined by the Court, as well as any other ground the Court deems just and proper.

DATED: February 24, 2012

BOIES, SCHILLER & FLEXNER LLP

________________________________
Steven C. Holtzman
Attorneys for Plaintiff
ORACLE AMERICA, INC

1


MEMORANDUM OF POINTS AND AUTHORITIES I. INTRODUCTION In October 2011, Google’s copyright damages expert, Dr. Alan Cox, calculated infringer’s profits at [REDACTED] and lost profits at [REDACTED] In a supplemental report served on February 17, 2012, Dr. Cox [REDACTED] and added an entirely new infringer's profits analysis that estimates recoverable profits at [REDACTED]. With his new report, Dr. Cox has thus [REDACTED]

These revised and new calculations should be stricken.

First, Dr. Cox’s revised and new infringer’s profits and lost profits calculations are impermissible under the Court’s prior orders. On January 9, 2012, the Court struck Prof. Cockburn’s reasonable royalty and lost license fee calculations, but did not strike his infringer’s profits or lost profits calculations. (Dkt. 685 at 9.) On January 20, 2012, the Court entered an order permitting Prof. Cockburn to revise the stricken calculations and update his discussion of the econometric and conjoint studies. (Dkt. 702 at 2–3.) The Court permitted Google to serve revised damages reports in response, but stated: “Only revisions directly responsive to new material by Dr. Cockburn will be allowed.” (Id.) Prof. Cockburn did not revise his opinions as to infringer’s profits, lost profits, or the conjoint analysis. Dr. Cox’s revisions to his infringer’s profits and lost profits calculations, and his entirely new infringer’s profits calculation, are all beyond the scope of what the Court permitted, and therefore should be stricken.

Second, Dr. Cox’s new calculations should be stricken because they are fundamentally flawed. For infringer’s profits, the copyright statute requires Dr. Cox to measure the portion of Android’s real world profits that are attributable to factors other than the copyrighted APIs. Dr. Cox does not perform that analysis. He does not measure the contributions of the other elements of Android at all. Instead, he simply assumes that 5.1% of Android’s profits would be attributable to Google’s copyright infringement. This number comes from the lowest bound of Prof. Cockburn’s hypothetical license fee determination, in which Prof. Cockburn determined that at least 5.1% of the value of the 2006 license that Sun offered Google was attributable to the copyrights in suit. Dr. Cox

1


applies the same 5.1% figure to reduce Oracle’s lost profits damages. But the Court has previously held that there is no relationship between the value of the Bundle and the value of actual Android profits (or Oracle’s actual lost profits) attributable to the infringed copyrights. (See Dkt. 685 at 7–9.) Indeed, Google previously accused Prof. Cockburn of mixing apples and oranges by using data showing the actual value to Android from 2011 to apportion the value of the 2006 Bundle. Dr. Cox has now done the same thing, only in reverse: by applying a 2006 apportionment to profits and losses through 2011, he has compared oranges to apples. His analysis must be stricken.

Dr. Cox’s unauthorized and illogical revisions are particularly pernicious given that the Court has indicated that it will not permit Oracle to depose Dr. Cox about his new opinions or permit Prof. Cockburn to rebut them at trial. (Dkt. 702 at 2.) Dr. Cox’s opinions that violate the Court’s order will be highly prejudicial to Oracle and particularly misleading to the jury.

Oracle respectfully requests that the Court strike Sections VIII.B and VIII.D from Dr. Cox’s supplemental report, which contain Dr. Cox’s improper and flawed revisions to his infringer’s profits and lost-profits calculations.

II. COPYRIGHT DAMAGES STANDARD

The copyright damages statute provides, in full:

(b) Actual Damages and Profits. — The copyright owner is entitled to recover the actual damages suffered by him or her as a result of the infringement, and any profits of the infringer that are attributable to the infringement and are not taken into account in computing the actual damages. In establishing the infringer’s profits, the copyright owner is required to present proof only of the infringer’s gross revenue, and the infringer is required to prove his or her deductible expenses and the elements of profit attributable to factors other than the copyrighted work.
17 U.S.C. § 504(b).

“Congress explicitly provides for two distinct monetary remedies—actual damages and recovery of wrongful profits. These remedies are two sides of the damages coin—the copyright holder’s losses and the infringer’s gains.” Polar Bear Prods., Inc. v. Timex Corp., 384 F.3d 700, 707–08 (9th Cir. 2004). “Actual damages are usually determined by the loss in the fair market value of the copyright, measured by the profits lost due to the infringement or by the value of the use of the copyrighted work to the infringer.” Id. (quotation marks and citations omitted).

2


This Court recognized this important distinction between actual damages and infringer’s profits in its November 2011 ruling on Oracle’s Daubert motion concerning the original reports of Google’s damages experts. In that ruling, the Court held that non-infringing alternatives are irrelevant to recovery of wrongful profits. (Dkt. 632 at 7.) The Court wrote:

Not acceptable, however, is allowing the existence of non-infringing alternatives to reduce recovery of wrongful profits. This is a distinct remedy for the purpose of disgorgement. Non-infringing alternatives have nothing to do with this. The motion to strike portions of Dr. Cox’s report that opined that non-infringing alternatives would provide a basis for calculating wrongful profit is GRANTED.
(Id. (emphasis in original).)

The infringer’s profits remedy eliminates incentives for would-be infringers, and prevents “the infringer from unfairly benefitting from a wrongful act.” Polar Bear, 384 F.3d at 708 (quotation marks and citation omitted). Pursuant to this policy, Congress expressly provided that apportionment of infringer’s profits is the defendant’s burden. 17 U.S.C. § 504(b). Oracle “is required to present proof only of the infringer’s gross revenue,” id., and then Google “bears the burden of apportioning the profits that were not the result of infringement.” Polar Bear, 384 F.3d at 711. Where there is “imprecision in the computation of expenses, a court should err on the side of guaranteeing the plaintiff a full recovery.” Three Boys Music Corp. v. Bolton, 212 F.3d 477, 487 (9th Cir. 2000) (quotation marks and citation omitted).

III. BACKGROUND

A. Prof. Cockburn’s Copyright Damages Calculations
In his September 2011 report, Prof. Cockburn calculated Google’s gross Android revenue attributable to its infringement to be [REDACTED] through the end of 2011. This calculation was based on his review of evidence demonstrating the importance of the copyrighted APIs to attracting Java developers and the importance of those developers to Android’s success. Any further apportionment of that [REDACTED] was and is Google’s burden under the Copyright Act.

Prof. Cockburn also calculated lost profits in two ways. [REDACTED]

3


[REDACTED] He calculated that Oracle’s combined lost profits from these two sources are $136.2 million.

In his February 2012 report, Prof. Cockburn did not revise either the lost profits or the infringer’s profits calculation. (See Cockburn Report ¶¶ 621–49, 650, 671–94, Ex. 20–22.)1 This was consistent with the Court’s orders. In its previous Daubert order striking portions of Prof. Cockburn’s damages analysis, the Court specifically noted that “calculations of copyright unjust enrichment and copyright lost profits are not stricken.” (Dkt. 685 at 13.) The Court permitted Prof. Cockburn only to revise “only those items stricken by the recent order” as well as the conjoint and econometrics analyses. (Dkt. 702 at 2–3.) Neither Prof. Cockburn nor Prof. Shugan revised any portion of the conjoint analysis; instead, Prof. Cockburn now uses that analysis in a manner that is consistent with the Court’s orders. (See Cockburn Report ¶¶ 470–79 (discussing conjoint analysis); id. ¶¶ 416–20 (describing current use of the conjoint analysis); id. Exh. 5.)

B. Dr. Cox’s Copyright Damages Calculations
On October 3, 2011, Google served a copyright damages expert report signed by Dr. Cox. On October 24, 2011, Google served an errata list with a revised Dr. Cox report. In that revised report, Dr. Cox responded to Prof. Cockburn’s infringer’s profits calculation and opined that [REDACTED] in Android revenue is profit attributable to the copyrights-in-suit. (Declaration of Meredith Dearborn (“Dearborn Decl.”) Ex. A (10/21/2011 Cox Report) at pp. 38–42 and Ex. 3a.) Dr. Cox also included alternative calculations in which he opined that [REDACTED] (Id. Ex. 2a, 2b.) Dr. Cox also opined that [REDACTED] (Id. pp. 58.)

Dr. Cox has never offered an affirmative infringer’s profits analysis. In his October 2011 report, Dr. Cox relied—“for the sake of argument” only, despite the fact that Google has the affirmative burden of proving apportionment—on the conjoint analysis conducted by Dr. Shugan. (10/21/2011 Cox Report at p. 38, Ex. 3a).) Dr. Shugan’s conjoint survey showed that consumers

__________________________________

1 All references to the Cockburn Report refer to the February 9, 2012 revision to the February 3, 2012 report, which was submitted to the Court.

4


place significant value on the number of applications available, and, if fewer applications were available, Android’s market share would have declined by at least 8% to 19%, depending on certain assumptions about the but-for world. (See Cockburn Report ¶ 653.) Ignoring the fact that Dr. Shugan’s measurements represented a floor, not a ceiling, and also ignoring the fact that Dr. Shugan measured the incremental benefit of Google’s copyright infringement over the next best alternative, not the total benefit Google derived from its infringement, Dr. Cox nonetheless borrowed that figure, [REDACTED] and applied that adjusted figure to his calculation of Google’s Android revenues. Dr. Cox did not perform any independent analysis that would enable him to apportion Google’s infringer’s profits.

On February 17, 2012, in “response” to Prof. Cockburn’s February 3 report, Google served a supplemental damages report by Dr. Cox that significantly changes his analysis as to both infringer’s profits and lost profits. (Dearborn Decl. Ex. B (Supplemental Expert Report of Dr. Alan J. Cox) (“Supp. Cox Report”)) ¶¶ 29–44, Ex. 2a, 2b, 4a.)

On infringer’s profits, Dr. Cox substantially reduces the figure that was included in his October report. He includes two alternative calculations.

First, Dr. Cox, again “for the sake of argument,” uses the lower end of the range of Prof. Cockburn’s “group and value” copyright apportionment percentage (5.1%) and applies that percentage to all of Google’s Android revenues, which ultimately yields infringer’s profits of [REDACTED]” [sic]. (Id. ¶¶ 31, 33.)

Second, Dr. Cox treats Sun’s 2006 proposal to license its intellectual property to Google in exchange for a 10% revenue share (plus a license fee) as establishing that no more than 10% of real world Android revenues could be attributable to Sun intellectual property. (Id. ¶¶ 37–39.) Dr. Cox assumes that only 5.1% of that 10% (or 0.51% of the total) is attributable to the asserted copyrights. (Id. ¶¶ 39–41.) Thus, he multiplies his estimate of Android’s revenues by 0.51%, and arrives at an amount of [REDACTED]. (Id. ¶¶ 37–42.) In this alternate approach, Dr. Cox artificially limits the value of the transaction to the 10% revenue share, disregarding the facts that the 2006 proposal he relies on (a) would have provided Sun an additional $60 million in license fees, (b) included a grant to Sun of rights to Google intellectual property, (c) would have required Google to promote the Sun

5


brand for the mobility market, and (d) would have allowed Sun to earn hundreds of millions of dollars from convoyed sales.

Although both of Dr. Cox’s calculations rely on Prof. Cockburn’s “group and value” lower bound copyright apportionment percentage, Dr. Cox concedes that percentage expressly applies only to the 2006 Bundle—not real world Android. (See id. ¶ 36.)

Dr. Cox also revises his lost profits figures, applying the same lower-bound “group and value” 5.1% figure with no explanation as to why it would be appropriate to use that number. (Id. ¶ 44, Ex. 4a.) His new opinion would award [REDACTED] a substantial reduction from his previous opinion.

IV. ARGUMENT

A. Cox’s Changes To His Infringer’s Profits And Lost Profits Calculation
Are Impermissible Under The Court’s Order
In its order permitting Prof. Cockburn to file a third report, the Court limited the scope of any reports filed by Google’s experts: “Only revisions directly responsive to new material by Dr. Cockburn will be allowed.” (Dkt. 702 at 2.) That limitation was important because the Court was seeking to “streamline” this process. (Id.) Further, the Court’s order limited Oracle’s ability to challenge any revised reports filed by Google’s experts, stating “no deposition shall be taken of Google’s experts, Oracle may not serve a further reply report, and Oracle will not be allowed to present Dr. Cockburn as a rebuttal witness on the new material.” (Id.) These limitations made it critically important that Google’s experts not stray beyond the scope of Prof. Cockburn’s revised report, because Oracle would not have a full opportunity to test any new opinions or revised calculations.

Dr. Cox’s supplemental report violates the Court’s order and prejudices Oracle. The Court did not permit Prof. Cockburn to make any changes to his infringer’s profits or lost profits calculations, and Prof. Cockburn did not make any changes to those calculations. (See Dkt. 685 at 12–13; Dkt. 702 at 2–3.) Neither Prof. Cockburn nor Prof. Shugan revised the conjoint analysis. Nonetheless, Dr. Cox’s supplemental report includes significant changes to his previous infringer’s profits and lost profits calculations. As described above, those changes [REDACTED]

6


[REDACTED]

Dr. Cox’s arguments for why it was appropriate to make these changes are meritless. Dr. Cox seeks to bypass the Court’s order by stating: [REDACTED] (Dearborn Dec. Ex. B (Supp. Cox Report) ¶ 31).) This is nonsense. Prof. Cockburn revised his apportionment analysis only for the purpose of calculating a reasonable royalty and lost license fee, and only as to the portion of the 2006 license bundle that Sun offered Google properly attributable to the intellectual property in suit. Prof. Cockburn offered no affirmative opinion at all as to the percentage of Android’s profits attributable to the copyrighted APIs in suit, other than to opine that the copyrighted APIs are a material and indeed very significant element in Android’s success. Any more is Google’s burden. As discussed below, conflating the 2006 license bundle and Android’s actual profits is precisely what the Court has previously rejected. Prof. Cockburn does not have a “copyright apportionment analysis” in the sense in which Dr. Cox is attempting to use that term, and he made no revision to any infringer’s profits analysis that warrants Dr. Cox’s changes.

Dr. Cox also asserts that [REDACTED] (Dearborn Decl. Ex. B (Supp. Cox Report) ¶ 36; see also id. ¶ 44 (claiming that revisions to lost profits are appropriate because [REDACTED] This is both false and inconsistent with what Google itself repeatedly argues in its Daubert motion— filed precisely the same day—regarding Prof. Cockburn’s third report. (See Dkt. 720 at 3, 14–15 (arguing that Prof. Cockburn “relies” on the conjoint analysis in multiple respects).) Prof. Cockburn continues to use the conjoint analysis. (See Cockburn Report ¶¶ 411–12, 416–20, 459, Exs. 5–11, 35–36; see also Dearborn Decl. Ex. C (Excerpts from the 2/10/12 Deposition of Iain M. Cockburn) at 30:15–23, 31:1–32:21.) Prof. Cockburn did not “back away” from the conjoint analysis at all, much less in any way that justifies Dr. Cox’s modifications to his infringer’s profits or lost profits calculations. Dr. Shugan’s conjoint analysis has not been changed in any respect at all.

7


But even if Prof. Cockburn completely abandoned conjoint analysis as a basis for his opinions—or if Google succeeded in its efforts to exclude —that would not justify Dr. Cox’s belated introduction of a completely new method of measuring infringer’s profits. It has always been Google’s burden to apportion infringer’s profits. Google decided it would meet that burden by relying on evidence that it simultaneously argues should never be admitted, and if admitted, should never be believed by a jury. Now Google faces the consequence of its own cynical tactics, but it cannot blame that consequence on Prof. Cockburn.

Dr. Cox also offers a “different way” to calculate infringer’s profits in his supplemental report, which never appeared in his prior report and which is in no way “directly responsive to new material by Dr. Cockburn.” (Dkt. 702 at 2.) Dr. Cox begins from the license the parties were actually negotiating in 2006, and he then suggests that the revenue share portion of that license (10%) was an “implicit proposal” that 90% of revenues would be “available to Google to recover all its costs and cover any profit attributable to factors other than Sun’s IP.” (Dearborn Decl. Ex. B (Supp. Cox Report) ¶ 37–39). The only step Prof. Cox takes that is borrowed from Prof. Cockburn’s new report is to apply the 5.1% lower-bound “group and value” apportionment figure to the 10% of revenues Sun would have earned. (Id. ¶ 41.) With or without the use of that apportionment figure, this is an entirely new framework by which Dr. Cox purports to calculate infringer’s profits. It is fatally flawed, as described below, but it is also a naked attempt to shoehorn a wholly new analysis into what should have been a narrowly cabined rebuttal report. Like his “revised” infringer’s profits analysis, this wholly new “different” approach should be stricken as procedurally improper, particularly in light of the fact that Oracle may not depose Dr. Cox or rebut this wholly new opinion at trial.

B. Dr. Cox’s Application of Prof. Cockburn’s Lowest Apportionment
Percentage To Calculate Infringer’s Profits Does Not Rest On A Reliable
Foundation And Should Be Stricken
Google’s burden for apportioning infringer’s profits is straightforward: it must prove “the elements of profit attributable to factors other than the copyrighted work.” 17 U.S.C. § 504(b). Dr. Cox did not try to do that. Instead, he assumed that the percentage of Android’s real world profits

8


attributable to Google’s infringement of copyrighted API specifications must be the same as the percentage of the 2006 Bundle attributable to copyrighted API specifications.

When the Court struck the apportionment analysis in Prof. Cockburn’s second damages report, it held that Prof. Cockburn erroneously “failed to account for [the] disconnect” between the universe of know-how included in Sun’s 2006 offer of a license and the universe of know-how included in Android during 2008–2011. (Dkt. 685 at 8.) The Court concluded that Prof. Cockburn’s apportionment of the 2006 Bundle had an “apples-and-oranges problem” because the 2006 Bundle “does not bear any relationship” to 2008–11 Android revenues, “for all the record shows.” (Id.)

Dr. Cox’s revised calculations are impermissible for the same reasons articulated in the Court’s order striking Prof. Cockburn’s second report. Dr. Cox borrows the lowest possible bound of the apportionment percentage from Prof. Cockburn’s analysis of the 2006 license bundle—5.1%, out of a range of 5.1% to 16.4% (see Cockburn Report ¶ 420)—and applies that apportionment percentage to actual Android revenues through 2011 to reduce his infringer’s profits calculation to [REDACTED] (Dearborn Decl. Ex. B (Supp. Cox Report) ¶ 33.) This is not appropriate. The Court has already rejected the notion that there is a direct correlation between the relative value of the copyrights to the 2006 license bundle, on the one hand, and the relative value of the copyrights to revenues generated for Google by Android, on the other hand. (Dkt. 685 at 7–9.) Dr. Cox just ignores the Court’s ruling.

In fact, Dr. Cox admits that his use of Prof. Cockburn’s apportionment percentage is inappropriate. Dr. Cox writes:

[REDACTED]
(Dearborn Decl. Ex. B (Supp. Cox Report) ¶ 36) (emphasis in original). Dr. Cox does not explain why his use of Prof. Cockburn’s apportionment analysis is scientifically or logically sound at all. Instead, he only claims, without any analysis or explanation, that his use of those percentages [REDACTED] (Id. ¶ 32.) This is, of course, the same

9


reasoning that the Court previously found inadequate . (See Dkt. 685 at 8 (“It is no answer to say, as Dr. Cockburn did in his footnote 327, that his approach is ‘conservative.’”).)

Even if Dr. Cox’s use of the 5.1 percentage figure were analytically sound—and it is not—it could not be conservative. Prof. Cockburn, through the group-and-value approach, measured the relative value of the copyrights and the patents as compared to the other elements of the 2006 license bundle. Google’s burden with respect to infringer’s profits, by contrast, is to present some methodology by which to measure the absolute value that the copyrights provided to Google’s advertising revenues, to avoid full disgorgement. Anything less would not accomplish the disgorgement aims of the statute. See Polar Bear, 384 F.3d at 708. Moreover, Prof. Cockburn determined that the proper apportionment percentage of the 2006 Bundle for the copyrights could be 16.4% or higher. (Cockburn Report ¶ 420.) Dr. Cox has not explained why the lower bound of Prof. Cockburn’s range is appropriate to use in the context of his infringer’s profits analysis.

C. Cox’s Alternative Calculation Of Infringer’s Profits Based On The
Parties’ Revenue Sharing Discussion Should Be Stricken
Dr. Cox’s new and “different way” to calculate Google’s profits attributable to the infringement fares no better. In his alternative calculation, Dr. Cox begins with the [REDACTED] in revenues that he claims Android had earned as of September 2011. Once again, he fails to prove the elements of profit attributable to factors other than the copyrighted work. Instead, he constructs the following argument: In 2006, Sun proposed to license its technology to Google in exchange for a 10% royalty. (Dearborn Decl. Ex. B (Supp. Cox Report) ¶ 39.) According to Dr. Cox, that establishes that no more than 10% of real world Android revenues could be attributable to Sun intellectual property. (Id.) Dr. Cox once again relies on Prof. Cockburn’s 5.1% lower bound apportionment figure to assert that no more than 5.1% of the 10% of real world Android revenues (0.51%) could be attributable to Google’s copyright infringement. Ergo, reasons Dr. Cox, 99.49% of Android’s real world profits are attributable to factors other than the copyrighted Java API specifications. (Id.) Thus, he multiplies his estimate of Android’s revenues by 0.51%, and arrives at an amount of [REDACTED]. (Id. ¶¶ 37–42.)

This method is even more profoundly flawed than Dr. Cox’s first method. No court has ever

10


approved of a defendant using a previously negotiated royalty rate to meet its burden to apportion the profits it subsequently earned through infringement.

First, this approach suffers from the same problem that undermines his other calculation, as explained above. The fact that Prof. Cockburn concludes that, at the lowest bound, the copyrighted APIs represented at least 5.1% of the 2006 license bundle does not permit Dr. Cox to conclude that the same apportionment percentage applies to Android’s real world revenues.

Second, Dr. Cox’s analysis cannot meet Google’s burden under Section 504(b). Whatever amounts the parties hypothetically might have agreed on in 2006, that ex ante division of value does not establish the amount by which Google subsequently actually profited from Android’s infringement of the Java APIs.

Third, using a royalty rate negotiated ex ante to apportion infringer’s profits will systematically understate copyright damages. The amount that Google rationally would have been willing to pay Sun for a license necessarily will be less than the amount Google expected to profit from that same license. Of course, the point of the infringer’s profits remedy is to deter infringement by forcing the infringer to disgorge whatever gains it earned as a result of the infringement. Polar Bear, 384 F.3d at 707–08 (discussing “infringer’s gains”). Dr. Cox’s approach would subvert that rule. A willful infringer would bargain for the lowest possible royalty, break off negotiations once favorable terms were proposed, and then deliberately infringe. If the infringer avoided getting caught, it would pay nothing. If it were caught, it would argue—just as Google does now—that the infringer’s profits remedy should be no greater than what Google was prepared to pay in the first place. This is even more brazen than the “Soviet-style negotiation” Google previously advocated, and the Court rejected. (Dkt. 230 at 10.)

Fourth, even if Dr. Cox’s second approach were sound—and it is not—his use of the 10% revenue share to cap the value of the copyrights is nonsensical. The 2006 proposal would have provided Sun with substantial value above and beyond a 10% share of Android revenues, including a license fee of $60 million for the first three years, rights to Google intellectual property, the advantage of having Google promote the Java brand in the mobile market, a compatible Android implementation, and hundreds of millions of dollars from convoyed sales.

11


Dr. Cox’s alternate approach to infringer’s profits is unsound factually, legally, and logically, and it should be stricken.

D. Cox’s Revisions To Lost Profits Do Not Rest On A Reliable Foundation
And Should Be Stricken
Dr. Cox has revised his lost profits analysis to apply the same 5.1% apportionment percentage, derived from Prof. Cockburn’s group-and-value apportionment analysis. (Dearborn Decl. Ex. B (Supp. Cox Report) ¶ 44).) This is unsound for the same reasons described above. Dr. Cox does not explain why it is appropriate to apply a figure derived from a group-and-value apportionment exercise to Sun’s lost profits, and there is no basis to do so. On this ground alone, as well as the fact that Prof. Cockburn made no revisions to his lost profits analysis and Dr. Cox therefore may not either, the Court should strike Dr. Cox’s supplemental lost profits opinions.

Moreover, the Copyright Act does not provide for any apportionment of lost profits at all. See 17 U.S.C. § 504(b). Prof. Cockburn already accounted for the degree to which Google’s infringement caused Sun, and now Oracle, harm in calculating the lost-profits figures shown in his Report. (Cockburn Report ¶¶ 671–94.) [REDACTED]

The Copyright Act provides for “actual damages” to account for the copyright holder’s actual losses. Lost profits are lost revenue less costs avoided. See On Davis v. The Gap, Inc., 246 F.3d 152, 164–65 (2d Cir. 2001). Prof. Cockburn already deducted avoided costs. (See Cockburn Report Exs. 20 & 21.) Oracle will show at trial that the entirety of the remaining losses are attributable to Android’s infringement. Google may argue otherwise, but Dr. Cox should not be permitted to present a misleading and unreliable calculation that mechanically applies an inapposite 5.1%

12


apportionment figure to Oracle’s lost profits with no explanation at all as to why that figure bears any relationship to the harm caused by Android’s infringement.

V. CONCLUSION

For all these reasons, the revisions in Dr. Cox’s supplemental report to the infringer’s profits and lost profits figures should be stricken, and Dr. Cox should not be permitted to testify regarding them.

Dated: February 24, 2012

BOIES, SCHILLER & FLEXNER LLP

By: /s/ Steven C. Holtzman
Steven C. Holtzman

Attorneys for Plaintiff
ORACLE AMERICA, INC.


12


737

UNITED STATES DISTRICT COURT
NORTHERN DISTRICT OF CALIFORNIA
SAN FRANCISCO DIVISION

ORACLE AMERICA, INC.
Plaintiff,
v.
GOOGLE, INC.
Defendant.

Case No. CV 10-03561 WHA

ORACLE AMERICA, INC.’S OPPOSITION
TO GOOGLE’S MOTION TO STRIKE
PORTIONS OF THIRD EXPERT REPORT
BY IAIN COCKBURN AND EXPERT
REPORT BY STEVEN SHUGAN

Dept.: Courtroom 8, 19th Floor
Judge: Honorable William H. Alsup


TABLE OF CONTENTS

I. INTRODUCTION 1
II. ARGUMENT 3
A. Prof. Cockburn’s “Independent Significance” Analysis Is Legally And
Factually Sound
3
B. Prof. Cockburn’s “Group And Value” Analysis Is Reasonable And
Admissible
7
1. Prof. Cockburn’s Consideration Of The Work Performed By
Oracle Engineers Is Appropriate
8
2. Prof. Cockburn’s Consideration Of Industry Studies Reflecting
Patent Value Distributions Is Appropriate
11
C. Prof. Cockburn Correctly Apportions The Value Of Copyrights In The
2006 Bundle
15
D. Prof. Cockburn Correctly Considered And Accounted For Claim-By-
Claim Differences Within The Portfolio
17
E. Professor Shugan’s Conjoint Analysis Is Methodologically Sound 19
1. Conjoint Analysis Is An Appropriate Tool In A Hypothetical
License Analysis
19
2. Prof. Shugan’s Conjoint Survey Was Methodologically Sound 20
F. Prof. Cockburn’s Econometric Analysis Is Methodologically Sound 23
1. Prof. Cockburn’s Econometric Analysis Is Based On Reliable
Data
23
2. Prof. Cockburn’s Analysis Is Based On Reasonable
Assumptions
24
III. CONCLUSION 25

i


TABLE OF AUTHORITIES

CASES
Arista Records LLC v. Lime Group, LLC,
06 CV5936KMW, 2011 WL 1674796 (S.D.N.Y. May 2, 2011)
15, 24
Boucher v. U.S. Suzuki Motor Corp.,
73 F.3d 18 (2d Cir. 1996)
24
Cornell Univ. v. Hewlett-Packard Co.,
609 F. Supp. 2d 279 (N.D.N.Y. 2009)
19
Daubert v. Merrell Dow Pharms., Inc.,
509 U.S. 579 (1993)
8, 20, 23
Domingo v. T.K.,
289 F.3d 600 (9th Cir. 2002)
24
Faulkner v. Gibbs,
199 F.2d 635 (9th Cir. 1952)
1, 4
Finjan, Inc. v. Secure Computing Corp.,
626 F.3d 1197 (Fed. Cir. 2011)
2, 4, 5, 6
General Electric Co. v. Joiner,
522 U.S. 136 (1997)
14
Hangarter v. Provident Life & Acc. Ins. Co.,
373 F.3d 998 (9th Cir. 2004)
11
i4i Ltd. P'ship v. Microsoft Corp.,
598 F.3d 831 (Fed. Cir. 2010), aff'd, 131 S. Ct. 2238 (2011)
19
In re Katz Interactive Call Processing Patent Litig.,
639 F.3d 1303 (Fed. Cir. 2011)
19
In re Phenylpropanolamine (PPA) Prods. Liab. Litig.,
289 F. Supp. 2d 1230 (W.D. Wash. 2003)
24
Johnson Elec. N. Am. v. Mabuchi Motor Am. Corp,
103 F. Supp. 2d 268 (S.D.N.Y. 2000)
23
Kennedy v. Collagen Corp.,
161 F.3d 1226 (9th Cir. 1998)
13, 23
LG Display Co., Ltd. v. AU Optronics Corp.,
722 F. Supp. 2d 466 (D. Del. 2010)
8, 11, 13

ii


Lucent Technologies, Inc. v. Gateway, Inc.,
580 F.3d 1301 (Fed. Cir. 2009)
4, 6, 7, 19
McLaughlin v. American Tobacco Co.,
522 F.3d 215 (2d Cir. 2008)
20
Medical Instr. & Diagnostics Corp. v. Elekta AB,
No. 97-CV-02271, Dkt. No. 464
24
Ruiz-Troche v. Pepsi Cola of Puerto Rico Bottling Co.,
161 F.3d 77 (1st Cir. 1998)
21
U.S. Gypsum Co. v. Lafarge N. Am. Inc.,
670 F. Supp. 2d 737 (N.D. Ill. 2009)
24
Uniloc USA, Inc. v. Microsoft Corp.,
632 F.3d 1292 (Fed. Cir. 2011)
2, 4, 6
Veritas Operating Corp. v. Microsoft Corp.,
No. C06-0703-JCC, 2008 WL 7404617 (W.D. Wash. Feb. 26, 2008)
4

iii


I. INTRODUCTION

In his revised damages report (“Report”), Prof. Cockburn applies two new apportionment methods, both grounded in this Court’s Orders and established case law. In addition, he provides further explanation of his approach to claim-by-claim and future damages, and minor revisions to his econometric analysis. Other than that, as required by the Court, his report is unchanged.

Based on the new apportionment methods, and after accounting for any reductions for failure to mark or for the limitation to only accused products, Prof. Cockburn calculates a range of minimum reasonable royalty damages for the patent claims ($17.7 to $57.1 million) and the copyrighted API claims ($34.7 to $111.9 million). (Report ¶¶ 5-6.) The “starting point” for those calculations is the combination of Sun’s $100 million offer and the $557 million that Sun expected to earn over three years from a compatible Android, both of which the Court previously ruled could be used to calculate damages based on a hypothetical negotiation. (Dkt. 685 at 4-7.) Prof. Cockburn’s copyright lost profits and infringer’s profits calculations and analyses are unchanged.

In its entire brief, Google cites only eight cases (not one of which supports the points Google advances), an oral argument transcript from a different case, and a losing in limine brief that Google mischaracterizes as a federal court decision. Google claims that Prof. Cockburn and Prof. Shugan failed to perform analyses that they did perform and expressly documented. Google makes factual assertions, and then quotes testimony that negates those assertions. Google speculates that there might be facts that contradict Prof. Cockburn’s analyses, but offers no evidence to support its musings. Nothing in Google’s brief justifies exclusion of any expert opinion in this case.

Google first attacks Prof. Cockburn’s “independent significance” approach to apportioning the value that would have been included in the 2006 transaction (the “2006 Bundle”). In that approach, Prof. Cockburn considers the available evidence concerning the attributes provided by the patents and copyrights and the importance of those attributes to Google as of 2006. Google’s attack has two prongs:

  • Google contends Prof. Cockburn’s conclusion that the patents in suit account for at least 25% of the value of the 2006 Bundle is impermissibly subjective because the figure is not based on a “quantitative algorithm or formula.” But “[t]here is no mathematical formula for the determination of a reasonable royalty.” Faulkner v. Gibbs, 199 F.2d 635, 639 (9th Cir. 1952). Prof. Cockburn’s approach is consistent with the Federal Circuit decision in Finjan,

1


    Inc. v. Secure Computing Corp., 626 F.3d 1197 (Fed. Cir. 2011). Google ignores Finjan.

  • Google argues that Prof. Cockburn’s use of 25% is an improper attempt to rely on the “25% rule of thumb” royalty rate that the Federal Circuit rejected in Uniloc. But this argument ignores that the independent significance approach does not calculate a royalty rate; it simply apportions the 2006 Bundle. Moreover, Prof. Cockburn relies on no rule of thumb; he examines the specific evidence in this case to determine proper apportionment percentages.

Google next attacks Prof. Cockburn’s “group and value” approach to apportionment. In that approach, Prof. Cockburn identifies, values, and accounts for all of the elements of the 2006 Bundle other than the patents and copyrights. He then relies on work done by Oracle’s Chief Architect for the Java Platform, Dr. Mark Reinhold, and other Java engineers to identify the patents that would be expected to contribute the greatest technical benefits to a smartphone platform in 2006. Numerous studies of patent value establish that most of the value of a group of patents is attributable to a very small number of patents in the group. Those studies calculate curves that allow one to estimate the percentage of portfolio value attributable to the most significant patents in the portfolio. Prof. Cockburn uses the engineers’ conclusions that three of the asserted patents are among the most technically important 4% of the 2006 portfolio to apportion economic value according to these curves. Google’s arguments in response are unfounded:

  • Google argues that the four engineers supporting Dr. Reinhold were biased by their previous analyses of Sun’s patents for litigation purposes. As this Court has previously made clear, the claim of bias is a cross-examination point, not a Daubert argument. But there was no bias. Google fails to mention that the engineers testified under oath that their previous analyses covered dozens of patents, not just the patents-in-suit.
  • Google argues that the engineers’ analysis is too vague because it does not rank each of the 569 patents individually. But false precision does not make for good evidence or good engineering analysis. Prof. Cockburn appropriately deals with this by having an upper bound and a lower bound measurement of the value of the patents-in-suit.
  • Google speculates that the portfolio of patents identified by the engineers may have a different value distribution than the patents in the studies that Prof. Cockburn cites. But Google cannot say whether the distribution really is different, whether any difference is material, or whether any difference would increase or decrease damages. Prof. Cockburn has made clear that the distribution of patent value in the studies – across countries, across industries, and within specific product groups – is the same as for the Sun portfolio and is consistent with his own observations of single company portfolios.
  • Google argues that Prof. Cockburn fails to subtract the full value of copyrighted source code (not in suit) that Sun would have conferred in the 2006 Bundle. Google is wrong. Prof. Cockburn identifies all of the relevant copyrighted material. For copyrighted source code not in suit, he calculates the value to Google of avoiding the expense of writing that code itself. With respect to code implementations, there is no basis for attributing greater value than the

2


    cost of writing them.

  • Google argues that Prof. Cockburn failed to account for unasserted claims within the asserted patents. But the Court’s requirement that Oracle separately calculate damages for the asserted claims was for purposes of determining a verdict if some asserted claims were more valuable than others, or if some claims were later found to be invalid. Prof. Cockburn was not required to calculate a value for every claim, asserted and unasserted, of a patent, particularly when neither real nor hypothetical negotiators in 2006 would have done so.

Google also argues that Prof. Shugan’s conjoint analysis should be excluded. Google claims that conjoint analysis is a market research tool, not a way to calculate damages. Google’s argument that only methods specifically developed for litigation are admissible, while those that are accepted by academics and professionals as rigorous and reliable are not, contradicts Daubert itself. Market surveys are routinely used in litigation to calculate damages, and the Federal Circuit has specifically approved their use to show what parties would have expected at the time of a hypothetical negotiation. Moreover, conjoint analysis has repeatedly been used in litigation, and many experts, including Google’s own copyright damages expert, Dr. Cox, have concluded that it is a proper way to calculate infringement damages. None of Google’s specific attacks on Prof. Shugan’s analysis here have any merit. Even if they did, those attacks are grounds for cross-examination, not exclusion.

Finally, Google attacks Prof. Cockburn’s econometric analysis. That study uses eBay data from smartphone auctions to assess the value that consumers would place on the performance benefits provided by the patents-in-suit (as shown by the benchmarking analyses). Google asserts that eBay purchasers are unrepresentative, without explaining how or why, while ignoring Prof Cockburn’s explanation why this is not so. Google also asserts that if Android phones were slower, their price would drop, mitigating the effect on its own market share. But Google never explains how it would dictate phone prices to the OEMs and carriers who must build and subsidize those mediocre phones. Moreover, Google’s assertion concedes that the performance enhancement provided by the patents in suit is valuable to consumers. Google’s arguments provide no basis for exclusion.

II. ARGUMENT

A. Prof. Cockburn’s “Independent Significance” Analysis Is Legally And Factually Sound
Prof. Cockburn’s independent significance approach evaluates (a) the specific attributes provided by the patents and copyrights in suit, (b) Google documents that demonstrate the importance

3


to Google of those specific features and functionality, and (c) quantitative measures of the performance and economic benefits that the parties would have expected from the infringement, as demonstrated by benchmark performance tests and the input of technical experts and generally supported by the conjoint survey and econometric data. (Report ¶¶ 421, 459, Exs. 6-11.) Based on all of that evidence, Prof. Cockburn concludes that at least 25% of the value of the 2006 Bundle should be attributed to the patents in suit, and at least 12.5% should be attributed to the copyrighted API specifications in suit. (Id. ¶¶ 5-6, 423, 670.)

Google does not dispute that the speed, memory and applications provided by the patents and copyrights were and are important to Android. Instead, Google argues that Prof. Cockburn’s quantification of the value of the property in suit is nothing more than a “subjective guess” derived “without using any quantitative algorithm or formula” and that Prof. Cockburn’s apportionment analysis runs afoul of the Federal Circuit’s rejection of the “25% rule of thumb” for a reasonable royalty in Uniloc. (Dkt. 718 at 1, 4-7.)

Like almost all of the arguments in its motion, Google’s insistence that a reasonable royalty calculation must be based on some “quantitative algorithm or formula” (Dkt. 718 at 1) rests on nothing but Google’s own say-so. Google cites no legal authority in support of its argument. To the contrary, the Ninth Circuit has held that “[t]here is no mathematical formula for the determination of a reasonable royalty.” Faulkner, 199 F.2d at 639; see also Lucent Tech., Inc. v. Gateway, Inc., 580 F.3d 1301, 1325 (Fed. Cir. 2009) (hypothetical negotiation “necessarily involves an element of approximation and uncertainty”) (internal quotation and citation omitted); Veritas Operating Corp. v. Microsoft Corp., No. C06-0703-JCC, 2008 WL 7404617, at *5 (W.D. Wash. Feb. 26, 2008) (denying motion to exclude expert damages; reasonable royalty calculation does not require “mathematical formula” or “mathematical precision”).

The Federal Circuit’s decision in Finjan, Inc. v. Secure Computing Corp., 626 F.3d 1197 (Fed. Cir. 2011) squarely supports Prof. Cockburn’s independent significance analysis. In Finjan, the jury found that defendant Secure Computing infringed three of Finjan’s software patents related to “proactive scanning” (techniques for defending against computer viruses), and awarded reasonable royalty damages based on the testimony of Finjan’s expert, Russell Parr. Id. at 1201. On appeal,

4


Secure challenged the damages award on numerous grounds, including the sufficiency of Dr. Parr’s testimony under Georgia Pacific factors 10 and 13, which concern “the nature of the patented invention” and “the portion of the realizable profit that should be credited to the invention.” Id. at 1211. Secure argued that Dr. Parr failed to fully account for the other software modules included in the accused products that also contributed to their profits. Id. The Federal Circuit disagreed:

Parr testified that, based on Secure’s internal documents, proactive scanning was “fundamentally important to the product,” while Degen [Secure’s expert] agreed after reviewing Secure and Finjan promotional materials that “[i]t’s important technology. It was perceived as the next wave.” Those materials included statements emphasizing Secure’s ability to “proactively protect” customers with “Webwasher Proactive Scanning.” From this evidence, the jury could infer that a substantial fraction of the accused products’ profits stemmed from proactive scanning.
Id. (internal record citations omitted). Dr. Parr offered no algorithm or formula for his apportionment of profits and no quantitative evidence, but nonetheless testified regarding a specific range of appropriate apportionment values. The Federal Circuit held that the jury was “entitled to hear the expert testimony” and affirmed the damages award based on that testimony. Id. at 1212. Google never addresses Finjan, though the Court and Oracle both have previously cited it.

Prof. Cockburn’s conclusion here considers and relies upon far more evidence than Dr. Parr did in concluding that Finjan’s patents accounted for a “substantial fraction” of the value of Secure’s virus detection software. Like Dr. Parr, Prof. Cockburn considers Google’s contemporaneous documents, which repeatedly emphasize the critical importance of the speed and memory that the asserted patents provide. (Report ¶¶ 63, 441-46, 448, 450.) Google does not try to rebut that evidence, but rather argues that speed, memory and applications are “obvious features” (Dkt. 718 at 1). But calling features “obvious” does not make them unimportant – indeed, they may be obvious because they are so critical to the commercial success of a smartphone.

Prof. Cockburn also considers the cost of memory and hardware needed to compensate for a non-infringing device’s larger footprint (Report ¶¶ 432, 453-55; Exh. 16) and benchmarking evidence showing the substantial incremental performance improvements provided by the patented technology. (Id. ¶¶ 425–38.) In addition, he considers the results of the conjoint survey and econometric analyses of consumer preferences as general quantitative indicators of how important those features are. (Id.

5


Exs. 6-11.)1 Because all of this data is precisely the type of information that could have been obtained at the time of the hypothetical negotiation (Report ¶¶ 426, 458), it is properly considered to establish a royalty. Lucent, 580 F.3d at 1333–34; see also Dkt. 718 at 15-16 (Google concession that conjoint analysis is frequently used when “designing or launching a new product”).)

In short, nothing in Prof. Cockburn’s testimony states or suggests, as Google contends, that Prof. Cockburn “made up [the 25%] number because it felt right to him.” (Id. at 6.) Instead, even the testimony Google quotes shows that his quantification is based on objective evidence, not a “subjective judgment,” as Google asserts. (Dkt. 718 at 7.)

Finjan analysis is a “disguised resurrection” of the “25% rule of thumb” rejected in Uniloc USA, Inc. v. Microsoft Corp., 632 F.3d 1292 (Fed. Cir. 2011). In support of its argument, Google relies only on coincidence: Prof. Cockburn’s patent apportionment percentage is 25%, and the “rule of thumb” was a 25% royalty. Google’s attempt to relate the two is off base. First, What Google describes as an impermissible “subjective” analysis is in fact the approach that the Federal Circuit approved in Finjan. Second, in Uniloc the court rejected the rule of thumb not because it was “subjective,” but because it was a constant, arbitrary value unrelated to the facts of the case, see 632 F.3d at 1313-18. In contrast, the independent significance approach rests upon consideration of the specific evidence in this case, not any generic rule of thumb. Third, the approach is an apportionment analysis of the percentage of the 2006 Bundle attributable to the patents and copyrights in suit, not the calculation of a royalty rate. It has nothing at all to do with the 25% rule of thumb at issue in Uniloc.

Google’s claim Prof. Cockburn “admitted at deposition that he included the caveat ‘at least’ to preserve his flexibility to argue a much higher percentage to the jury” (Dkt. 718 at 1) is also false. Google pressed Prof. Cockburn to identify an upper limit for the apportionment percentage for the IP in suit, and he stated that the percentages could be substantially greater (see id. at 6-7 (quoting

_________________________________

1 Google incorrectly claims that Dr. Cockburn does not discuss the conjoint or econometric analyses in the independent significance section of his report. (Dkt. 718 at 5.) In fact, paragraph 459 of that section makes clear that the evidence on which he relies is summarized in Exhibits 6-11 of his report, which refer to the results of the conjoint and econometrics analyses.

6


testimony)). That testimony in no way undermines his damages calculations, which are based on the minimum apportionment percentages that he derived based on his review of the evidence, and which make no adjustment for fragmentation or the litigation premium. Google’s claim that Prof. Cockburn’s analysis cannot be “replicated” (id. at 7) is also factually and legally wrong. Google engineers could have conducted performance tests on Android phones; Google could have hired a survey expert; Google’s experts could have conducted an econometric study; and Google’s experts could, like Prof. Cockburn, consider the evidence. But in any event, because expert valuation of intellectual property, unlike a chemistry experiment, “necessarily involves an element of approximation and uncertainty,” Lucent, 580 F.3d at 1325, its admissibility cannot hinge on whether it can be “replicated.” See Kumho Tire Co., Ltd. v. Carmichael, 526 U.S. 137, 150-51 (1999) (expert testimony that relies on “personal knowledge or experience,” such as valuation of property, may not be susceptible to scientific standards of reliability); American General Life Ins. Co. v. Schoenthal Family, LLC, 555 F.3d 1331, 1338 (11th Cir. 2009) (“Standards of scientific reliability, such as testability and peer review, do not apply to all forms of expert testimony.”).

B. Prof. Cockburn’s “Group And Value” Analysis Is Reasonable And Admissible
In the “group and value” approach, Prof. Cockburn identifies the elements of the 2006 Bundle and subtracts the value of the elements other than the patents and copyrights in suit. (Report ¶¶ 361- 90.) To help apportion value between the patents in suit and other patents, Dr. Cockburn asked Dr. Reinhold to evaluate what Sun patents would have been included in the 2006 Bundle and which of those patents would have been expected to contribute the greatest technical benefit to a smartphone platform. (Id. ¶¶ 391-97.) Prof. Cockburn considers studies showing that the distribution of value within a group of patents is highly skewed; that is, a very small number of patents will account for a very large percentage of the value of a portfolio. Dr. Cockburn applies the engineers’ ranking of the technical benefits provided by the patents in Sun’s portfolio to the distribution curves described by the studies, and calculates an apportionment range on that basis. (Id. ¶¶ 403-15.) This method carefully applies established economic principles to the facts of this case. Indeed, allocating value within a company’s patent portfolio based on evidence of the asserted patents’ relative value and established distribution curves has been accepted by other courts. See LG Display Co., Ltd. v. AU

7


Optronics Corp., 722 F. Supp. 2d 466 (D. Del. 2010).

Google challenges Prof. Cockburn’s (a) reliance on Oracle engineers to select and rank the patents that likely would have been included in the 2006 Bundle; and (b) method of evaluating the relative economic value of the patents-in-suit based on studies concerning the distribution of value among patents. (Dkt. 718 at 7-11.) Neither challenge warrants exclusion under Daubert. The core of Google’s argument is an unreasonable and insupportable notion of what is required for expert analysis to survive attack under the Daubert standard. Google first takes Dr. Reinhold to task for not “offer[ing] even a guess as to the ranking [of each individual patent] among these top 22,” (Dkt. 718 at 9), and then criticizes Prof. Cockburn for applying a well-accepted economic rationale for attributing value to the patents group-ranked by the engineers. Google would have this Court require a standard of false precision that lies beyond the ability of engineering and economics and still further beyond the gate-keeping requirements of the law.

1. Prof. Cockburn’s Consideration Of The Work Performed By Oracle Engineers Is
Appropriate
Google badly mischaracterizes the work that Oracle’s Java engineers performed to assist Dr. Cockburn. That work is described in detail in the engineers’ depositions and in declarations that each has submitted.2 All five engineers were Sun employees with over 70 years combined experience in overlapping elements of the Java language, Java virtual machines, and Java virtual machines on small embedded devices. All five are the inventors of Java-related patents (none of which are in suit) that would have been included in the 2006 portfolio. All five are deeply experienced in evaluating and implementing inventions related to the performance of Java virtual machines. ((Reinhold Decl. ¶¶ 4- 7; Rose Decl. ¶¶ 3-6; Wong Decl. ¶¶ 3-7; Kessler Decl. ¶¶ 3-8; Plummer Decl. ¶¶ 3-7.) One was directly involved in the 2006 negotiations with Google. (Wong Decl. ¶ 5; Wong Tr. 23:24-25:18.) Dr. Kessler has also served on Sun’s and Oracle’s Patent Review Committee, in which he assesses

___________________________________

2 Excerpts of the deposition transcripts of the five Oracle engineers are attached to the Norton declaration as Exhibits F (Dr. Mark Reinhold), G (John Rose), H (Hinkmond Wong), D (Peter Kessler), and E (Christopher Plummer). We refer to those deposition transcripts and declarations in this brief by last name (e.g., “Reinhold Tr.” and “Reinhold Decl.”).

8


the technical merit of many Java-related inventions. (Kessler Decl. ¶ 6.)3

The engineers began by applying their experience, as well as the Android Product Requirements Document that Google gave Sun in 2006, to identify 22 “technology blocks” covered by Sun’s Java patents in 2006 relevant to a smartphone at that time. (See, e.g., Reinhold Decl. ¶ 10.) They then ranked those 22 groups, in terms of their importance to a smartphone platform, by considering their importance to startup time, speed, memory and security – all criteria they regularly use to judge Sun’s and Oracle’s own Java implementations. (Id. ¶ 19.)

The engineers also reviewed a deliberately over-inclusive list of over 1300 patents issued to Sun prior to June 30, 2006 that included the term “Java” or “bytecode” anywhere in the patent or listed James Gosling or Nedim Fresko as an inventor. (Simion Decl. ¶¶ 3-9.) The engineers were already familiar with the overwhelming majority of these patented inventions, as they were (or knew) the inventors, were directly or indirectly involved in the implementations of those inventions, [REDACTED] (Reinhold Decl. ¶¶ 4-7; Rose Decl. ¶¶ 4-8; Wong Decl. ¶¶ 4-7; Kessler Decl. ¶¶ 4-8; Plummer Decl. ¶¶ 4-7; see also footnote 3.) By reviewing the titles, abstracts, inventors, application dates, and where necessary, the specifications and claims, the engineers were able to evaluate which patents would have been relevant to the parties’ 2006 smartphone platform negotiations, and to assign the relevant patents to the appropriate technology blocks. (See, e.g., Reinhold Decl. ¶ 17; Kessler Decl. ¶ 9.) The engineers confirmed the accuracy of their initial categorization of the patents by reviewing the patents both together and individually in categories in which they had special expertise. (Reinhold Decl. ¶ 17.) This process left 569 potentially relevant patents.

The engineers rated the technical importance of each of the 569 patents on a 3-point scale, considering the patent’s expected contribution to a smartphone platform’s startup, speed, or footprint. (Id. ¶ 21.) The engineers could provide this ranking in a matter of days because the scale was

____________________________________

3 The engineers each detailed the breadth of their knowledge and experience during their depositions. (Reinhold Tr. 6:4-8:10, 23:13-23, 48:3-49:16; Rose Tr. 12:22-13:18, 26:22-28:9, 28:18-24, 77:15- 78:11, 80:9-14; Wong Tr. 13:1-12, 15:13-17:22, 39:13-19, 64:17-65:8, 66:17-67:6, 69:4-70:6, 145:4- 146:6, 147:3-148:11, 153:15-154:8; Kessler Tr. 22:2-22, 23:19-24:4, 37:20-38:6; Plummer Tr. 10:10- 11:2, 16:18-17:10, 35:9-36:15, 96:2-19, 97:11-98:5, 99:14-101:17.)

9


manageable, they were already familiar with the patented inventions based on years of experience, and they had already spent an intensive week examining and discussing the patents. (Reinhold Decl. ¶ 22; Wong Decl. ¶¶ 11-21; Plummer Decl. ¶¶ 11-20; Kessler Decl. ¶¶ 12-21; Rose Decl. ¶¶ 12-21; Wong Tr. 64:11-66:6, 146:18-148:21; Rose Tr. 77:15-78:11; Plummer Tr. 97:11-98:5.) Each patent’s rating was independent of the importance of the technology block to which the patent was assigned. (Reinhold Decl. ¶ 19.) By identifying the patents and blocks rated most highly, the engineers were able to identify the 22 most important patents of the 569. Three out of six (now five) of the asserted patents – the ’104, the ’720, and the ’205 – were among the top 22 patents, comprised of patents rated a “1” on the individual rating scale and assigned to one of the three top-ranked technology blocks.

Google argues that the four engineers who assisted Dr. Reinhold were “biased” [REDACTED] ) There is simply no factual basis for any assertion of bias in favor of the patents that happen to be at issue in this one lawsuit.

Even were there a factual basis for a claim of bias, it is not cognizable in this motion. Google’s own damages experts in this case rely heavily on party employees for information, and the Court has already ruled that they may do so provided that the foundational facts are established at trial. (Dkt. 632 at 2-3 (“If Oracle is worried about bias, then it should make its arguments on crossexamination.”).) The same reasoning applies here.

Google also suggests that the engineers had too little time to rate the patents, and should have conducted “quantitative testing” of all 569. (Dkt. 718 at 7-9.) Each of the engineers has rejected that contention in sworn testimony. (Reinhold Tr. 41:23-42:11; Wong Decl. ¶ 24; Plummer Decl. ¶ 23; Kessler Decl. ¶ 24; Rose Decl. ¶ 24.) Google ignores that testimony, as well as the engineers’

10


testimony that they relied on years of experience with implementations of the patents and longstanding familiarity with many of the inventions to inform their technical assessment of the engineering benefits of the patents in the portfolio. Google’s arguments at most go to the weight of the evidence, not admissibility. See Hangarter v. Provident Life & Acc. Ins. Co., 373 F.3d 998, 1017 n.14 (9th Cir. 2004) (factual basis for expert opinion goes to credibility of testimony, not admissibility; opposing party may examine basis for opinion in cross-examination).

2. Prof. Cockburn’s Consideration Of Industry Studies Reflecting Patent Value
Distributions Is Appropriate
Based on the engineers’ identification of the 22 technically most significant patents that would have been included in the 2006 Bundle, Prof. Cockburn applied published studies that examine the distribution of patent value to estimate the percentage of the total portfolio value that would have been attributed to those top patents. (Report ¶¶ 403-12 & Exs. 34-35.)

This approach was approved in LG Display Co., Ltd. V. AU Optronics Corp., 722 F. Supp. 2d 466 (D. Del. 2010). There, AU Optronics (AUO) asserted that LG Display infringed four patents. AUO’s damages expert, Jonathan Putnam, reviewed industry portfolio licensing practices to determine the amount that LG Display would have paid AUO for its entire portfolio. To determine the value of the four specific patents in suit, Putnam assumed that those four patents were in the top 5% of AUO’s patent portfolio and calculated their likely value based on a distribution curve from a single paper that examined patent values for the entire electronics industry. (Norton Decl. Exh. A (Putnam declaration describing analysis) at 7 n.4, 8.) Putnam’s distribution curve showed that the average patent in the top 5% was worth 29 times more than the average patent in the bottom 95%, and on that basis he calculated a royalty for the four asserted patents. (Id. at 7–8.) The court found Putnam’s “testimony and methodology to be credible and consistent with Federal Circuit case law and the Georgia-Pacific factors, despite LGD’s assertions to the contrary,” and awarded damages consistent with that testimony. LG Display, 722 F. Supp. 2d at 472.

Prof. Cockburn’s analysis is more rigorous. He does not simply assume the importance of the patents in suit; he relies on the informed assessments of experienced Java engineers. He does not use a single distribution curve; he considers three such curves, based on detailed academic studies across

11


countries, industries, and products. He does not simply assume that the Java patent portfolio at issue here would share a similar distribution; he has the engineers’ own assessment, which confirms that a relatively small number of the patents would have the greatest technical value, while the majority would have little importance to a smartphone platform. He also relies on his own experience having observed the same highly skewed value distribution of single-company patent portfolios. (Report ¶ 404; Norton Decl. Exh. B (Cockburn Tr. 107:20-110:19); Cockburn Decl. ¶¶ 4-12.)

Google contends that the three studies cited by Prof. Cockburn have “nothing to do with the Sun portfolio at issue” and “are likely to have different distributions of value than the Sun portfolio.” (Dkt. 718 9-10). Notably, Google does not challenge the conclusion that the distribution of value in the Sun portfolio would be highly skewed, a fact that has been observed, documented, and studied for decades. (See Cockburn Decl. ¶¶ 4-6.) Instead, Google speculates – without a shred of analysis or factual support – that superficial differences between the relevant Sun portfolio and the patents in the studies are “critical” and make it “likely” that the degree of skewness in the Sun portfolio might be different from the studies. (Dkt. 718 at 10.) Lacking any support for this conjecture, Google does not even try to argue that the curves really would be different, that the supposed difference would be material, or that accounting for the supposed difference would actually decrease damages.

First, Google argues that none of the three studies examined distribution of patent value within a single company’s portfolio. (Dkt. 718 at 10.) But Prof. Cockburn specifically explained at his deposition why this is not a meaningful distinction:

As I’ve said a few times, it’s striking that these types of studies done for many different sets of patents and in many different contexts using many different methodologies all point to a conclusion which I don’t think is controversial, which is that the value distribution is highly skewed. So based upon that, it is my opinion that I have no reason to believe that the 569 patents of interest here would have a value distribution which is any less skewed than that which has been found so many times in so many different circumstances. . . .

I’m very familiar myself with looking at portfolios of patents held by specific companies or organizations that I’ve done research projects on which show this similar degree of skewness. For example, a number of years ago, had a doctoral student who was studying patents licensed or offered for license by M.I.T. so he could see the entire portfolio of inventions and worked on – he knew what the licensing payments were, he was able to come up with a valuation based upon a calculation or a projection of licensing revenue. And I think that -- I can’t recall the precise number

12


from memory, but I think that portfolio displayed exactly the property that a handful of patents of the order of 1 or 2 percent constituted 50 percent of the economic value.

More recently I worked on research projects again looking at licensing data collected for all the patents, several thousand patents in the portfolio which is held by two large academic medical research centers who prefer to remain nameless but are not a million miles from here, and I've looked closely at the data on their licensing. And again they display the same phenomenon: that a large number of the patents are essentially valueless by the metric of are they able to attract licensing revenue. And even among those licensed, a handful constitute the vast majority of the value and have been licensed or sold for very large sums of money. (Norton Decl. Exh B (Cockburn Tr. 107:20-110:19).)

Dr. Cockburn’s conclusion that single-company portfolios exhibit the same distribution of value as larger populations of patents is supported both by LG Display, 722 F. Supp. 2d at 472 (approving apportionment analysis based on patent value distribution for entire electronics industry) and the academic literature (Cockburn Decl. ¶¶ 7-12 & Exh. B at 560 (article observing similar distribution of values for Harvard patent portfolio).) Prof. Cockburn’s explanation of his reasoning is more than adequate. See Kennedy v. Collagen Corp., 161 F.3d 1226, 1230 (9th Cir. 1998) (refusing to exclude testimony where expert had “set forth the steps he took in arriving at his conclusion at his deposition”). Google provides no affirmative basis on which the Court could conclude otherwise.

Second, Google argues that the 569 patents are confined to software patents for smartphone functionality, unlike the studies cited by Prof. Cockburn. (Dkt. 718 at 10.) Again, Prof. Cockburn explained why the Sun portfolio at issue here would be expected to show the same distribution as other patent populations. That conclusion is supported by the published research, which confirms the same highly skewed distributions for patents limited to specific product areas. (Cockburn Decl. ¶ 8.) Google provides nothing to show the distinction makes any difference at all, much less a “critical” one that “means” the distribution here differs from the studies Prof. Cockburn cites.

Third, Google complains that the 569 patents were “selected deliberately for relevance to this case,” whereas the cited studies looked at randomly selected patents. (Dkt. 718 at 10.) But the purpose of the “selection” was to identify the entire set of patents that would have been relevant to a smartphone in 2006, based on objective indicia of relevance. (Reinhold Decl. ¶¶ 17-18.) Google cites no example of any patent that was improperly excluded.

13


Fourth, Google notes that two of the studies cited by Prof. Cockburn relied on European patents, whereas one relied on US patents. (Dkt. 718 at 10.) Google tries to argue that “this is not a trivial difference” because one of the studies observed that citation counts – not distributions of value – look very different for German patents than US patents. (Id.) On that basis, Google claims that “the text of one of [Prof. Cockburn’s] source studies refutes” the assumption that the value distribution is comparable. (Id.) Differences between European and US citation practices have nothing at all to do with whether there is any difference in the distribution of patent values. Moreover, as the authors of the study explain immediately after the words Google quotes – but Google omits from its brief – “applicants at the EPO are not required to supply a full list of prior art” so they cite far fewer patents than US applicants. (Zimmer Decl. Exh. C (Harhoff study) at 1355.)

In fact, the PatVal survey cited by Prof. Cockburn finds distribution curves to be consistent across countries. (Cockburn Decl. Exh. A at 45.) To the extent that there is a difference between US and European patents, the Barney study Prof. Cockburn cites suggests that the value of US patents is even more skewed, which would tend to increase damages. (Report Exh. 35 (showing damages ranges for each study’s curve).) Other studies demonstrate that distribution curves are similar across a sample of German, US, and Harvard patents. (Cockburn Decl. Exh. B at 560.)

Google neither addresses Prof. Cockburn’s explanations, examines the extensive relevant literature, nor provides any analysis that calls Prof. Cockburn’s conclusions into question. Instead, it says Prof. Cockburn’s reasoning must be rejected as mere ipse dixit, in conflict with General Electric Co. v. Joiner, 522 U.S. 136, 146 (1997). As Joiner held and Google concedes, “experts commonly extrapolate from existing data.” 522 U.S. at 146.4 Numerous decisions confirm that it is reasonable

___________________________________

4 Joiner is of no help to Google. In that case, the Court held that the district court was within its discretion in excluding expert testimony that the plaintiff’s small cell lung cancer was caused by his exposure to small quantities of PCBs, rather than his history of smoking. The plaintiff’s expert relied on inapposite studies that found that (a) infant mice injected directly with massive doses of PCBs developed a different kind of cancer; (b) factory workers exposed to PCBs had an elevated incidence of cancer, but there was no basis to find a causal connection; (c) factory workers exposed to PCBs had a slightly elevated incidence of cancer, but again made no finding of a causal connection; (d) a study of Norwegian workers exposed to mineral oil, not PCBs, and (e) a study that found a significant increase in cancer among Japanese workers exposed to PCBs, but those workers had been exposed to many other carcinogens as well, including “toxic rice oil they had ingested.” Joiner, 522 U.S. at 145-46.

14


for experts to rely on industry studies when calculating damages, provided they apply their judgment and analysis. See, e.g., Arista Records LLC v. Lime Group, LLC, 06 CV5936KMW, 2011 WL 1674796, at *11 (S.D.N.Y. May 2, 2011) (the “fact that these conclusions are based, in part, upon a review of surveys and the relevant literature in the area does not render [expert’s] testimony inadmissible, as long as [he] bases his conclusions on his own expertise and analysis where he considered both survey evidence, relevant literature, and his expertise to calculate damages”). Prof. Cockburn uses these studies to inform his analysis and conclusions, and applies his expert judgment and analysis in doing so. While Google would preclude the application of expertise to interpret preexisting studies, such application is precisely what the law both contemplates and requires.

C. Prof. Cockburn Correctly Apportions The Value Of Copyrights In The 2006 Bundle
Prof. Cockburn accounts for the value of the copyrights in the 2006 Bundle in three ways. First, he begins with the draft contract between the parties, which limits the grant of intellectual property to technology that would be specific to the Android smartphone platform. [REDACTED] Both calculations likely overstate the cost of Google’s independent development. Third, Prof. Cockburn values the copyrighted API specifications, using the independent significance approach. Google’s claim that Prof. Cockburn “failed to make any attempt to value all of the copyrights that would have been part of the 2006 intellectual property package” (Dkt. 718 at 11) is false.

Google attacks Prof. Cockburn’s analysis by claiming that he “admitted in his deposition that

15


did not even know what Java-related copyrights Sun owned in 2006” (Dkt. 718 at 12). The deposition testimony Google quotes does not say what Google claims: it simply recites the irrelevant point that Prof. Cockburn does not know precisely how many copyrights Sun owned for all of Java. Prof. Cockburn certainly does know, and analyzes, what is relevant: what copyrighted material, including both the copyrighted APIs in suit and the source code implementations (of both API libraries and the virtual machine) on which Google now focuses, would have been included in the 2006 Bundle and provided value to Google. (See Report ¶¶ 365-82.)

Google erroneously claims that Prof. Cockburn “made no systematic effort to measure the value of the millions of lines of code in the API libraries that would have been part of the 2006 bundle” (Dkt. 718 at 13). Prof. Cockburn did account for the value of that copyrighted code. He quantified the cost incurred by Google to write the code, which it might have otherwise obtained as part of the 2006 Bundle. As explained below, Prof. Cockburn’s evaluation of those costs tracks the value that Sun and Google would have assigned to that aspect of the 2006 Bundle. Prof. Cockburn conducted the same analysis with respect to the development of the Dalvik virtual machine. Google has not identified any other source code that the parties would have valued as part of the 2006 Bundle or that would not have been accounted for in Sun’s estimate of $86.15 million in engineering costs.

Google complains that Prof. Cockburn “never had anyone from Oracle examine the code libraries to determine their value in relation to the API specifications.” (Dkt. 718 at 13.) But Prof. Cockburn had no need to do so. In calculations Google does not challenge, Prof. Cockburn uses the amounts Google paid contractors to write source code for the Java class libraries to calculate the value of all of the Sun class library source code to Google. No further analysis is necessary to account for the value of the code. Google’s claim that Prof. Cockburn “avoids any specific valuation of those copyrighted materials at all” (Dkt. 718 at 13) ignores the plain language of the report.

Finally, Google attacks Prof. Cockburn for using Sun’s projected future R&D costs “as a proxy for the value of Sun’s then-existing intellectual property (the copyrighted class libraries and source code).” (Dkt. 718 at 13.) Here again, Google misrepresents Prof. Cockburn’s analysis. [REDACTED]

16


[REDACTED] Prof. Cockburn’s analysis also obviates Google’s arguments about cost versus value and “apples and oranges.” (Dkt. 718 at 13.) Google could have written source code without infringing, and Google would not place a value on Sun’s code that was greater than the cost to Google of doing the work itself. Accordingly, the value of that code is no greater than Google’s saved costs.

In any event, as both Prof. Cockburn and Google's copyright expert, Dr. Astrachan, explain, Google derived value by including Oracle’s copyrighted API packages to attract the Java developer community by giving them a familiar development environment. (Report ¶¶ 635-36 (quoting Dr. Astrachan).) The Android website boasts that, “Android includes a set of core libraries that provides most of the functionality available in the core libraries of the Java programming language.” (Id. at 479.) Copying Oracle’s source code implementation of the core libraries (API packages) was not needed to create that familiar development environment because, as Dr. Astrachan points out, application developers do not care how an API is implemented, as long as it works properly. (Astrachan Report at p. 56.) [REDACTED]

D. Prof. Cockburn Correctly Considered And Accounted For Claim-By-Claim Differences
Within The Portfolio
Prof. Cockburn’s revised report considers and addresses the issue of calculating damages on a claim-by-claim basis. (Report ¶¶ 493-533.) Based on input from Prof. Mitchell, Prof. Cockburn explains how Oracle asserts the same infringement theory for every asserted claim of each patent, and concludes that each of the asserted claims in any of the patents-in-suit should account for the full value associated with each of the patents-in-suit. (Id. ¶¶ 494-95.) Prof. Cockburn’s report does not fail to apportion damages on a claim-by-claim basis, but rather explains, based on the evidence and

17


input from Prof. Mitchell, how damages should be calculated if the jury finds Google liable on some but not all asserted claims. Notably, Google’s own experts have never argued that damages for one claim of a patent differ from other claims of that patent. (Report ¶ 496.)

Google now claims that Prof. Cockburn’s apportionment analysis is flawed because he did not “attribute any value to any of the unasserted claims of the patents-in-suit.” (Dkt. 718 at 14 (emphasis added).) But neither this Court (nor any other) has ever required such an analysis, and it makes no sense to do so here. The Court ruled that it was necessary to value the individual asserted claims for a number of reasons, including the possibility that a jury might find infringement as to some but not all of the asserted claims. (Dkt. 685 (1/9/12 Order) at 9-10.)

Reducing the value of the patents-in-suit based on the presence of some unasserted claims makes no sense where the relevant inquiry is one of apportionment – what proportion of the 2006 Bundle value the parties would have assigned to those patents. Parties to an actual licensing negotiation would not have evaluated every claim – certainly Google and Sun never negotiated over individual claims – and there is no reason to do so now. (Report ¶¶ 336-37.)

The granular analysis demanded by Google would introduce false precision without generating more reliable results. The Oracle engineers evaluated each of the 569 patents that might have been included in the 2006 Bundle by considering, among other things, the patent abstracts and descriptions, which describe what invention is covered by each of the patents. To the extent asserted claims deliver the benefits promised by the abstract and description – which they do – they are entitled to the full value of the patent. No reduction is necessary to account for any unasserted claims, each of which would be expected to deliver the same benefits promised by the abstracts.

Google suggests that Oracle’s decision to narrow the number of claims in suit somehow proves that Prof. Cockburn’s analysis is flawed, because Oracle “cannot be heard to argue that those [no longer asserted] claims have no value to Google.” (Dkt. 718 at 14.) But it is not Oracle’s or Prof. Cockburn’s position that those claims had no value to Google. Rather, it is Prof. Cockburn’s opinion that the claims would have no additional value to Google, above and beyond that of the asserted claims. The evidence indicates that the different claims each address the same functionality.

Finally, requiring some additional reduction to account for the value of unasserted claims

18


would be particularly problematic in this case, where Oracle has been required to drop claims to narrow this case for trial. If Oracle is forced to reduce damages for the remaining claims based on the existence of unasserted claims, that would mean that those claims presented unique issues as to damages, and the Court’s case management orders would be constitutionally suspect. In re Katz Interactive Call Processing Patent Litig., 639 F.3d 1303, 1312-13 (Fed. Cir. 2011).

E. Professor Shugan’s Conjoint Analysis Is Methodologically Sound
Conjoint analysis is an accepted method of valuing individual product attributes using a consumer survey. (Report ¶¶ 471; 662; Shugan Report p. 5-6.) By considering multiple attributes jointly, survey participants make the implicit tradeoffs one would make in real world purchasing decisions. Comparing respondents’ choices when presented with different feature sets allows for the estimation of the relative importance of specific features and their effect on market shares.

1. Conjoint Analysis Is An Appropriate Tool In A Hypothetical License Analysis
Google initially attacks the use of conjoint analysis based on the overbroad argument that such analysis should never be admissible to calculate damages in litigation. (Dkt. 718 at 15.) But it is settled law that survey-based studies are admissible for just such a purpose. In Lucent, the Federal Circuit held that an expert may rely on surveys to calculate damages:

Usage (or similar) data may provide information that the parties would frequently have estimated during the negotiation. . . . Such data might, depending on the case, come from sales projections based on past sales, consumer surveys, focus group testing, and other sources. . . . This quantitative information, assuming it meets admissibility requirements, ought to be given its proper weight, as determined by the circumstances of each case.
Lucent Technologies, Inc. v. Gateway, Inc., 580 F.3d 1301, 1333–34 (Fed. Cir. 2009) (emphasis added). A conjoint study is precisely the kind of analysis that hypothetical parties could have used during the 2006 negotiations. The Federal Circuit has also approved of survey evidence to estimate infringing use, i4i Ltd. P'ship v. Microsoft Corp., 598 F.3d 831, 856 (Fed. Cir. 2010), aff'd, 131 S. Ct. 2238 (2011); ee also Cornell Univ. v. Hewlett-Packard Co., 609 F. Supp. 2d 279, 289 (N.D.N.Y. 2009) (Rader, J.) (faulting plaintiff for failing to produce “customer surveys” that would have backed “predictive claims” that patent-in-suit was “competitive requirement”).

Many experts have concluded that conjoint analysis is a proper way to calculate intellectual property damages. (Shugan Decl. ¶¶ 10–18 (listing sources).) Indeed, Google’s damages expert Dr.

19


Cox has described conjoint analysis as a “rigorous” method of calculating damages.5 Conjoint analysis has been used extensively in litigation. (Shugan Decl. ¶¶ 15-16 (listing cases); Green, P.E. and V. Srinivasan, “Conjoint Analysis in Marketing: New Developments with Implications for Research and Practice,” Journal of Marketing, October 1990, pp. 3-19 (article published over 20 years ago, noting increasing use of conjoint analysis in litigation).)

Seeking some disapproval of conjoint analysis, Google mischaracterizes a footnote in McLaughlin v. American Tobacco Co., 522 F.3d 215 (2d Cir. 2008). (Dkt. 718 at 15-16.) McLaughlin says absolutely nothing about the admissibility of conjoint analysis, much less its application to patent damages.6 Google provides no legal basis on which to exclude conjoint analysis, and it is wrong to claim that a “dearth of opinions” suggests that such analysis is inadmissible. Daubert does not require that a methodology first be accepted by other courts to be admissible. To so hold would be to return to the “general acceptance” test that Daubert expressly rejected. Daubert v. Merrell Dow Pharms., Inc., 509 U.S. 579, 597 (1993).

2. Prof. Shugan’s Conjoint Survey Was Methodologically Sound
Google next presents a series of specific challenges to the reliability of Prof. Shugan’s conjoint analysis. (Dkt. 718 at 17-20.) Google made every one of these arguments in its opposition

____________________________________

5 Norton Decl. Exh. I (Dr. Alan Cox, “A Better Consumer Survey for Better Damages,” IP Value Commentator (April 2003), http://www.nera.com/extImage/5958.pdf (“In a choice modeling exercise, survey respondents compare the features of a product and report their preferences for a particular combination of features. The results can be translated into monetary values. . . . This willingness-to-pay and predicted market penetration can then be used to develop a damage estimate.”)); id. Exh. J (Dr. Alan Cox & Louis Guth, Survey Techniques for Rigorous Measurement of Damages in Trade Dress Confusion Cases, NERA (Jan. 8, 2007), http://www.nera.com/67_4876.htm (choice modeling is a “widely accepted, testable technique that has been peer reviewed, has known rates of error and which is often used in market research -- not merely in litigation.”)).

6 McLaughlin concerned an effort to extend a presumption of class-wide reliance to a RICO class action alleging that class members had relied on statements about the health benefits of “light” cigarettes. 522 F.3d at 223-25. Although plaintiffs had offered a conjoint analysis in the trial court, that analysis was not at issue on the appeal. Rather, in the course of explaining why it would not import a presumption of reliance into RICO cases, the Second Circuit observed in a footnote that the plaintiffs also had difficulty coming up with direct proof of class-wide reliance. Id. at 225 n.6. Google incorrectly suggests that the Second Circuit found the study inadequate to prove damages “with sufficient precision to allow a jury award.” (Dkt. 718 at 16.) The Second Circuit was not commenting on the adequacy of the studies; it was holding that the plaintiffs’ overall approach to damages – a “fluid recovery” that relied on [r]oughly estimating the gross damages to the class as a whole and only subsequently allowing for the processing of individual claims” – would expose defendants to excess damages and violate due process. McLaughlin, 522 F.3d at 231.

20


to Oracle’s Daubert motion to preclude Google’s patent damages expert from testifying about the conjoint analysis. (Dkt. No. 581 at 14-17; Leonard Decl. (Dkt. No. 581-1) ¶¶ 26-46, 64, 66, 73.) At that time, Google emphasized that these challenges were “a dispute between experts” and “factual disputes or disagreements between experts are not grounds for striking expert testimony.” (Dkt. No. 581 at 17.) The Court instructed Oracle to “save its critiques” of Dr. Leonard “for trial.” (Dkt. No. 632 at 7.) The same result should apply now. See Ruiz-Troche v. Pepsi Cola of Puerto Rico Bottling Co., 161 F.3d 77, 85 (1st Cir. 1998) (“As long as the expert’s scientific testimony rests upon ‘good grounds, based on what is known,’” it should be permitted “rather than excluded from jurors’ scrutiny for fear that they will not grasp its complexities or satisfactorily weigh its inadequacies.”). But Google’s Daubert arguments are not only misplaced jury arguments. They are also wrong.

First, Google claims that Prof. Shugan’s selection of features to be tested was driven by litigation. This argument is both false and irrelevant. Before including any of the features relevant to this litigation, Prof. Shugan first confirmed that the features were actually relevant to consumer choice, through interviews, market research, a focus group, industry analyst reports and buyers’ guides, as he explained in his deposition. (Shugan Decl. ¶¶ 22–24; Shugan Report p. 7–8, App’x D at 5–6.) After he identified 36 features that real-world consumers said mattered in making a purchasing decision, Prof. Shugan did include features in his survey that were not at issue in this litigation, including brand and price. It is not necessary in conjoint analysis to test every distinguishing feature that may matter to consumers, because conjoint assesses relative importance. (Shugan Decl. ¶ 25.)

Second, Google complains that some of the tested features were “spoonfed” to Prof. Shugan by Prof. Cockburn. (Dkt. 718 at 18.) By that Google means that Prof. Cockburn told Prof. Shugan which infringement-enhanced features had to be tested. These are the same features that Google calls “obvious.” (Id. at 1.) That is not a flaw in the study. It is the point of the study.

Third, Google claims that Prof. Shugan erred by assessing the value of features “rather than the incremental benefit to those features allegedly enabled by the technology at issue.” (Id. at 18.) This argument is nonsense. The measurements that Prof. Shugan took were of the incremental benefits provided by the patents and copyrights as they affect features, measured by the benchmark tests upheld by the Court. Prof. Shugan measured, for example, how much market shares would shift

21


if applications opened in 0.2 versus 2.0 versus 4.0 seconds – the precise differences demonstrated by the benchmarking. (Shugan Report Exh. 3a–f; Report Exh. 5 (showing benchmarking results).) Prof. Shugan could have measured nothing more granular. (Shugan Decl. ¶¶ 28–32.)

Fourth, Google argues that Prof. Shugan’s instruction that survey respondents assume that any feature not listed was the same for all phones “undermines the survey’s ability to predict realworld behavior” because “all phones are not the same in the real world” (Dkt. 718 at 16, 18). Google further claims consumers in fact did not hold constant all unnamed features, as is supposedly shown by the purported fact that 24% of respondents “prefer a smartphone costing $200 to a putatively identical one costing $100.” (Id. at 19). As Prof. Shugan explained in his October reply report, Google misinterprets his data. A proper reading indicates that 8.8%, not 24%, of respondents were insensitive to the $100 price increase. (Shugan Decl. ¶ 39, Exh. A (Shugan Reply Report at 19).) But even that 8.8% does not give substance to Google’s claims. As Prof. Shugan explains, the ability of a Bayesian model to predict aggregate consumer behavior is not tested by focusing on individual outlier cases. (Id. ¶ 41-42.) In any event, it is widely recognized that consumers may associate a higher price with prestige in a conspicuously consumed item like a smartphone, or with durability. (Dkt. No. 595 (11/1/11 Shugan Decl.) at 10.) Google’s complaint is that when it reads the data the wrong way, the study results do not conform to Google’s uninformed assumptions about consumer behavior. Google’s mistakes say nothing about the reliability of Prof. Shugan’s work.

Fifth, Google argues that Prof. Shugan “conceded” that respondents did not hold nonspecified features constant because some implicitly associated prestige or durability with price. This argument misconstrues Prof. Shugan’s testimony. (Shugan Decl. ¶ 33.) Prof. Shugan explains:

When respondents implicitly attribute aspects of other attributes to price or brand name, that is not inconsistent with holding constant all other variables that are not included in the conjoint study. There is no reason to believe that respondents who do enrich the value of the price or brand with variables not included in the conjoint study vary their evaluation of price or brand between the 16 choice sets from which they choose their preferred smartphones. Therefore, even if respondents enrich the value of price or brand, I am still able to isolate the incremental benefit of the features at issue accurately.
(Id. ¶ 33.) In other words, the meaning of price and brand may differ slightly for some consumers, but each individual consumer can be expected to have a constant view of the meaning of price and brand. That is all that this survey requires.

22


Finally, Google claims that “[c]ourts routinely recognize that a ‘common sense’ understanding of real-world consumer behavior is an important check against the reliability of surveys,” (Dkt. 718 at 20, citing Johnson Elec. N. Am. v. Mabuchi Motor Am. Corp., 103 F. Supp. 2d 268, 286 (S.D.N.Y. 2000)), but does not cite a single case where a court actually did so. Johnson Electric involved no survey at all, contrary to Google’s representations. Even so, a rule that admissibility of expert testimony depends on the court’s “common sense” reaction to the results violates the principle that Daubert does not test results, but methods. Daubert, 509 U.S. at 595 (court must focus “solely on principles and methodology, not on the conclusions that they generate”). More importantly, if properly understood, nothing about the conjoint analysis here offends common sense.

F. Prof. Cockburn’s Econometric Analysis Is Methodologically Sound
Prof. Cockburn’s econometric analysis uses eBay smartphone auction data to determine how likely it is that consumers would switch to a different phone if Android lacked the performance benefits provided by the patents-in-suit. Even if Google’s arguments to exclude this evidence had any merit, which they do not, they would at most go to the study’s credibility. Indeed, as explained in the attached Cockburn Declaration, Google’s expert Dr. Leonard makes grievous errors in his treatment of the eBay data. (See, e.g., Cockburn Decl. ¶ 34–40 (explaining that Dr. Leonard, inter alia, improperly drops variables, misunderstands collinearity, and applies linear tests to nonlinear models).) This conflict among the experts should be reserved for trial.

1. Prof. Cockburn’s Econometric Analysis Is Based On Reliable Data.
Google argues that the preferences of people who buy used smartphones on eBay are not representative of purchasers of new phones bundled with a service agreement (Dkt. 718 at 22), but provides no reason to believe that the two groups value incremental performance gains differently. (See Cockburn Decl. ¶¶ 9–11.) See Kennedy, 161 F.3d at 1230–31 (defendant failed to introduce any evidence that extrapolation at issue was not scientifically valid). Economics and business literature teaches that there are strong relationships between new and used goods. (Cockburn Decl. ¶ 11.) Google’s own Chief Economist, Hal Varian, has emphasized how useful eBay data is for understanding consumers’ willingness to pay for products they demand, stating that “Online auctions offer a wonderful laboratory for experimental economists.” (Cockburn Decl. ¶ 20.)

23

Extrapolation is acceptable science. “Studies involving similar but not identical situations may be helpful, so long as an expert sets forth the steps used to reach the conclusion that the research is applicable.” In re Phenylpropanolamine (PPA) Prods. Liab. Litig., 289 F. Supp. 2d 1230, 1245 (W.D. Wash. 2003) (quoting Domingo v. T.K., M.D., 289 F.3d 600, 605–06 (9th Cir. 2002) (internal punctuation omitted)). Contrary to Google’s claims, Prof. Cockburn did explain why online consumers of used phones could be expected to behave similarly to brick-and-mortar purchasers of new phones (Report App’x C ¶¶ 9–11), and did control for possible difference between new and used phones. (Id. ¶ 29; see also Cockburn Decl. ¶ 11.) His analysis is sound.

2. Prof. Cockburn’s Analysis Is Based On Reasonable Assumptions.
Google next challenges two aspects of Prof. Cockburn’s econometrics analysis: his assumption that poorly performing, non-infringing Android phones would be sold at the same price as real world Android phones, and his use of a ten-day window to observe data for each bidder.

“[C]laims that ‘the assumptions relied on by an expert are unfounded is generally an argument that goes to the weight rather than the admissibility of the evidence.’”. Arista Records, 2011 WL 1674796, at *3. See also U.S. Gypsum Co. v. Lafarge N. Am. Inc., 670 F. Supp. 2d 737, 740–41 (N.D. Ill. 2009) (rejecting Daubert motion where infringers took issue “not with Davis’s subsequent calculations, but with her initial assumptions;” holding that “the jury will decide whether to accept or reject that factual predicate.”) Google’s two “authorities” for the proposition that Daubert excludes assumptions do not help it. The first emphasizes that testimony is excludable only if the assumptions are “so unrealistic and contradictory as to suggest bad faith.” Boucher v. U.S. Suzuki Motor Corp., 73 F.3d 18, 21 (2d Cir. 1996). The second is not an opinion at all. Google quotes and cites the defendant’s losing brief on a motion in limine – a motion that the court denied. (Norton Decl. Exh. C (Medical Instr. & Diagnostics Corp. v. Elekta AB, No. 97-CV-02271, Dkt. No. 464, at 14–15).)

Google argues that in a world in which Android phones performed more poorly, the price of Android phones would have declined, blunting the effect on its market share. But with the exception of a small number of Nexus One phones that Google sold directly, Google would have no influence over the price at which OEMs sold phones or the extent to which carriers subsidized them. Google cites no evidence to the contrary. Indeed, Prof. Cockburn will testify that the more likely scenario is

24


that Android would have failed if Google required OEMs or carriers to sell (or subsidize) poorly performing phones at a lower prices, while still facing the same costs. (Cockburn Decl. ¶¶ 13–16.) Consequently, Prof. Cockburn’s assumption does not inflate damages; it understates them. Indeed, the conclusion Google assumes – that poorer-performing Android phones would command a lower price – corroborates the conclusion of a significant reasonable royalty in this case.

Nor does Prof. Cockburn’s choice of a ten-day window to observe bids overestimate the number of people who would not purchase a smartphone. The number of buyers who would switch away from Android phones is not tied to a particular auction; the analysis compares successful bidders’ revealed valuations to the average prices for the phones on which they bid during the year the auction took place. (Cockburn Decl. ¶¶ 17–20.) Google merely asserts that if the data shows that 15% of prospective purchasers would have decided not to buy a smartphone in the but-for world, the data must be wrong. Google does not support that assertion or explain why that result is so surprising. In fact, in 2009, almost 80% of U.S. wireless subscribers used feature phones, not smartphones, and the majority of US phones today still are feature phones. (Cockburn Decl. ¶ 32.) Moreover, any supposed bias in Prof. Cockburn’s assumption is mitigated, if not eliminated, by Prof. Cockburn’s conservative assumption that Android did not increase the overall size of the smartphone market. (Id.) Google’s argument is at most a cross-examination point.

III. CONCLUSION

Oracle respectfully requests that the Court deny Google’s motion to strike.

Dated: February 24, 2012

BOIES, SCHILLER & FLEXNER LLP

By: /s/ Steven C. Holtzman
Steven C. Holtzman

Attorneys for Plaintiff
ORACLE AMERICA, INC.

25



739

UNITED STATES DISTRICT COURT
NORTHERN DISTRICT OF CALIFORNIA
SAN FRANCISCO DIVISION

ORACLE AMERICA, INC.
Plaintiff,
v.
GOOGLE, INC.
Defendant.

Case No. CV 10-03561 WHA

DECLARATION OF IAIN M. COCKBURN
IN SUPPORT OF ORACLE AMERICA,
INC.’S OPPOSITION TO GOOGLE’S
MOTION TO STRIKE

Dept.: Courtroom 8, 19th Floor
Judge: Honorable William H. Alsup


I, IAIN M. COCKBURN, declare as follows: 1. I have been retained by Oracle America, Inc. (“Oracle”) as an expert in this matter. In this declaration, at the request of counsel for Oracle, I address certain issues raised in Google’s Motion to Strike Portions Of Third Expert Report By Iain Cockburn And Expert Report By Steven Shugan (“Motion to Strike”) (Dkt. No. 720).

2. My background and qualifications, the terms of my retention, and the documents I have reviewed are set forth in the report I submitted in this matter on February 3, 2012 (as revised February 9, 2012) (“Cockburn Report”), and I incorporate them herein by reference. (See Cockburn Report ¶¶ 8–10; App’x A, B.)

My Use Of Certain Patent Value Distribution Curves In Connection With My “Group
And Value” Apportionment Analysis Was Reasonable
3. Google claims that the three studies I used to evaluate the expected distribution of value among the 569 patents included in the 2006 Bundle—the Harhoff,1 Barney,2 and PatVal3 studies—are inapplicable because they “have nothing to do with the Sun portfolio at issue” and “are likely to have different distributions of value than the Sun portfolio.” (Motion to Strike at 9–10.) In my opinion, Google’s criticisms are misplaced, and my use of these studies in my analysis was appropriate.

4. As I explained at my deposition, it is my opinion that the distribution of value among the 569 patents included in the 2006 Bundle would have been similar to the value distribution reflected in these studies, which is also consistent with a number of other studies and my own experience evaluating different patent portfolios, both in my academic work and in connection with various licensing negotiations.

__________________________________

1 Harhoff D., F. Scherer, K. Vopel, "Citations, family size, opposition and the value of patent rights" Research Policy 32, October, 2002, pp. 1343-1363 (“Harhoff”).

2 Barney, J. A., "A Study of Patent Mortality Rates: Using Statistical Survival Analysis to Rate and Value Patent Assets," AIPLA Quarterly Journal, Vol. 30, No. 3, Summer 2002. (“Barney”).

3 Gambardella A., P. Giuri, and M. Mariani, "The Value of European Patents - Evidence from a Survey of European Inventors," Final Report of the PatVal EU Project, January 2005 (“PatVal”).

1


5. It is well-established and widely recognized that patent values are generally highly skewed, with a relatively small number of patents having significantly greater value. The three studies that I cited in my report each recognize this principle, and cite to other studies that confirm this same principle:

  • “Consistently with previous findings in the literature, the economic value (measured in monetary terms) of the PatVal-EU patents is skewed: a small share of patents yields very high economic returns.” (PatVal, 2005, p. 5 (true and correct copy of article attached to this declaration as Exhibit A).);
  • “The task of assessing the value of patent rights is a particularly difficult one, since the distribution of these values is highly skew. The skewness property has been discussed by numerous authors, e.g. Scherer (1965), Pakes and Schankerman (1984, p 79), Pakes (1986), and Griliches, 1990.” (Harhoff, 2003, p. 1344.);
  • “[T]he model supports the view, long held by many in the field, that patent values are highly skewed. A relatively large number of patents appear to be worth little or nothing while a relatively small number appear to be worth a great deal.” (Barney, 2002, p. 329.).
6. Many other studies reinforce this conclusion

  • “[T]he distribution of the value of patented innovations is known to be extremely skew implying that a few patents are very valuable, and many are worth almost nothing.” (Hall, 1999, p. 14.4 );
  • “This paper draws implications for technology policy from evidence on the size distribution of returns from eight sets of data on inventions and innovations attributable to private sector firms and universities. The distributions are all highly skew; the top 10% of sample members captured from 48 to 93 percent of total sample returns.” (Scherer, 2000 (true and correct copy of this article attached to this declaration as Exhibit B).5 );
  • "[I]n 1980 the top 10% of licensed technologies account for 95% of total gross income, and in 1990 the top 10% account for 88%. This feature of the distribution - that outlying tail values account for a large proportion of cumulative revenue - is consistent with previous evidence on the distribution of returns from industrial innovations (see Scherer and Harhoff 2000 for an excellent review), and university
__________________________________

4 Bronwyn Hall, “Innovation and Market Value,” NBER Working paper 6984, 1999.

5 Scherer, F. M. and D. Harhoff, “Technology policy for a world of skew-distributed outcomes,” Research Policy, v. 29, 2000, pp. 559-566.

2


    inventions (Mowery et al. 2001; Mowery and Sampat 2001).” (Sampat and Ziedonis, 2005, pp. 285-286.6 );

  • “The distribution of the private value of patent rights is sharply skewed in all technology fields, with most of the value concentrated in a relatively small number of patents in the tail of the distribution.” (Schankerman, 19987 ).
7. Google’s contention that a single-firm portfolio, such as the portfolio that I considered in this case, is “likely” to have a different distribution curve (Motion to Strike at p. __) is, in my opinion, incorrect. Google has not cited any study or paper in support of this proposition, and I know of none. In contrast, one study from 2000 shows that the patents of a single firm (Harvard) have a similar degree of skewness as the patents shown in the broader studies of both German and US patents, with the top 10 percent of each set capturing more than 80 percent of the value of that set.8

8. There have been a number of studies that focus on the distribution of value across various different patent populations, and they all come up with similar distribution curves. That includes published research that confirms the same highly skewed distribution for patents limited to specific product areas.9 Every paper I have reviewed suggests that the distribution of value of patent is highly skewed, and they confirm that the distribution curves that I used in my analysis for this case are reasonable.10 I would expect such similar distribution curves to apply to single-firm portfolios, including the one that I evaluated in this case.

9. My opinion in this respect is reinforced by the work that I did with a doctoral student, Ajay Agarwal, at University of British Columbia, who was studying patents licensed or offered for

____________________________________

6 Bhaven Sampat and Arvids Ziedonis, “Patent Citations and the Economic Value of Patents,” Handbook of Quantitative Science and Technology Research 2005, Part 2, 277-298.

7 M. Schankerman, “How Valuable is Patent Protection? Estimates by Technology Field,” Rand Journal of Economics, Vol. 29, 1998, pp. 77-107.

8 Scherer, F. M. and D. Harhoff, “Technology policy for a world of skew-distributed outcomes,” Research Policy, v. 29, 2000, pp. 559-566.

9 Harhoff D., F. Scherer, K. Vopel, “Citations, family size, opposition and the value of patent rights” Research Policy 32, 2003, pp. 1343-1363.

10 See also, e.g., Gambardella A., P. Giuri, and M. Mariani, “The Value of European Patents - Evidence from a Survey of European Inventors,” Final Report of the PatVal EU Project, January 2005.)

3


license by the Massachusetts Institute of Technology. As I explained in my deposition, Dr. Agarwal studied licensing payments received by MIT for all of the patents in its portfolio. Such payments provide a basis for economic valuation based on received or projected licensing revenue, and the distribution of value within that single portfolio displayed the same property that I rely on in my report: a very small number of patents, 1-2%, commanded more than 50% of the value of the portfolio. That work is consistent with the Scherer article, and confirms that single-firm portfolios demonstrate similar value distributions as the larger samples reflected in other studies.11

10. Additionally, as I explained in my deposition, I recently worked on a research project in which I reviewed licensing data collected for several thousand patents held by two large medical research centers. These portfolios owned by a specific institution again display the same phenomenon that is described in the Harhoff, PatVal, and Barney papers: the distribution of value in these single-company portfolios was highly skewed, in that a small number of patents had a substantially greater value that the rest of the patents. Even among those patents that were licensed, a handful constitute the vast majority of the value, and have been licensed or sold at a premium.

11. Based on my conversation with Alfonso Gambardella, the author of the PatVal study, the distribution of the value within the portfolios of individual companies is very similar to the distribution of value within the larger sample of patents in the published paper.

12. This phenomenon has been repeatedly described to me in the course of conversations and research interviews with licensing executives for large US-based technology companies, including IBM, United Technologies, and Pitney Bowes.

My Calculation Of The Value of Copyrights Included In The 2006 Bundle Other Than
The Infringed APIs Was Reasonable
13. Google claims that I confuse cost with value based on my analysis of the copyrighted materials at issue in the 2006 Bundle, and argues that my analysis is therefore inadequate. (Motion to Strike at 13.) Google is wrong. The “millions of lines of code” that Google claims would have been

__________________________________

11 Bhaven Sampat and Arvids Ziedonis, “Patent Citations and the Economic Value of Patents,” Handbook of Quantitative Science and Technology Research 2005, Part 2, 277-298

4


copyrighted in 2006 would not have any value to Google separate and apart from the code itself, which I have fully accounted for in my assessment of the cost to write that code. As explained in my report, I fully accounted for the copyrighted materials included in the 2006 Bundle but which are not in-suit by calculating the full cost of developing that copyrighted code.

14. Google’s patent damages expert, Dr. Leonard, uses the same proxy as I do to calculate value. In Dr. Leonard’s October 2011 report, he also used the cost of writing a certain amount of code—the cost of building a virtual machine—which he pegs at approximately $11 million. If there is significant “value” on top of the pure code of a virtual machine, Dr. Leonard should have accounted for such value in this calculation.

15. I have not confused cost and value. If there had been an agreement between the parties, it might have cost Sun next to nothing to provide Google some of the source code Google wanted for the Java virtual machine and associated libraries. However, given that all these items provide value to Google, I calculate the value to Google as the cost of writing that code from scratch. Effectively, the value is the avoided cost. By calculating the cost to Google of writing that source code, I have fully accounted for the value to Google.

The Conjoint Analysis
16. I understand that Prof. Steven Shugan, who conducted a conjoint study to evaluate the enhancements enabled by the use of the copyrights and patents that Google is alleged to have infringed, has submitted a declaration in support of Oracle’s opposition to Google’s motion in which he explains how Google’s critiques of the conjoint analysis are badly misinformed. I have reviewed Prof. Shugan’s critiques, and I agree with them.

The Econometrics Analysis
17. Google claims that my econometric analysis should be excluded for two reasons. First, Google claims that my analysis is based on unrepresentative data. Second, Google claims that my calculation of market share is based on unreasonable assumptions about price and consumer choices. Google is wrong on both points.

A. My calculations are based on representative data.

5


18. Google claims that eBay purchasers are not representative of purchasers in brick-and-mortar stores. Google’s position that online sales data do not provide meaningful information about consumer preferences is contrary to Google’s own practices. As a company, Google makes money by analyzing online consumer behavior and selling that information to firms interested in reaching potential consumers. 12 19. Google has emphasized that it uses data from online purchases, rather than brick-and-mortar stores.

  • Google has proposed a Google Price Index (GPI) based on online prices as an improved way of tracking inflation. Rather than waiting for answers from brick and mortar stores (as the U.S. government does in producing the CPI), GPI tracks inflation through online prices. Generally, Google’s index and the U.S. Government’s index closely track one another.13 An independent effort to use online price information to track pricing and inflation has found very similar results. MIT’s Billion Prices Project’s (BPP) uses online prices to construct a measure of inflation that tracks the Bureau of Labor Statistic’s measure of the CPI very closely, suggesting that online prices have a close correspondence to those observed in traditional retail outlets.14
  • Google Shopper searches the web (including the websites of both online and brick-and-mortar stores) to deliver the best prices for keyword-searched goods. It is difficult to reconcile Google’s claim that consumers who buy their phones from eBay versus mobile carriers and brick and mortar stores are not linked when Google, itself, provides a shopper that allows consumers to compare prices between the two.15
  • When Google commissions studies to better understand purchasing behavior, it has, itself, looked at eBay data. For example, eBay was among the sites analyzed in a study
__________________________________

12 Firms that advertise on Google include both brick and mortar and online firms.

13 See, e.g., Robin Harding, Google to map inflation using web data, Financial Times, ft.com (Oct. 11, 2010), available at http://www.ft.com/intl/cms/s/2/deeb985e-d55f-11df-8e86- 00144feabdc0.html#axzz1mvc6HBDa; Alexis Madrigal, Google Price Index Highlights Slowness of Economic Data Collection, The Atlantic, (Oct. 12, 2010), available at http://www.theatlantic.com/technology/archive/2010/10/ google-price-index-highlights-slowness-ofeconomic- data-collection/64393/; Tavia Grant, Google eyes trends as economic indicators, The Globe And Mail, (Oct. 12, 2010), available at http://www.theglobeandmail.com/report-onbusiness/ economy/google-eyes-trends-as-economic-indicators/article1754248/.

14 http://bpp.mit.edu/usa/

15 See, e.g., http://www.google.com/mobile/shopper/

6


    commissioned by Google to better understand consumer purchases of e-readers, netbooks, tablets, and laptops.16
20. Google’s Chief Economist, Hal Varian, has even emphasized how useful eBay data could be for economic use. Indeed, as the two examples below make clear, Dr. Varian clearly endorses the use of eBay data to understand consumers’ willingness to pay for products they demand. Dr. Varian’s biggest concern appears to be that because bidders seem reluctant to use eBay’s automated “bidding agent” that is intended to eliminate any incentive for late bidding, eBay data leads to an underestimate of consumers’ willingness to pay (e.g., Dr. Varian seems to believe that my econometric analysis is an appropriate way to conservatively estimate consumers’ willingness to pay).

  • Dr. Varian has used online data from eBay to test his own theories. In a New York Times article,17 he wrote: “Online auctions offer a wonderful laboratory for experimental economists: the participants are intelligent adults, spending real money, who hope to purchase goods in which they have an intense interest. This is a far cry from the reluctant sophomores whom experimentalists have had to rely on in the past to test economic theories. Analysis of online auction data has yielded a wealth of insights, and a few puzzles. A particularly intriguing puzzle has been the tendency for ‘late bids.’ In a representative sample of eBay auctions, researchers found that 37 percent of them exhibited bids in the last minute and 12 percent had bids in the last 10 seconds. These data understate the actual number of bids submitted in the closing seconds of the auction, since bids that arrived at eBay after the auction had closed were not counted. The late-bid puzzle is particularly interesting, since eBay offers an automated ''bidding agent'' that is intended to eliminate any incentive for late bidding. I only need to tell my bidding agent the most I am willing to pay for an item, along with my initial bid. If someone bids more, my agent will automatically increase my bid by the minimal bid increment, as long as this doesn't raise my bid over my maximum. In theory, each bidder should report his true maximum willingness to pay and let the agent do the work of bidding. In the end, the person with the highest willingness to pay for the item will win the auction, paying a price equal to the second-highest willingness to pay plus the bid increment.”
___________________________________

16 See, e.g., Google.com, Portable PC Shopper, October 2010, available at http://www.thinkwithgoogle.com/insights/library/studies/portable-pc-shopper/ Among the paper’s conclusions are that shoppers looking for e-readers, netbooks, tablets, and laptops check the following different sites when shopping: Amazon, eBay, Sam‘s Club, Best Buy, Fry’s, Staples, Circuit City, Newegg, Target, Comp USA, Office Depot, Tiger Direct, Costco, Office Max, Walmart, CDW, Radio Shack. See id. at p. 38.

17 Hal Varian, Online auctions as a laboratory for economists to test their theories,
New York Times (Nov 16, 2000), available at
http://people.ischool.berkeley.edu/~hal/people/hal/NYTimes/2000-11-16.html

7


  • Dr. Varian has also written a chapter in a microeconomics textbook discussing auction design that discusses how in a Vickrey auction (which eBay’s proxy bidder system effectively is), it is in the bidder’s interest to submit his or her true willingness to pay.8
21. In addition to being contrary to its own business practices, Google’s suggestion that the econometric analysis is based on unrepresentative data is incorrect for a number of reasons.

22. First, this criticism misunderstands the purpose of the econometric analysis. I analyzed eBay data not to model the price at which OEMs or mobile carriers sell smartphones to their customers, nor to measure the discount that a used or unlocked smartphone sells at compared to a new smartphone. In either of these scenarios, the difference in price point between an unlocked, used phone and a new phone bundled with a carrier plan might have some significance, because the analysis would be sensitive to differences between the two populations. However, that was not the purpose of my analysis. Instead, I measured how consumers value the incremental benefit of features, particularly performance and speed. I therefore tested an eBay purchaser’s incremental willingness to pay for an incrementally faster phone against the same population’s incremental willingness to pay for an incrementally slower phone. The difference is important, because the population I analyzed was held constant. The only question one needs to ask is whether there is any reason to suppose the incremental value that online purchasers place on these features is any more or less than the incremental value of purchasers who buy in traditional in-store sales outlets. There is, in fact, no reason to suppose there are any meaningful differences, nor have I seen Google offer any evidence to the opposite.

23. Second, Google’s criticism of the representativeness of eBay data ignores the role of resellers in the eBay data. Resellers, who either purchase their stock on eBay or use eBay as a outlet to sell stock purchased elsewhere, provide a link between eBay and other channels for purchase. These resellers, in order to make a profit, must reflect end consumers’ preferences. It is a wellestablished proposition in economics that the ability to arbitrage across customer subgroups makes it

__________________________________

18 Varian, H. R., Intermediate Microeconomics: A Modern Approach, Seventh Edition, pp. 315-316.

8


impossible for firms to differentially price across the subgroups.19 In other words, this ability (and potential for) arbitrage leads to one price, regardless of where or how a consumer purchases a product.20 Google fails to understand this fundamental link. Instead, Google suggests that my results are unreliable because I base them on data which includes resellers. Thus, Google’s criticisms—on the one hand, criticizing eBay data for being unrepresentative and on the other hand, criticizing the inclusion of resellers—are inconsistent and show little understanding of how markets work.21

24. Third, Google’s concerns regarding the inclusion of used phones in the eBay data is not well-grounded. Google fails to recognize the important economic relationship between new and used goods. Exactly the same models of phones are sold on eBay as are sold in physical stores, whether in new condition, used, factory reconditioned, or otherwise. There is a long literature in economics and business that teaches us that there are strong relationships between new and used goods.22 This is true whether new and used goods are sold in the same channels or through different

___________________________________

19 “Under arbitrage, a consumer who is offered a lower price for a good by a firm purchases an excess quantity of the good and resells the good to consumers who are denied the lower price by the firm. Under perfect arbitrage, the firm would be forced to sell all its output at the lowest price to consumers offered the lowest price, who would then resell to other consumers. Thus, arbitrage effectively turns price discrimination into offering a single price.” Preston MacAfee, “Price Discrimination,” Issues in Competition Law and Policy, Volume I (ABA Section of Antitrust Law 2008), p. 467. See also: Hal Varian, “Price Discrimination,” Chapter 10, Handbook of Industrial Organization, Volume I, Edited by R. Schmalensee and R.D. Willig , Elsevier Science Publishers B.V., 1989, p. 599 and Jean Tirole, The Theory of Industrial Organization, MIT Press, 1988, pp. 134- 135.

20 In economics, this is referred to as the “Law of One Price” – which demonstrates that in a competitive market, all transactions between buyers and sellers occur at a single, common market price. Preston MacAfee, Price Discrimination, Issues in Competition Law and Policy, Volume I (ABA Section of Antitrust Law 2008), p. 467. See also: David Besanko and Ronald R. Braeutigam, Microeconomics (4th edition), Wiley, 2010, at p. 331.

21 As Dr. Leonard noted, several bidders bid on many auctions – probably with the objective to resell these phones. Furthermore, many sellers on eBay operate on several electronic stores– such as eBay, Amazon Marketplace, and Buy.com. Some eBay sellers operate both in their own electronic stores or in brick-and-mortar stores. The evidence of the many linkages between eBay and more traditional distribution channels – electronic or not – can only lead to one conclusion: the law of one price will prevail thanks to arbitrage.

22 See, for example, “Exploring the Relationship between the Markets for New and Used Durable Goods: The Case of Automobiles. Devavrat Purohit. Marketing Science Vol. 11, No. 2 (Spring, 1992), pp. 154-167. ; Benjamin D, Kormendi R. “The Interrelationship between Markets for New and Used Durable Goods,” Journal of Law and Economics, October 1974, 17(2):381-401.

9


channels, as they are in this case.23 Google also claims that whether or not a phone is locked to a particular carrier is a significant issue. Again, it is important to note that the same models of phone are sold locked and unlocked in traditional outlets as well as on-line, and purchasers can always unlock the phone themselves or have this done for them at a nominal fee. Prices of locked and unlocked phones are therefore closely related. In any case, although there may be some differences between the prices of new versus used or locked versus unlocked phones, Google fails to recognize that my econometric analysis explicitly accounts for whether a phone is new or used and whether a phone is locked or unlocked.24 The econometric modeling will absorb any systematic differences, if they exist. Yet, Google fails to acknowledge these facts.

B. My calculations are not based on any unreasonable assumptions.
25. Google claims that my calculation of market share is based on unreasonable assumptions about price and consumer choices. First, Google claims that contrary to my assumption, the price of Android phones wouldn’t have remained unchanged in a but-for world in which Android phones would be slower. Second, Google claims that my choice of a ten-day window to observe bids leads to an overestimate of the number of people who would not have bought the slower Android phone. Neither one of these critiques has merit.

1. The assumption that the price of Android would remain constant is reasonable.
26. Google claims that my assumption that the price of Android phones would have remained unchanged in a but-for world is incorrect and that it leads to an overestimate of damages. Google’s apparent logic is that in a world in which Android phones were slower and performed more poorly, the price of Android phones would have declined. This is an alternative assumption to the way I conduct my primary analysis. In making this argument, Dr. Leonard and Google completely ignore the obvious implication that any decline in prices of Android phones due to the reduced

_______________________________

23 Again, we expect the law of one price to hold, regardless of channel. See, for example, Shulman, Jeffrey D., Coughlan, Anne T. “Used goods, not used bads: Profitable secondary market sales for durable goods channel,” Quant Market Econ, (2007) 5:191–210, June 5, 2007.

24 As I explained in Appendix C to my February 3, 2012 report, I include controls for whether a phone is used or new, and whether a phone is locked or unlocked.

10


quality of the operating system itself is not irrelevant in measuring the reasonable royalty due to Sun. Quite the opposite – this price decline itself is one such measure.

27. I have considered and discussed this possibility in my report. In Exhibits 6-9 I have discussed the change in users’ willingness-to-pay for Android phones with diminished performance. For example, I have determined that disabling the ‘104 patent would translate into a decrease in users’ willingness-to-pay of $24-$29 (Exhibit 6). Presumably, if the price of Android phones decreased by that amount in the absence of features enabled by the ‘104 patent, sales of Android phones would remain unchanged.

28. In that case, one has to consider how this price differential would be financed. The wireless industry is intensely competitive. Android OEMs and carriers compete with very similar products and therefore likely earn small margins. If these companies were forced to reduce price by $24-$29 to compensate for lower quality of the Android OS , they would require Google to compensate them, for example through additional technical assistance or (increased) revenue sharing or both. In the end, it would be Google’s burden to absorb the majority, if not all, of the price decline, or face the likely failure of Android. Therefore, I conclude that Google would have been willing to pay a patent royalty to Sun in the amount up to the decrease in the users’ willingness-topay. For the case of the ‘104 patent, Google would be willing to pay Sun up to $29 per unit. The upper bound of my calculated royalty rate for the ‘104 patent is just $2.36 (see Exhibit 12b) – over 10 times less than this measure of Google’s willingness-to-pay for that technology. This confirms that my analysis of royalty rates is reasonable and conservative.

2. My assumption that the ten-day window to observe bids is representative is
reasonable.
29. Google next claims that my choice of a ten-day window to observed bids leads to overestimate the number of people who would not purchase a smartphone. This claim is wrong. It betrays a fundamental misunderstanding of my model.

30. Google ignores the fact that the computation of the number of buyers who would switch away from Android phones is not tied to a particular auction, but compares bidders’ revealed valuations to the prevailing average prices for the phones on which they bid.

11


31. In my model, individual bidders’ valuations of the slower phones are compared to the average price of that phone. If the price is higher than the bidder’s valuation he simply doesn’t buy that phone. Contrary to what Google claims, this bidder will not go to AT&T or Verizon to buy the same phone: that slower phone is simply too expensive relative to that bidder’s valuation for them to be willing to buy it, whether it is sold on eBay, by a wireless carrier, or in a brick-and-mortar store.

32. Finally, Google contests my conclusion that 15% of Android phone buyers in 2010 wouldn’t have acquired a smartphone at all if the Android phones had been dramatically slower. Google’s critique is unrealistic, misguided and shows a fundamental misunderstanding of my market share calculations. In fact, in 2009, almost 80% of wireless subscribers in the United States used feature phones, not smartphones, and the majority of US phones today still are feature phones.25 Stated differently, my results imply that, if Android phones had performed less well, consumers would have been slower to switch from their cell phones to smartphones, or alternatively, would have been slower to upgrade their smartphone to a newer model. This is hardly a surprising finding. To put it in context, in the but-for world I construct, I find that the sales of smartphones in the United States would have increased from 46 to 101 million units between 2009 and 2011, instead of the actual increase from 46 to 107 million units, a reduction of 5.5% of overall smartphone sales. My model estimates that sales of Android smartphones in the but-for world would still have increased from 4 to 22 million units instead of the actual increase of 4 to 46 million units. My model is also conservative based on my assumption that Android did not increase the overall size of the smartphone market. These estimates are hardly “unreasonable” as Google alleges.

C. Google’s own expert, Dr. Leonard, makes elementary errors in his econometric
analysis.
33. Upon close analysis, it is Google’s expert Dr. Leonard, not I, who has made elementary mistakes in his econometric analysis. I only describe those errors briefly here, but it is

_________________________________

25 See http://blog.nielsen.com/nielsenwire/consumer/smartphones-to-overtake-feature-phones-in-u-sby- 2011/; http://gigaom.com/2011/09/01/four-in-ten-u-s-phones-are-now-smartphones/

12


my opinion that his analysis is flawed and unreliable, and his criticisms of my approach fundamentally erroneous.

34. First, Dr. Leonard falsely insists that the results from his Android-only model are meaningful. Yet, Dr. Leonard must know that by restricting his sample to only the 13 Android phone models he is unable to estimate the impact of more than twenty different smartphone features, as he claims to do. Mathematically, one simply cannot do this. The problem arises because there is insufficient variation in these features across the small number of phone models in Dr. Leonard’s restricted sample. Dr. Leonard’s “work-around” is to rely on the econometrics software package to arbitrarily drop certain features from the analysis. One result of this “work-around” is that the contribution of these omitted features to valuation “loads” on features that are retained, making those estimates unreliable. Furthermore, the result of this “work-around” is that Dr. Leonard’s model is nearly collinear, and displays the familiar symptoms of near collinearity.26 In these circumstances, his estimated coefficients are not reliable and are economically meaningless. Further, Dr. Leonard’s “results” are not based on the variability in phone features (which would provide meaningful results); rather, his “results” are based on the different timing of auctions (which does not provide meaningful results).

35. Second, Dr. Leonard wrongly insists that multicollinearity only affects standard errors. 27 This is nonsense; multicollinearity or near multicollinearity is generally recognized to also make

_______________________________

26 Greene provides an excellent summary of the current understanding of near-multicollinearity:

“When the regressors are highly correlated, we often observe the following problems:
  • Small changes in the data can produce wide swings in the parameter estimates.
  • Coefficients may have very high standard errors and low significance levels even though they are jointly highly significant and the R2 in the regression is quite high.
  • Coefficients will have the wrong sign or an implausible magnitude”.
(Greene, William H. (1997): Econometric Analysis, 3rd edition, London, Prentice Hall, p. 420).

27 Note that I do not calculate standard errors formulaically from the regression estimates, but instead report “boot-strapped” standard errors which are much less sensitive to the impact of multicollinearity.

13


parameter estimates highly sensitive to small changes in sample or specification.28 This is particularly true in non-linear models which are estimated here.

36. Third, Dr. Leonard insinuates that he ran a “Chow test.” I note that the Chow test per se is only appropriate for use on linear models; yet, both my model and Dr. Leonard’s version of my model are nonlinear models. If Dr. Leonard, in fact, used an analogue appropriate to this context, he has cited the wrong paper.29

37. Fourth, Dr. Leonard makes deceptive claims about the stability of estimated coefficients in different subsets of the data. Because the specification of the model, that is to say the list of explanatory variables, is not held constant across the different subsets of the data that Dr. Leonard considers (there is no or insufficient variation in some variables within subsets of the data), his analysis conflates the effect, if any, of variation over time in the economic relationship between valuations and the explanatory variables with the effect of changing the set of explanatory variables. For example, Dr. Leonard claims to observe variation in the Linpack coefficient for different time periods. Yet, the evidence Dr. Leonard presents is generated from models which severely limit the size of the sample used and therefore the set of models present in each subset. As it happens, in some of Dr. Leonard’s monthly regressions, there are simply no Android phones; in others there are only a

_______________________________

28 There are two types of near-multicollinearity. The first results from high correlation among regressors (structural issue.) The second is numerical. The regressor data matrix is ill-conditioned. This problem gives rise to erratic volatility. As Spanos and Guirk explain:

The presence of ill-conditioning in (XTX) indicates that the sample information is ‘nearlyinsufficient’ for reliable inference concerning X. In such a case the modeler should answer the question whether he/she can live with the potential erratic volatility as quantified by the norm bounds. If not, the only obvious way to proceed is to ameliorate the sample information in an attempt to render the data matrix well-conditioned. In the case of observational data this amounts to changing the units of measurement and/or getting additional or better quality data; see Silvey (1969). In the case of experimental data repeating or re-designing the experiment are additional potential options. (Spanos, A. and McGuirk A. (2002): The Problem of Near- Multicollinearity Revisited: Erratic vs Systematic Volatility, Journal of Econometrics, Vol. 108, pp. 365-393.)
Dr. Leonard’s analysis violates this well-established principle. Rather than using the existing welldesigned and acceptable sample, he divides the sample into sub-samples that suffer from erratic volatility.

29 Again, by reducing the sample artificially, Dr. Leonard generates subsamples with ill-conditioned regressor matrices. The fact that his so-called Chow test concludes that the subsamples are not identical only proves that the subsamples are unreliable.

14


few smartphones overall. During the time period covered by my analysis, many new models were introduced, and during any specific month not all models will be present in the data. Further, Dr. Leonard should know to attempt to measure 11 unique phone features in a month in which there are data for only five phones is impossible. As with his analysis of an Android-only subset of the data, by reducing the sample artificially, Dr. Leonard again generates subsamples with ill-conditioned regressor matrices, which, in turn, lead to unreliable and highly unstable estimates. Contrary to Dr. Leonard’s claim, his results simply do not support his claim that my requiring coefficients be equal across months is incorrect.

38. Fifth, Dr. Leonard wrongly claims that any bias in some coefficient estimates is necessarily transmitted to other coefficient estimates. The extent of any such “contamination” is an empirical question. Dr. Leonard offers no evidence to this point.

39. Sixth, Dr. Leonard further alleges that I made a comparison between his Android-only model and my revised Android-only specifications based on the Schwartz criterion (Dr. Leonard supplemental report, p. 12). I made no such comparison in my report. Dr. Leonard focuses on this imaginary comparison instead of addressing the very valid concerns I presented related to collinearity in both my rebuttal and 3rd damages report. On these substantive issues Dr. Leonard remains silent.

I declare under penalty of perjury that the foregoing is true and correct and that this declaration was executed on February 24th, 2012 at Boston, Massachusetts.

DATED: February 24, 2012

/s/ Iain M. Cockburn
IAIN M. COCKBURN

15


ATTESTATION OF FILER

I, Steven C. Holtzman, have obtained Dr. Iain Cockburn’s concurrence to file this document on his behalf.

Dated: February 24, 2012

BOIES, SCHILLER & FLEXNER LLP

By: /s/ Steven C. Holtzman
Steven C. Holtzman

Attorneys for Plaintiff
ORACLE AMERICA, INC.

16



740

UNITED STATES DISTRICT COURT
NORTHERN DISTRICT OF CALIFORNIA
SAN FRANCISCO DIVISION

ORACLE AMERICA, INC.
Plaintiff,
v.
GOOGLE, INC.
Defendant.

Case No. CV 10-03561 WHA

DECLARATION OF STEVEN M.
SHUGAN IN SUPPORT OF OPPOSITION
TO GOOGLE’S THIRD DAUBERT
MOTION

Dept.: Courtroom 8, 19th Floor
Judge: Honorable William H. Alsup


I, STEVEN M. SHUGAN, declare as follows:

1. I have been retained by Oracle America, Inc. (“Oracle”) as an expert in this matter. My background and qualifications, the terms of my retention, and the documents I have reviewed are set forth in the Expert Report I submitted in this matter on September 12, 2011, and in declarations I submitted on October 21, 2011, and November 1, 2011; I incorporate those documents herein by reference. (See Expert Report of Professor Steven M. Shugan (Sept. 12, 2011) (“Shugan Report”) p. 1–2; App’x A, B, C; Declaration of Steven M. Shugan In Support Of Oracle America, Inc.’s Motion to Exclude The Expert Reports Of Gregory K. Leonard and Alan J. Cox (Dkt. No. 560); Declaration in Support of Oracle America, Inc.’s Reply To Google, Inc.’s Opposition to Motion to Exclude Portions of the Expert Reports of Gregory K. Leonard and Alan J. Cox, November 1, 2011, p. 8-10. (Dkt. No. 595.) In addition to these materials, I submitted a Reply Report in this matter on October 10, 2011 (Expert Reply Report of Professor Steven M. Shugan (“Reply Report”)), which I am providing to the Court as Exhibit A to this Declaration. I incorporate by reference all of the analyses I performed in each of the four documents referenced in this paragraph, for completeness.

2. To review my qualifications in brief, I am currently the McKethan-Matherly Eminent Scholar and Professor at the University of Florida, where I teach multivariate statistics, marketing models and advanced marketing management. I hold a Ph.D. in Managerial Economics from Northwestern University and my research includes services marketing (integrating operations), statistics, metrics, entertainment marketing, advance-selling, normative methods for modeling competition, markets for evaluative information, and models of selling and product policy. I was formerly a full professor at the University of Chicago for 13 years, and I have taught marketing, econometrics, statistics, and computer science at various universities.

3. I was editor-in-chief of Marketing Science for six years, editor of Journal of Business and associate editor of Management Science, and I served on over 10 editorial boards including the Journal of Consumer Research, Journal of Marketing and Journal of Marketing Research. I have numerous publications (including 27 editorials and commentaries) and have made over one hundred professional presentations in more than 22 countries. Much of my work involves the evaluation of marketing research tools including surveys.

1


4. My fields of specialization within marketing include marketing strategy, marketing research, quantitative models, and consumer decision making. In the course of my scholarly research, teaching, editorial work, and consulting work, I have studied issues of marketing research, product design, product and feature development, branding, and pricing, and their roles in consumer preferences and choice. During my career, I have taught managers, M.B.A. students, and doctoral students about, written textbook chapters on, evaluated articles for publication on, and conducted conjoint analysis.

5. I was asked by counsel for Oracle to evaluate consumer preferences for certain smartphone features. Specifically, Prof. Iain Cockburn, on behalf of Oracle, asked me to conduct a conjoint study that evaluates the effect of particular features on the demand for smartphones, including those devices that run the Android operating system. (Shugan Report p. 3.) I conducted the survey at his direction, and reported the results of the survey and my analysis thereof in the Shugan Report, which I filed on September 12, 2011. I submitted a Reply Report on October 10, 2011, responding to certain critiques from Google’s damages expert, Dr. Gregory Leonard. I have not revised either report since then.

6. I understand that Google has moved to exclude the conjoint analysis, claiming that the analysis is unfit for litigation and had certain methodological flaws. (Google’s Motion to Strike Portions of Third Expert Report by Iain Cockburn And Expert Report by Steven Shugan. (Dkt. No. 720).) At the request of counsel for Oracle, I address certain issues raised by Google in that motion.

A. Google misunderstands the market share analysis I performed.
7. Google claims that “Dr. Shugan essentially converts the consumers’ preference share into projected market shares—essentially, he concludes that, if 20% of consumers value application start time more than other tested features, an increase in application start time on Android phones would mean a 20% drop in Android market share.”1

____________________________________

1 Google’s Notice of Motion and Motion to Strike Portions of Third Expert Report by Iain Cockburn and Expert Report by Steven Shugan; Memorandum of Points and Authorities in Support Thereof, Oracle America, Inc. v. Google, Inc., Case No. 3:10-cv-3561-WHA, N.D. Cal, San Francisco Division, February 17, 2012 (“Google’s Motion to Strike”), p. 15.

2


8. Google’s assertion is incorrect. It demonstrates a lack of knowledge of the analysis that I undertook and, more generally, of how conjoint studies are typically used to evaluate new products and features. Google misunderstands and mischaracterizes my standard market share analysis. Exhibit 3a to the Shugan Report concludes that an incremental change to the application start time on an Android device from two to four seconds would decrease sales by 20 percent. This 20 percent change is the result of estimating, based on the choices respondents made in the study, the number of buyers who would select a different phone when the start up time is changed by two seconds. Thus, the analysis estimates preference share movements when changing from one level to another within one feature, holding all other features and levels constant. It is nonsensical to conclude that this result occurs because 20 percent of consumers value application start time “more than other tested features,” when those “other features” are held constant within the model.

9. Contrary to Google’s assertions, my conjoint study does not simply convert feature importance into preference share changes. It relies on the expected distribution of respondent choices given that a particular set of profiles were available in the market place. Each respondent’s set of preferences in the various scenarios is evaluated such that the most preferred profile, characterized by all seven features, including application start time, is identified for that respondent. By summing these selections for all respondents under the two scenarios, with different start times for Android devices, holding all else equal, I am able to evaluate the expected loss in market share using standard analyses with standard software.

B. Conjoint analysis has been accepted by marketing professionals, academics, and
courts.
10. Conjoint analysis is a combination of data collection through surveys and analysis of the data provided by respondents using standard, well-tested methods relied upon by academics and commercial entities for more than 40 years.

11. Specifically, conjoint analysis employs a two-step procedure, both of which have been widely employed in litigation. In the first step, data are collected using standard survey methods. Consumers are simply asked to choose among different sets of “products” defined by different features levels. In the second step, the data is analyzed using a standard choice modeling technique,

3


using commercially available and publicly tested software, to analyze survey data. In this case for the data analysis, I used a logit model, which is a type of regression model, in a software package developed by Sawtooth Software Inc. (“Sawtooth Software”).

12. Sawtooth Software is a leading provider of software packages used by marketing researchers and ultimately relied upon by corporations for making numerous product development and marketing decisions. Sawtooth Software developed the most commonly used software to conduct conjoint analysis. The software package was created with the assistance of leading academics in the field of Bayesian analysis and is constantly being improved and enriched in its capabilities based on the input that Sawtooth Software receives from academics and practitioners during seminars and conferences that the firm organizes.2 Because of Sawtooth Software’s close connection to the academic world, the software is highly reliable and provides accurate results.

13. Software developed by Sawtooth Software has been used in litigation matters before and specifically for the assessment of damages. Bryan Orme, the current President of Sawtooth Software, states in his introductory book on conjoint analysis that conjoint analysis is used in the assessment of damages in litigation matters.3

14. Well-known academics have also noted the use of conjoint surveys in litigation matters: “In addition to the use of conjoint analysis for marketing and strategic analysis, its applications are becoming increasingly diverse. One area of growing interest is litigation. Recently, conjoint studies provided primary input to the settlement of disputes in the telecommunications (foreign dumping of equipment in the U.S.), Pharmaceuticals (lost profits through misleading competitive product claims), and airline (alleged brand position bias in travel agents’ reservation computer screens) industries.”4

_________________________________

2 Bryan Orme, “Getting Started with Conjoint Analysis: Strategies for Product Design and Pricing Research,” Second Edition, Research Publishers LLC, 2010, pp. 29-37.

3 Bryan Orme, “Getting Started with Conjoint Analysis: Strategies for Product Design and Pricing Research,” Second Edition, Research Publishers LLC, 2010, p. 137.

4 Green, P.E. and V. Srinivasan, “Conjoint Analysis in Marketing: New Developments with Implications for Research and Practice,” Journal of Marketing, October 1990, pp. 3-19 at 15.

4


15. According to Green, Krieger, and Wind, conjoint analysis has been used in the following cases:5

  • Antidumping litigation – AT&T vs. Pacific-rim manufacturers legal dispute regarding small business telephone equipment; case adjudicated in AT&T’s favor
  • Intermittent windshield wipers litigation – study to determine consumer evaluations of the derived “willingness to pay” for the intermittent wiper feature
  • Continental vs. American Airlines litigation – conjoint study of travel agents’ tradeoffs among airline flight selections
  • TiVo v. EchoStar, in the Eastern District of Texas
  • U-Haul Int’l v. Jartran Inc., in the District of Arizona
  • Robert Kearns v. Ford Motor Co, in the Eastern District of Michigan
16. Furthermore, conjoint analysis has also been used in reports that were submitted to Congress and used by the U.S. Navy: 6

  • Health maintenance plans: study conducted by the American Association of Retired Persons; results submitted to Congress
  • U.S. Navy benefit packages for reenlistment: conjoint used to develop menu of new reenlistment plans based on individual differences in types of duties, health needs, and sign-over bonuses.
C. Conjoint analysis is frequently used in practice to model consumer behavior.
17. Google claims that “[c]onjoint analysis measures consumer preference for product features; it does not capture how consumers actually behave when purchasing a product. Consumers’ stated preference for a given feature may be one of many factors a company considers in designing or launching a new product, but they are not a proxy for market share.”7 In fact, conjoint analysis has

_________________________________

5 Green, P.E., A.M. Krieger, and Y. Wind, “Thirty Years of Conjoint Analysis: Reflections and Prospects,” Interfaces, Vol. 31, No. 3, Part 2 of 2, May-June 2001, pp. S56-S73 at S67.

6 Green, P.E., A.M. Krieger, and Y. Wind, “Thirty Years of Conjoint Analysis: Reflections and Prospects,” Interfaces, Vol. 31, No. 3, Part 2 of 2, May-June 2001, pp. S56-S73 at S67.

7 Google’s Motion to Strike, p. 16.

5


been used by firms and other organizations in deciding the specific design features of multi-million dollar projects as it is widely recognized as a reliable tool for predicting consumer behavior:

  • Marriott used conjoint analysis to design Courtyard Marriott hotels, “an ‘optimal’ hotel chain catering primarily to business travelers who had no need for many of the features provided by up-scale hotels, such as Marriott and Hyatt.”8 Most of the design recommendations from the conjoint analysis were used to create the Courtyard Marriott and by 2001, there were 450 Courtyard Marriott locations worldwide “with annual sales in the billions of dollars.”9 Marriott has continued to use conjoint analysis in other projects, such as “designing time-share vacation units and in room and amenities pricing.”10
  • General Motors Corporation has used conjoint analysis since the early 1970s. Conjoint analysis has been used for “products such as the Cadillac Northstar engine, OnStar, XM Radio, Bumper to Bumper warranty, as well as many vehicles, such as the Chevrolet Avalanche.”11
  • Boeing Employees Credit Union, which is the fifth largest credit union in the United States with 385,000 members, used conjoint analysis to estimate members’ preference for money market accounts versus regular saving accounts.12
18. Additionally, the external validity of conjoint analysis has been demonstrated in several circumstances. Several applications in industry, in which conjoint analysis has been applied,

____________________________________

8 Green, P.E., A.M. Krieger, and Y. Wind, “Thirty Years of Conjoint Analysis: Reflections and Prospects,” Interfaces, Vol. 31, No. 3, Part 2 of 2, May-June 2001, pp. S56-S73 at S67.

9 Green, P.E., A.M. Krieger, and Y. Wind, “Thirty Years of Conjoint Analysis: Reflections and Prospects,” Interfaces, Vol. 31, No. 3, Part 2 of 2, May-June 2001, pp. S56-S73 at S68.

10 Green, P.E., A.M. Krieger, and Y. Wind, “Thirty Years of Conjoint Analysis: Reflections and Prospects,” Interfaces, Vol. 31, No. 3, Part 2 of 2, May-June 2001, pp. S56-S73 at S68.

11 Orme, B., Getting Started with Conjoint Analysis: Strategies for Product Design and Pricing Research , Second Edition, Research Publishers LLC, 2010, p. 130.

12 Orme, B., Getting Started with Conjoint Analysis: Strategies for Product Design and Pricing Research , Second Edition, Research Publishers LLC, 2010, p. 132.

6


have confirmed ex-post that the predictions from conjoint analyses were close to how the market actually behaved after the respective products had been introduced:

  • EZ-Pass toll collection project used conjoint analysis to determine how EZ-Pass should be configured and what level of resources should be allocated to its implementation: “The overall (at equilibrium) forecast made in 1992 of ‘take rate’ was 49-percent usage. The actual take rate (seven years later) was 44 percent; future usage is expected to be higher than 49 percent.”13
  • “Robinson (1980) reports a multinational conjoint study of North Atlantic air travel involving airfares, discounts, and travel restrictions. His results indicate that conjoint analysis had a substantial ability to predict market shares. Srinivasan et al. (1981) describe a conjoint study of individually computed conjoint functions that are used to predict work-trip modes (auto, public transit, and car pool). Travel mode shifts were forecasted for various policy-level gasoline tax surcharges. The authors’ forecasted results were roughly consistent with the actual subsequent increase in transit ridership resulting from a serendipitous rise in the price of gasoline. Benbenisty (1983) describes a conjoint study involving AT&T’s entry into the data terminal market. The simulator forecasted a share of 8% for AT&T four years after launch. The actual share was just under 8%.”14
  • Sunbeam Appliance Co. used conjoint analysis in conjunction with market research and product line simulations in redesigning its product lines. “The correlation between the predicted brand shares and the reported shares was very high (r = .96), which builds confidence in the quality of the market share predictions produced from the conjoint based simulations.”15
__________________________________

13 Green, P.E., A.M. Krieger, and Y. Wind, “Thirty Years of Conjoint Analysis: Reflections and Prospects,” Interfaces, Vol. 31, No. 3, Part 2 of 2, May-June 2001, pp. S56-S73 at S68.

14 Green, P.E. and V. Srinivasan, “Conjoint Analysis in Marketing: New Developments with Implications for Research and Practice,” Journal of Marketing, October 1990, pp. 3-19 at 13.

15 Page, Albert L. and Harold F. Rosenbaum, “Redesigning Product Lines With Conjoint Analysis: How Sunbeam Does It,” Journal of Product Innovation Management, Vol. 4, 1987, pp. 120-137 at 120.

7


D. Google misunderstands the design of my survey.
19. Google claims that “Dr. Shugan used respondents’ selections to rank and measure the relative importance of the seven features to consumers. He then plugged these ranked values— referred to in conjoint parlance as ‘partworths’—into a statistical software program in order to assess general consumer preference for an Android phone lacking the application volume, startup time, and multitasking capabilities allegedly provided by the patents- and copyrights-in-suit.” 16

20. Google misunderstands and misstates the analysis that I conducted. I calculated partworths using the most widely recognized and used regression-based statistical model for evaluating respondent choices in the framework of choice-based-conjoint. These partworths are not “ranked values,” rather they are coefficients from the Hierarchical Bayes regression estimation based on the actual “choices” respondents made in the conjoint exercise. This misstatement further demonstrates that Google’s counsel and experts do not understand the fundamentals of choice-based conjoint analysis. The “partworths” are not plugged “into any statistical software,” they are the coefficient estimates generated through regression analysis based on the data from the choices made by respondents over 16 choice tasks. They are not generated from simple rankings or from the assignment of points to various features or feature levels. Conjoint analysis is used because it allows one to analyze how features contribute to actual decision making when the relevant set of features are considered jointly.17

E. The selection of features for the conjoint survey was methodologically proper.
21. Google attacks the choice of features in my survey. Google’s attack mischaracterizes my analysis and misses the point of conjoint analysis.

22. First, Google falsely characterizes my description of how I selected the features and feature levels to be included in my 2011 Smartphone Survey. While I clearly included the features that were relevant to Oracle’s case and to Professor Cockburn’s analyses, I first confirmed that these

__________________________________

16 Google’s Motion to Strike, p. 17.

17 Green, Paul E., Abba M. Krieger, Yoram Wind, “Thirty Years of Conjoint Analysis: Reflections and Prospects,” Interfaces, Vol. 31, No. 3, Part 2 of 2, 2001, pp. S56-S73.

8


features were relevant to consumer choice. I testified in deposition that I relied upon several sources in making his decision of which features and feature levels to include:

Q. Do you recall who communicated to you specific features that ought to be included in the conjoint analysis?

A. Well, the -- your question is not really clear in the sense that there are different features in the analysis. Now, some of the features were communicated to me through Analysis Group that they were required features and need to be there. Other features I decided should be there, and there were other features that Cockburn decided needed to be there. And then in the end, I put it all together and decided which ones to actually include in the analysis. So the -- there wasn’t one source where all of the features came from.18 (emphasis added.)

23. Second, as I explained in the Shugan Report, Reply Report, and deposition, in addition to speaking with Professor Cockburn, I considered interviews, market research, and a focus group to identify the relevant features and their levels to be included in his 2011 Smartphone Survey. Exploratory interviews determined what product characteristics matter to consumers’ smartphone selections and the appropriate vocabulary for the questionnaire. I explained the purpose of the exploratory interviews in deposition:

Q. BY MR. PURCELL Did you formulate a list of questions that then went into the one-on-one interviews?

A. Yeah, basically the major use of that research was to identify the features that consumers would likely use in choosing between smartphones. And specifically my goal in this was to try at the end of the process to come up with the most important features so that we could identify the relative importance of the features involved in the case, that is, the ones involved in the patent and copyright infringement and how those -- important those would be compared to the important features. We wouldn’t be really that concerned about how important

_______________________________

18 Deposition of Steven M. Shugan, Ph.D., September 26, 2011, p. 29.

9


they were compared to the unimportant features, but important to the relevant features, and then be sure to include the least to most important of the features in the conjoint analysis, mainly, as I say, to explain variants and decrease the variability of the estimates more than it is to change the estimates. The other objective, the main objective was to -- in the focus groups, to be able to provide aided recall measures, where you actually tell people -- I think we had -- in different categories, we had people that, you know, that we asked them, you know, if you were considering this phone, would you consider the attribute, and then if you were a purchasing decision, would that be something that would be involved in a purchase decision. And then the aided recall requires certain -- you go into the focus group with sort of some attributes that you could cue people on. But first they were asked the unaided before the aided.19 (emphasis added.)
24. I complemented this qualitative research with a review and analysis of third-party sources that provided further information about the attributes that consumers consider when selecting a smartphone and confirmed that the feature enhancements enabled by the patents-in-suit and the Java copyrights are relevant to their decisions. Specifically, I reviewed industry analyst reports and buyers’ guides to determine how manufacturers differentiated their smartphones and underlying operating systems and which product features were emphasized in product reviews.20

Q. What was done to prepare for the conduct of that focus group?

A. To prepare for the focus group, there was background research done by myself and Analysis Group looking at the literature and publicly available information on smartphone purchases, specifically looking for the attributes that

________________________________

19 Deposition of Steven M. Shugan, Ph.D., September 26, 2011, pp. 40-42.

20 See for example, “J.D. Power and Associates Reports: Average Length of Time Wireless Customers Keep Their Mobile Phones Increases Notably,” J.D. Power and Associates, September 23, 2010; and “Top Reasons for Choosing Device – Smartphone vs. Feature Phone: What Were the Top 3 Reasons for Selecting Your Current Cell Phone?” The Nielsen Company, 2011, http://www.fiercemobilecontent.com/pages/nielsen-top-reasonschoosing-device-smartphone-vsfeature- phone.

10


were identified by either reviewers of smartphones, manufacturers of smartphones in their advertising, the consumers, any surveys that have taken place, and I believe it was also some quality -- companies that were looking at the qualities of various phones, the information, or a good sample of that information, including expert report. And then I also had some personal knowledge, as other people did, of smartphones, given that the category is a fairly common one. And then the focus group was set up. It was also -- before the focus group, it was done one-on-one in reviews with customers to sort of look at their purchasing decision. People who have recently purchased phones and what attributes they considered. I think there were seven one-on-one decisions within six people in the focus group. The focus -- the one-on-one decisions and focus group decisions are complementary in the sense that their goal was to identify an exhaustive set of features with aided and unaided recall, unaided recall being consumers just being asked what they considered, and aided recall being here are some features, which ones of these did you consider. And then that was sort of the preparation for the focus group.21 (emphasis added.)
25. Additionally, I did, in fact, include several features in my survey that are not at issue in this litigation. Importantly, it is not necessary in conjoint analysis to test every feature that may matter to consumers, because conjoint analysis assesses relative importance. I took this additional step to estimate a well-specified model. After I identified 36 features that real-world consumers said were important in making a purchasing decision, I intentionally, and conservatively, included in my survey the two features that consumers cared about most: brand and price, among other features consumers in the focus groups indicated that they required. This approach is methodologically sound, as I included the two most critical features that generally derive the greatest value in any estimation and that capture the benefits of Google’s reputation. Further, including brand in 2011 is conservative, as it allows Google to benefit from Android’s reputation for feature-rich devices with

___________________________________

21 Deposition of Steven M. Shugan, Ph.D., September 26, 2011, pp. 39-40.

11


broad applicability (i.e., wide availability of applications) built, in part, off of its infringement of the patents and copyrights in suit.

F. Google does not understand the difference between choice-based conjoint and self-
explicated approaches.
26. Google quotes a comment made by Judge Posner in the Apple v. Motorola case regarding the use of consumer surveys to measure the value of a patent. The approach that Judge Posner was commenting on is referred to as a self-explicated approach, where respondents are asked to evaluate the levels of each feature one-by-one and then asked to allocate a number of points, e.g., 100, across all features to represent their relative importance to the respondent. In contrast, the choice-based conjoint approach uses choice exercises that ask a respondent to perform repeated choice tasks. In each choice task, respondents are asked to choose their most preferred product from among several products (i.e., a set of products) where each product is described in terms of the product’s features. I did not use the approach that Judge Posner objected to in the Apple v. Motorola case. In fact, I testified in deposition that I considered using a self-explicated approach, but I dropped the approach in favor of using the choice-based conjoint approach.22 As noted above, a choice-based- conjoint study is optimal for this type of analyses, as it avoids the artificial “focus” that a selfexplicated method may require. By presenting product profiles characterized by both the features that are relevant to the copyright and patent allegations in this case and features that are generally considered by consumers when making decisions (e.g., brand and price), I did not draw specific attention to the features at issue in this case. They are presented as part of a broader product profile.

_____________________________________

22 Deposition of Steven M. Shugan, Ph.D., September 26, 2011, pp. 24-25.

12


G. I did evaluate the “incremental benefit” of phone features.
27. Google claims that my study “measures the value consumers place on certain phone features as a whole, rather than the incremental benefit to those features allegedly enabled by the technology at issue.” 23 (emphasis in original.) This statement is false.

28. Again, Google has failed to understand how my data and analyses function. The conjoint study created a “set of utilities or part-worths that quantify respondents’ preferences for each level of each attribute.”24 With this data on individuals’ partworths, I utilized a market simulator to simulate respondents’ market choices among a set of defined products characterized by both product features and the differentiated feature levels.25

29. Using the market simulator tool, I evaluated how preference shares for an Android device changed depending on whether the patented technologies were enabled (“Base case”) or disabled (“Scenario 1,” “Scenario 2,” and “Scenario 3”). (See Exhibit 3a to the Shugan Report.) By altering the levels of only one or two features at a time, the preference shares between the base case and various alternative scenarios could be compared to measure precisely the incremental benefit (or cost) of those feature level changes.26 According to Bryan Orme, this “one-attribute-at-a-time

___________________________________

23 Google’s Motion to Strike, pp. 18-19.

24 Orme, B., Getting Started with Conjoint Analysis: Strategies for Product Design and Pricing Research, Second Edition, Research Publishers LLC, 2010, p. 89.

25 “The simulator is used to convert raw conjoint (part-worth utility) data into something much more managerially useful: simulated market choices. Products can be introduced within a simulated market scenario and the simulator reports the percentage of respondents projected to choose each product. A market simulator lets an analyst or manager conduct what-if games to investigate issues such as new product design, product positioning, and pricing strategy.” (Orme, B., Getting Started with Conjoint Analysis: Strategies for Product Design and Pricing Research, Second Edition, Research Publishers LLC, 2010, pp. 89 and 91.)

26 “Conducting sensitivity analysis starts by simulating shares of choice among products in a base case market. Then, we change product characteristics one level at a time (holding all other attributes constant at base case levels). We run the market simulation repeatedly to capture the incremental effect of each attribute level upon product choice.” (Orme, B., Getting Started with Conjoint Analysis: Strategies for Product Design and Pricing Research, Second Edition, Research Publishers LLC, 2010, p. 81.)

13


approach to sensitivity analysis provides a good way to assess relative preferences of product attributes.”27

30. Contrary to Google’s assertions, the simulation results presented in my report, on which I understand Prof. Cockburn relied, do not measure the benefit of the at-issue features as a whole, as I do not consider Android phones with and without applications – I consider Android phones with varying speeds of application launch times and varying numbers of available applications.

31. Google may have confused the simulation results reported in Exhibits 3a – 4c of my Expert Report (on which I understand Prof. Cockburn relied) and the results reported in Exhibit 5 of my Expert Report which measure the relative importance of each feature as a whole. This particular analysis was not created for the purpose of estimating preference shares but rather to demonstrate stability across sample sensitivities; I understand that these results are not relied upon by Prof. Cockburn in his analysis.

H. I never “conceded” that the survey respondents did not hold levels constant.
32. Google claims that my instruction to survey respondents to “[a]ssume any features not listed are the same for all alternatives” “undermines the survey’s ability to predict real-world behavior.”28 This critique is not valid as researchers have found that respondents are able to consider such hypotheticals.29

33. Google argues that survey respondents did not hold constant those variables that were not part of the conjoint analysis’s choice task. Google misconstrues the following excerpt from my Reply Report to support its argument: “For example, respondents will tend to implicitly attribute to

____________________________________

27 Orme, B., Getting Started with Conjoint Analysis: Strategies for Product Design and Pricing Research, Second Edition, Research Publishers LLC, 2010, p. 84.

28 Expert Report of Steven M. Shugan, Appendix E, p. E-19; and Google’s Motion to Strike, p. 18.

29 See for example Wilhelm, W.B., “Encouraging Sustainable Consumption through Product Lifetime Extension: The Case of Mobile Phones,” International Journal of Business and Social Science, Vol. 3, No. 3, February 2012, pp. 17-32 at 22.

14


the brand name any excluded attributes.”30 Google mistakenly characterizes this statement as a concession that, contrary to the survey instructions, participants in my survey did not hold constant the non-included features that might be relevant to a smartphone purchase. That is not the case. When respondents implicitly attribute aspects of other attributes to brand name it is not inconsistent with holding constant all other variables that are not included in the conjoint study. There is no reason to believe that respondents who do enrich the value of the brand name with variables not included in the conjoint study vary their evaluation of brand between the 16 choice sets from which they choose their preferred smartphones. Therefore, even if respondents enrich the value of brand, I am still able to isolate the incremental benefit of the features at issue accurately. In other words, the meaning of price and brand may differ slightly for some consumers, but each individual consumer will have a constant view of the meaning of price and brand as they make their choices in the survey. Moreover, as I explained in my Reply Report, some of the value intrinsic to the patents and copyrights at issue as well as variables not included in the conjoint study may likely be captured in the Android brand feature level and would lead to my estimates of the value intrinsic to the patents and copyrights at issue to be understated.31 The same rationale applies to the price feature: should price capture any other attributes, such as quality, my market share simulation would lead to my estimates of the value intrinsic to the patents and copyrights at issue to be understated.

I. Google fails to show that respondents did not follow my instruction to hold constant
all unnamed features.
34. Google presents no support to demonstrate that survey respondents did not follow my instruction to hold any features not listed constant. In fact, the stability of my results and the accurate representation of real-world market shares by my preference shares in the base case is a strong signal

_______________________________

30 “Such an effect would increase the value of the Android brand name, and so such biases would reduce the value of the relevant features.” (Reply Report, Exhibit A, pp. 17.)

31 Shugan Report, pp. 16-17.

15


that the data was collected without deviating from scientific standards and resulted in reliable estimates.32

35. Google ignores this external validation, however, and instead states that “almost one quarter of all respondents claimed that they would prefer a smartphone costing $200 to a putatively identical smartphone costing $100.”33 (emphasis added.) What Google offers as “proof in the pudding”34 is not only technically wrong, but it is also a gross misstatement of my survey. First, I did not use any open-ended question that would have allowed respondents to express their preferences directly. Consequently, no respondent “claimed” to prefer a smartphone costing $200 to a similar one that costs $100. Second, my survey purposefully avoids statements in which respondents explain their preferences, opinions, or perceptions. Instead, respondents’ choices are used in a statistical, regression-based model to generate a distribution of partworths that finally – in their entirety – are used to predict preference shares in a simulated market. Notably, such preference shares have been shown in many cases to reflect market shares; and if they do so, they offer a statistically valid and reliable description of a market in either the real world or in a but-for world. In this case, the market shares were within a rounding error of the real world market shares, suggesting that respondents understood the instructions and were able to “hold all else constant.” The evidence Google supposedly provides is based on flawed analyses and flawed logic.

36. Google claims that “[n]o rational person, much less 24% of all rational people, would prefer to pay $200 for a phone they could have for $100.”35 I responded to this argument in my

___________________________________

32 Kellogg, Don, “In U.S. Market, New Smartphone Buyers Increasingly Embracing Android,” NielsenWire, September 26, 2011, http://blog.nielsen.com/nielsenwire/online_mobile/in-u-s-market-new-smartphone-buyersincreasinglyembracing- android/; and Exhibit 3a to Expert Report of Professor Steven M. Shugan. “One, I forecasted or predicted what the actual market shares could be from the conjoint analysis and found that the observed shares in the market were very close to what the conjoint analysis predicts they should be.” (Deposition of Steven M. Shugan, Ph.D., September 26, 2011, p. 115).

33 Google’s Motion to Strike, p. 19.

34 Despite claiming to present a proof, Google presents a proposition that indicates a lack of understanding not only of the concept of (scientific) proof, but also of the estimation and the interpretation of the estimation of partworths in a choice model. (Google’s Motion to Strike, p. 19.)

35 Google’s Motion to Strike, p. 19.

16


declaration from November 1, 2011, as well as in my Reply Report, and I incorporate those documents herein by reference. (See Exhibit A.)

37. In the declaration to which I responded in November 2011, Dr. Leonard pointed out that some individual utility functions contradict what he touts as rational preferences.36 In doing so, Dr. Leonard errs in four ways.

38. First, it is critical to note that Dr. Leonard is confusing estimated parameters with actual data and is then making false inferences about estimated parameters rather than analyzing actual respondent choices.

39. Second, for the majority of the 24 percent of respondents the estimates for the two prices are so close that that a diligent statistician would consider the difference to be zero, rather than representing some form of irrationality. As explained in my Reply Report, I excluded respondents with utilities associated with $100 and $200 that are within one standard deviation of the difference in utilities between levels as a sensitivity analysis. When these respondents with such utility comparisons are excluded from the analysis, only 8.8 percent of respondents, not 24 percent, prefer a price of $200 over a price of $100.37

40. Third, what Dr. Leonard touts as “irrational,” is, in fact, commonly accepted in marketing and consumer research: Some consumers display a tendency to shy away from a cheaper product in favor for a higher price; they are price-insensitive within a reasonable range of prices.38

This fact is well documented in the literature and represents kinks in the demand curve.39

41. Fourth, the most crucial argument in the discussion of so-called irrational preferences, Google’s experts misunderstand how the Bayesian approach informs individual estimates based on

____________________________________

36 Declaration in Support of Google’s Opposition to Motion to Exclude Portions of the Expert Reports of Gregory K. Leonard and Alan J. Cox, October 28, 2011, p. 24.

37 See Reply Report, p. 19.

38 See, for example, Mohr, Jakki, J., Sanjit Sengupta, and Stanley F. Slater, Marketing of High-Technology Products and Innovations, Third Edition, New Jersey: Pearson Prentice Hall, 2010; and “Smartphones: Building Profitability and Loyalty in the Mass-market,” WDSGlobal, 2010, http://www.wds.co/enlightened/smartphone_profitability/smartphone_profitability_and_loyalty_wdsglobal.pdf.

39 Lambert, Z.V., “Product Perception: An Important Variable in Price Strategy,” The Journal of Marketing, Vol. 34, No. 4, October 1970, pp. 68-71, pp. 68-71.

17


aggregate population parameters during the multi-stage estimation process and that one should not evaluate isolated respondent estimates, as these methods are intended to characterize populations, not individuals. In these critiques, Google’s experts are emphasizing outliers (which have little weight in the analysis), rather than recognizing the close approximation or fit of my results with actual market data.

42. By using such examples from individual respondents, Google’s experts purposefully focus on individual parameters, and therefore ignore the idea behind a Hierarchical Bayesian model, such as the one that I use for my estimations. Specifically, Google’s experts forget or ignore that while one could very well estimate individual parameters by fitting the individual respondent’s preferences, the aggregate results would almost always suffer from overfitting of the data.40 To avoid this overfitting in favor of precise aggregate results, the chosen Hierarchical Bayes approach allows individual parameters to be influenced by information from the whole sample. Hence, while some individual parameters might reflect small amounts of statistical noise, the aggregate estimation outcome is precise; in fact, it is considered the gold-standard in choice modeling by many marketing experts.41 I have stated in deposition that it is the aggregate outcome of the sample, not individual values, that matters for my analysis: “[t]he analysis was done at the sample level, and so we looked at all the – the estimation of all the individual responses of all the individual respondents. I didn’t examine individual questionnaires.”42

43. Furthermore, given that this is an issue involving estimation and not the quality of the data, it can easily be eliminated by modifying the estimation. If one tried to account for Dr. Leonard’s erroneous proposition that “No rational person, much less 24% of all rational people,

____________________________________

40 Overfitting describes a model’s characteristic to fit particular data too precisely and therefore lose its predictive power.

41 Allenby, G. and P.E. Rossi, “Perspectives Based on 10 Years of HB in Marketing Research, Sawtooth Software Research Paper Series, 2003, http://www.sawtoothsoftware.com/download/techpap/allenby.pdf.

42 Deposition of Steven M. Shugan, Ph.D., September 26, 2011, pp. 138-139.

18


would prefer to pay $200 for a phone they could have for $100,”43 the outcome of a model using monotonic preferences for price yields similar results to those reported in the Shugan Report.44

44. Finally, it is not scientific to conclude that my survey is flawed simply because its results do not confirm Google’s assumptions about consumer behavior. I have pointed out this flaw in Dr. Leonard’s reasoning before. It is axiomatic in science that one should not substitute one’s own view of what is correct for what the data show.45

DATED: February 24, 2012

/s/ Steven M. Shugan
STEVEN M. SHUGAN

____________________________________

43 Google’s Motion to Strike, p. 20.

44 “According to these results, Android sales but-for the feature enhancements enabled by the patents-in-suit and the Java copyrights would have been at least 7.6 (instead of 7.9) percent lower if availability of applications was reduced and at least 20.0 (instead of 19.9) percent lower if application startup time was increased. Collectively, Android sales would have been at least 25.7 (compared to 25.7) percent lower but-for these infringements.” (Reply Report, p. 20.)

45 Lehmann, D.R., S. Gupta, and J.H. Steckel, Marketing Research, Massachusetts, Addison Wesley, 1998, p. 68.

19


ATTESTATION OF FILER

I, Steven C. Holtzman, have obtained Dr. Steven Shugan’s concurrence to file this document on his behalf.

Dated: February 24, 2012

BOIES, SCHILLER & FLEXNER LLP

By: /s/ Steven C. Holtzman
Steven C. Holtzman

Attorneys for Plaintiff
ORACLE AMERICA, INC.

20



745

UNITED STATES DISTRICT COURT
NORTHERN DISTRICT OF CALIFORNIA
SAN FRANCISCO DIVISION

ORACLE AMERICA, INC.
Plaintiff,
v.
GOOGLE, INC.
Defendant.

Case No. CV 10-03561 WHA

DECLARATION OF HINKMOND WONG
IN SUPPORT OF ORACLE AMERICA,
INC.’S OPPOSITION TO GOOGLE’S
MOTION TO STRIKE PORTIONS OF
THIRD EXPERT REPORT

Dept.: Courtroom 8, 19th Floor
Judge: Honorable William H. Alsup


I, HINKMOND WONG declare as follows:

1. I am an employee of Oracle America, Inc. (“Oracle”). My title is Consulting Member of the Technical Staff.

2. I have personal knowledge of the facts set forth herein. If called upon to testify, I could and would testify as follows.

3. I hold a bachelor’s degree in electrical engineering from the University of Michigan and a master’s degree in computer engineering from Santa Clara University. I have been working as a software engineer since 1989.

4. I have been employed as an engineer with Sun Microsystems, Inc. (“Sun”), now Oracle, since approximately 1994. Through my years of working at Sun, I have become very familiar with the Java technology, and I have helped to design and improve the Java platform as part of my regular work responsibilities at various points throughout my tenure at Sun, now Oracle. I now work primarily with Java ME technology, including the editions of Java that run on smartphones, cell phones, and embedded devices. I am the inventor of several patents related to Java technology. In addition, my work at Sun and now Oracle has included development of a smartphone platform for a third party customer. As a result of my work over the years, I am quite familiar with and knowledgeable about the technical requirements of mobile devices in general and smartphones in particular.

5. For example, in addition to my ordinary work with Java ME, I was one of the engineers involved in the actual Sun-Google negotiations in the 2005–2006 time frame, when I was involved in planning the technical specifications, as well as estimating the work effort involved for some of the tasks that were required, for the mobile platform that the parties jointly considered developing and marketing. At that time, I was also responsible for advising Sun executives on the technical components that Sun could offer and that were essential to a smartphone platform such as Android. In the course of those negotiations, I participated in technical meetings with Google engineers to determine what Google’s requirements were for the Android platform and what portions of Sun’s Java technologies would be important in that effort. I also participated in face to face meetings with Google’s Andy Rubin concerning the negotiations.

1


6. I was therefore employed at Sun, and knowledgeable about the Java technology, in the 2006 time frame, when Sun and Google were actually negotiating for a Java license for Android.

7. In addition to the experience I outlined above, I have direct experience with a number of Java patents, and have direct experience with the performance benefits that certain Java patents provide. I have ranked patents in terms of their engineering performance before.

8. I, along with four of my colleagues at Oracle, was asked by counsel for Oracle to conduct an analysis of certain patents held by Sun Microsystems, Inc. in the spring of 2006. Specifically, counsel requested that we determine which of Sun’s Java patents would have been potentially relevant to a smartphone platform in 2006, and then determine which of those patents would have been expected, from an engineering perspective, to provide the greatest benefits to such a platform.

9. Along with the other engineers, I concluded that we could provide the analysis requested by counsel through the following steps:

  • Identify relevant “blocks” of technology corresponding to functions that would be expected to be useful to a smartphone platform in 2006;
  • Identify what subset of Sun’s patent portfolio would have been useful to a smartphone platform in 2006;
  • Assign those potentially relevant patents to the technology blocks;
  • Rank the relative importance of those technology blocks; and
  • Rate each of the patents on a three-point scale.
10. I was not asked to provide any economic valuation of the patents in suit. Instead, my analysis of “value” was confined to the engineering benefit that the patents would be expected to provide a smartphone platform.

11. Throughout the process, I understood that Dr. Mark Reinhold was the leader of the team. I understood that he would make all final decisions, and that I was helping him to review the patents in the interest of time and to ensure that the final analysis was as accurate and comprehensive as possible.

2


12. Dr. Reinhold asked me to begin working on this project on or about January 24, 2012. I had not done any work for Prof. Cockburn prior to that date.

13. At the outset of the process, I understand that George Simion of Oracle ran searches through Sun’s patent databases to acquire a list of over 1,300 Java related patents that we were to review. I reviewed this list to make sure that it included all of the patents that I would expect to see, and confirmed that it did.

14. In the first step of the process, our team, led by Dr. Reinhold, identified 22 technology blocks that would have been relevant to a smartphone platform in 2006. We brainstormed the list of logical functional components on or about January 24, 2012; over the course of the next few days, we refined that list to come up with the final list of 22 blocks.

15. In order to contribute to this exercise, I used my personal experience, expertise, understanding of Google’s objectives in 2006 based on a Product Requirements Document that I understand Google provided to Sun in 2006, and knowledge of the Java platform. I am confident that the 22 technology blocks that we identified represent the full range of Java technology that would have been relevant to a smartphone in 2006.

16. In the second step of the process, our team as a whole, led by Dr. Reinhold, reviewed every one of the patents in the list of 1,300+ patents captured by the searches discussed above to determine which patents would have been relevant to a smartphone platform in 2006.

  • Our team divided up the group of 1,300+ patents to ensure that every patent was evaluated by at least one person. We could not all look at every patent because of limitations on time. I personally looked at each patent I was assigned to evaluate with care, and I have no reason to believe my colleagues did not do the same.
  • We reviewed the patents by looking at the titles, abstracts, descriptions, application dates, and inventor names. When we believed it would be useful to do so, we also reviewed the specifications and the claims by retrieving the patent from the USPTO web site. This is a reasonable way to ascertain a patent’s rough usefulness in a smartphone, because, in my experience, the abstract and description are written by engineers, will provide the most useful information to another engineer, and will

3


    explain the general purpose of a patent and its claimed invention with enough specificity to understand its scope, application, and potential advantages. In contrast, in my experience information in the claims is less useful.

  • Although reviewing and categorizing the patents was time-consuming, it was not hard. Patents filed by Java engineers relating to improvements in Java technology use familiar terms, and in many cases I myself or another member of the team had direct experience with the invention or implementations of the invention.
  • I finished reviewing my portion of the 1,300+ patents on or about January 26, 2012.
  • When we had finished classifying each patent we were assigned to review, the members of the team who had special expertise with one or more of the technology blocks reviewed and confirmed the accuracy of our categorization. We discussed any inconsistently classified patents, and discussed any patents that I or one of my colleagues had indicated needed further attention or discussion.
17. At the end of this part of the process, I again reviewed the final list of all patents that could have been relevant in a smartphone platform in 2006. The final responsibility for deciding on the list of potentially relevant patents fell to Dr. Reinhold. After this process, Dr. Reinhold determined, and I agreed, that 569 patents out of the original 1,300+ would, in fact, be potentially relevant to a smartphone platform in 2006, and had classified each of those patents into one of the 22 groups.

18. In the third step of the process, our team, led by Dr. Reinhold, ranked the 22 technology groups. We distinguished among them by determining the benefits they would be expected to provide a smartphone platform in terms of speed, startup, footprint (i.e. memory requirements), and security.

  • We determined that these four criteria were reasonable criteria on which to rank the patents because software patents in the Java portfolio are almost always designed to provide one of these benefits. In my experience, and based on my knowledge of the Java platform, if I were to consider what benefits I would want to provide a Java-based

4


    smartphone in 2006, these are the criteria I would choose. These are the criteria that we routinely used at Sun and now at Oracle to evaluate our own Java implementations.

  • We ranked these groups independent of the patents that were contained within them.
  • Based on our ranking system, two or more groups occasionally tied.
19. We had completed this process on or about January 31, 2012.

20. In the fourth step of the process, our team, led by Dr. Reinhold, reviewed the specific patents, and evaluated each patent on a three-point scale, in which the best score was a 1. The ratings reflected the benefit that our team, led by Dr. Reinhold, would have expected the patented inventions to provide a smartphone platform in 2006. The top rating, a 1, was reserved for those patents that were either required by the Java platform specification for compatibility, or that would bring an order of magnitude improvement to the key metrics of speed, startup, footprint, or security. The middle rating, 2, was assigned to patents that would bring a significant improvement to the key metrics of speed, startup, footprint, or security. The bottom rating, 3, was assigned to patents that would have been relevant but would not have provided the benefits of a 1 or 2 rated patent.

21. By the time we began to rate the patents, we had reviewed many of them multiple times, in addition to our pre-existing familiarity with the underlying inventions. Again, I am confident that we had sufficient information to make an informed engineering assessment of the likely benefits of each patent. I and other senior engineers at Oracle, including each of the other four engineers on this assignment, regularly are called upon to make assessments of proposed innovations and improvements to Java technology. I and other engineers frequently make those assessments using information that is no more detailed than what is disclosed in the patent abstracts and descriptions. In addition, I was able to apply my own knowledge of how the inventions had been implemented, and the engineering benefits (or lack thereof) that resulted. The other engineers often had similar experiences that they shared with the team as we reviewed the patents and that informed our collective assessment. I agreed with all of the ratings that we settled on as a group.

22. In summary, we concluded:

  • A group of 569 patents would be relevant to a smartphone in 2006.
  • Those 569 patents fit into 22 technology groups.

5


  • The most important technology group is the Boot group.
  • There were seven patents ranked as 1 in the Boot group.
  • The second most important technology group is the JIT group.
  • There were twelve patents ranked as 1 in the JIT group.
  • The third most important technology group is the interpreter group.
  • There were three patents ranked as 1 in the interpreter group.
  • It would not be possible, based solely on engineering considerations that would be knowable in 2006, to say which of the 22 patents we identified in the top three groups was the most or least valuable to a smartphone platform such as Android.
23. I believe that both the process we employed, and the results we came to, are reasonable and accurate. If I were on a team assigned to design a Java-based smartphone platform in 2006, the three most important groups of technology in terms of startup, speed, and footprint would have been boot, JIT, and interpreter.

24. I am confident that the team had sufficient time and information to ascertain the correct rating for each patent, and that we had the collective expertise with the Java platform to understand the relative importance of every one of the technology groups in a smartphone platform in 2006 and every one of the patents in the list of 569. I do not believe that there is anyone at Oracle that we could have added to the team who would have improved the accuracy of the process or the results.

25. Although I have some knowledge of the Android platform, and I have assisted lawyers in analyses to determine whether Android infringes Oracle’s patents, neither aspect of that work had any effect on my work for this assignment. My assistance in that regard was not confined to the seven patents on which Oracle has sued Google – I had looked at numerous other Java patents as well. I understand that Google has also claimed in papers filed with the Court I and my colleagues are “the very engineers who selected litigation patents at the outset.” I know that statement to be false with respect to me. I have never selected any patents for this litigation or any other.

26. I understand that Google has also claimed in papers filed with the Court that I and my colleagues “admitted in deposition that they spent next to no time compiling their rankings and were

6


influenced by their prior work in this case.” I understand that Google also asserts in its papers that I and my colleagues “favor[ed] the patents they had already analyzed as part of this case.” I know that all of those assertions are entirely untrue with regard to me, and are contrary to everything I observed as I watched my colleagues perform their work alongside me. Together we spent significant time categorizing, evaluating, and rating the patents, and applied decades of directly relevant engineering experience to do so. There was sufficient time to do the analysis that we were asked to do, and to do so in a reliable and responsible manner. I did not “favor” any of the asserted patents for any reason related to the litigation, and I was not “influenced” by my prior work in the litigation. The assistance I provided to help determine whether Android infringes Java patents did not cause me to miscategorize any patent, nor did it cause me to rate any patent higher or lower than its technical merits warranted. I considered each patent and technology group based on the objective engineering benefit I would have expected it to provide to a smartphone platform in 2006, in light of my engineering experience and my knowledge of Java technology as described above.

I declare under penalty of perjury that the foregoing is true and correct.

DATED: February 23, 2012

/s/ Hinkmond Wong
HINKMOND WONG

7


ATTESTATION OF FILER

I, Steven C. Holtzman, have obtained Hinkmond Wong’s concurrence to file this document on his behalf.

Dated: February 24, 2012

BOIES, SCHILLER & FLEXNER LLP

By: /s/ Steven C. Holtzman
Steven C. Holtzman

Attorneys for Plaintiff
ORACLE AMERICA, INC.

8



748

[Keker & Van Nest LLP letterhead]

February 24, 2012

VIA ECF

Honorable William Alsup
U.S. District Court, Northern District of
California
Courtroom 8 - 19th Fl.
450 Golden Gate Avenue
San Francisco, CA 94102

Re: Oracle America, Inc. v. Google Inc., No. 3:10-cv-03561 WHA

Dear Judge Alsup:

In light of yesterday's order instructing counsel to keep the period from April 16 until late June available for trial, I am writing to re-notify Your Honor of two conflicting federal trial commitments on my schedule during that period. Both trials were set prior to the December 21, 2011 pretrial conference in this case, and they were set forth in the joint pretrial statement and discussed at the pretrial conference. In understand from Your Honor's remarks during th hearing that I am to notify the Court of any changes in this schedule.

I am presently scheduled to commence trial in Commonwealth Scientific and Industrial Research Organisation v. Lenovo, Inc. et al., Case No 6:09-cv-399 LED (Hon. Leonard Davis) on April 2, 2012. The trial is expected to last for approximately three weeks, and the pretrial conference is set to occur on March 22.

I am also set for trial in Genentech v. The Trustees of the University of Pennsylvania, 5:10-cv-02037 LHK (PSG) (Hon. Lucy Koh) on June 11, 2012. Judge Koh recently advised us that she expects the trial to last through the end of June, and she has scheduled a bench trial concerning certain issues on June 29, 2012. The pretiral conference in that case is schedule for May 30.

I am also planning to be out of the country from April 26 through May 12, but I understand from Your Honor's remarks during the pretrial conference that, if my conflicting trial commitments are vacated or reschedule, you might insist that trial proceed notwithstanding.


Honorable William Alsup
February 24, 2012
Page 2

I will notify the Court of any changes in this schedule, and I am continuing to hold the period September - December available, as Your Honor requested. In the meantime, if I have misunderstood Your Honor's instructions, please let me know.

Sincerely,

KEKER & VAN NEXT LLP

Robert A. Van Nest
Robert A. Van Nest


  View Printable Version


Groklaw © Copyright 2003-2013 Pamela Jones.
All trademarks and copyrights on this page are owned by their respective owners.
Comments are owned by the individual posters.

PJ's articles are licensed under a Creative Commons License. ( Details )