Memorandum in Opposition re [229] MOTION to Compel Production of Successor-Custodian Documents and Relevant Source Code filed by Google LLC. (Attachments: # (1) Exhibit A)(Reilly, Craig)
Page 1 PageID#
UNITED STATES DISTRICT COURT
EASTERN DISTRICT OF VIRGINIA
ALEXANDRIA DIVISION
UNITED STATES, et al.,
Plaintiffs,
vs.
No: 1:23-cv-00108-LMB-JFA
GOOGLE LLC,
Defendant.
GOOGLE LLC’S MEMORANDUM OF LAW IN OPPOSITION
TO PLAINTIFFS’ MOTION TO COMPEL PRODUCTION OF
SUCCESSOR CUSTODIAN DOCUMENTS AND SOURCE CODE
The Court should deny Plaintiffs’ Motion to Compel (the “Motion”) production of
documents that they now describe as key to their claims but which the DOJ failed to request
during its three-year-plus investigation of Google. Having failed to seek these documents when
they had the opportunity to do so, Plaintiffs now demand them within the compressed timeframe
for fact discovery. Plaintiffs should not be allowed to use this Court’s well-earned reputation for
speed and efficiency to unduly burden Google with discovery concerning information they did
not see fit to seek during their investigation.
Plaintiffs’ request that Google be compelled to add successor custodians in addition to
the 31 custodians that Google has already agreed to conditionally add should be denied.Plaintiffs represented to the Court that they “don’t intend to duplicate the efforts that have
As described below, Google’s agreement to add the 31 custodians was conditional on the
collection process resulting in a reasonable and proportionate universe of additional documents
to review. Page 2 PageID#
already occurred during the investigation phase, and we will be as efficient as possible in seeking
third-party requests and seeking a moderate amount of refreshing of data from Google as well,”
March 24, 2023 Tr. at 6:7-11, Dkt. 70, but Plaintiffs’ requests to date require Google to review
two million additional documents beyond what was already reviewed and produced in Plaintiffs’
pre-complaint investigation.
For the 31 custodians Google has voluntarily agreed to add,
Plaintiffs were able to articulate their connection to the claims and defenses in the case. But for
the successor custodians, Plaintiffs have not articulated how these individuals are relevant or
proportional to the needs of the case. Plaintiffs’ request should therefore be denied.
As for Plaintiffs’ motion to compel source code, Google has already agreed to search for
and make available (to the extent it exists) source code for the specific features and algorithms at
issue. Specifically, Google has agreed to make available nine of the twelve categories of source
code requested.2 Google has also proposed to conduct broad searches for technical explanatory
materials that Plaintiffs seek, and Plaintiffs have not challenged the sufficiency of those
proposals. The remaining three categories of source code identified by Plaintiffs do not relate to
any specific feature or algorithm. They are ill-defined, catch-all categories of source code that
are not susceptible to a reasonable search, and Plaintiffs have not explained how they are
relevant or necessary—much less proportional to the needs of the case.
BACKGROUND
This case comes on the heels of an expansive three-year-plus investigation into several of
Google’s businesses—including digital advertising—led by the United States Department of
Justice Antitrust Division (“DOJ”). From August 2019 to October 2020, the DOJ issued five
Civil Investigative Demands (“CID”) containing 73 requests. Following its filing of a lawsuit
Google has agreed to make source code available pursuant to the procedures described in
Appendix C of the Protective Order, Dkt. 203. Page 3 PageID#
regarding Google’s Search business in October 2020,3 DOJ continued its ad tech investigation
with another CID containing 37 broad requests related to Google’s advertising technology
products.
During that time, the DOJ had tremendous leverage over Google to demand
documents and information. In response to the DOJ’s CIDs, Google produced nearly 3 million
documents and the DOJ deposed at least 30 Google employees.
Plaintiffs filed their lawsuit in this case on January 24, 2023. On March 27, 2023,
Plaintiffs issued their First Set of Requests for Production (the “Requests”). Google timely
served its Objections on April 11, 2023 and its Responses on April 26, 2023. The parties have
since met and conferred on six separate occasions to discuss Plaintiffs’ Requests, including
Plaintiffs’ Requests for successor custodians and source code.
Plaintiffs’ Request for Successor Custodians
Along with its Objections, on April 11, 2023, Google provided Plaintiffs with a
comprehensive plan to meet its discovery obligations (the “Discovery Proposal”). Google’s
Discovery Proposal involved searching the files of 119 custodians over a ten year period—more
than had been negotiated and agreed to during the course of the DOJ’s expansive ad tech
investigation. As a result, Google is in the process of reviewing more than 2 million additional
documents to respond to Plaintiffs’ Requests.
On April 14, 2023, Plaintiffs responded to Google’s Discovery Proposal and requested
that Google: (1) add 15 identified custodians for specific reasons that DOJ provided, (2) identify
and add custodians with information relevant to Plaintiffs’ Federal Agency Advertiser claims, (3)
identify and add custodians from Google’s publisher partnership group, and (4) identify
successors to all of the 119 custodians proposed by Google.
See Compl., United States v. Google LLC, No. 1:20-cv-03010-APM (D.D.C. Oct 20, 2020),
ECF No. 1. Page 4 PageID#
Google agreed to the vast majority of Plaintiffs’ requests and, as a result, is currently in
the process of adding more than 30 custodians in response to Plaintiffs’ specific requests.
Google made clear to Plaintiffs that the addition of each incremental custodian is a
time-consuming process that can materially increase the population of over 2 million documents
already being reviewed by Google. But despite that burden, Google: (1) conditionally agreed to
add all 15 of the custodians specifically identified by Plaintiffs,4 (2) identified and conditionally
agreed to add 14 custodians likely to have relevant information on the sale of digital advertising
to Federal Agency Advertisers, and (3) identified and conditionally agreed to add 5 custodians
from Google’s publisher partnership group. In each case, Google’s agreement to add the
identified custodians was conditional on the collection process resulting in a reasonable and
proportionate universe of additional documents to review.
Given Google’s agreement to make a substantial custodial production, that leaves only
Plaintiffs’ fourth request—that Google identify post-October-2020 successors to all custodians included in Google’s Discovery Proposal—which is the subject of the instant Motion.
Google told Plaintiffs that their request to identify 119 successors is burdensome because Google
does not systematically maintain information on successors in the ordinary course of business,
and so the only way to address the request is through individualized manual inquiries. That is
because roles and responsibilities at Google are not static, reorganizations happen with some
frequency, and teams often redistribute work or direct reports as employees change teams or
depart the company.
Given this fluidity, there is no centralized system tracking successor
information.
Google intends to add one of those custodians for a limited time period because he left his role
in the pertinent business in 2014. Page 5 PageID#
In light of the burdensome nature of this request and the fact that the conduct at issue
occurred over a long period of time before 2020, Google asked Plaintiffs to narrow their request
and identify a priority set of custodians who are likely to have relevant information post
October-2020. Plaintiffs responded with a list of 63 custodians, but still did not explain why
post-October 2020 documents for these 63 individuals would be relevant to Plaintiffs’ claims.
Despite Plaintiffs’ failure to address relevance and the burden involved in responding, on
May 24, Google provided successor information for six senior employees that were the focus of
Plaintiffs’ queries, and stated that it would provide further information by the end of the week.
As promised, on May 26, Google provided additional information for about 39 of the 63 Google
employees identified in Plaintiffs’ May 10 email.5 This included confirming that either no
equivalent successor existed or the successor was already captured in the existing custodian list
for the vast majority of employees. Google also confirmed that it was continuing to investigate
other possible successors.
Just two hours later, and without further meet and confer, Plaintiffs filed the instant
Motion to “compel Google to complete its search for and produce relevant materials from the
custodial files of successor employees . . . or otherwise confirm that the employees are already
included on Google’s custodian list.” Mtn. at 3-4. The Motion does not assess relevance or
narrow the relief sought in light of the successor information that Google had already provided
prior to the Motion being filed.
Plaintiffs’ Requests for Source Code
Plaintiffs also requested source code and associated explanatory materials, which they
now characterize as being “at the Heart of This Case” and “crucial to allow Plaintiffs a fair
An anonymized version of the successor information Google provided on May 26 is provided
in Exhibit A to protect the privacy of non-party employees. Google has already provided
Plaintiffs with the non-anonymized information in its May 26, 2023 letter to K. Garcia. Page 6 PageID#
opportunity to prosecute their case and understand Google’s potential defenses.” Mtn. at 4-5.
But the DOJ never included similar requests for this “crucial” information during its
three-year-plus investigation.
Google timely served its Objections on April 11, 2023 and its Responses on April 26,
2023.
Notwithstanding its Objections, Google agreed to meet and confer with Plaintiffs
regarding all five Requests that are the subject of the instant Motion.
At the outset of the meet and confer process, Google told Plaintiffs that it was
considering Plaintiffs’ belated demands for source code in good faith, but explained that
Google’s source code is extremely sensitive, and that identifying code potentially responsive to
the Plaintiffs’ requests, some of which spans over 15 years, would be a burdensome process that
could take weeks to resolve. Google also explained that the process of identifying potentially
responsive lines of code (out of a much larger universe of source code) involved speaking with
multiple employees across Google.
To complicate matters further, Plaintiffs’ Requests for source code have been a moving
target.
After initially demanding both source code and pseudocode, on May 5 Plaintiffs
proposed that, in lieu of a production of source code, Google could instead produce a “complete
repository” of pseudocode, which Plaintiffs proposed to evaluate over a two-week period and
then inform Google whether (and to what extent) they would continue to press their request for
source code.
Google worked diligently to address Plaintiffs’ newly-revised Requests. On May Google informed Plaintiffs that it had been unable to identify any such “complete repository” of
pseudocode requested by Plaintiffs. In the same meet and confer, Google told Plaintiffs that it
was nearing the end of its inquiries regarding production of the source code itself and anticipated Page 7 PageID#
giving Plaintiffs a substantive response by May 30, 2023. Plaintiffs said that they would not wait
and planned to file a motion to compel.
Google asked that Plaintiffs refrain from filing any motion because the parties were not at
an impasse. Providing further context, Google explained to Plaintiffs that May 30 was the
deadline for substantial completion of production in the parallel proceedings in the Southern
District of New York and that Google was working diligently to respond in tandem to similar
requests for source code made by the MDL plaintiffs. Google specifically asked that—consistent
with the Coordination Order negotiated by the parties across the two cases (and agreed to by the
DOJ)—it be allowed to respond to these requests at the same time to make the process more
efficient. Plaintiffs refused.
On May 26, 2023, Google sent Plaintiffs a letter intended to resolve all remaining
disputes as to source code. Google explained that, subject to the provisions of the governing
Protective Order and associated Source Code Protocol, Google is prepared to search for and
make available source code sufficient to describe the following features, to the extent it exists
and can be located: Dynamic Allocation, Enhanced Dynamic Allocation, Reserve Price
Optimization, Dynamic Revenue Share, Minimum Bid to Win aka Highest Other Bid, Smart
Bidding, Poirot, and Bernanke (including Global Bernanke and Bell).
That letter concluded:
In light of Google’s extensive good faith cooperation in discovery, we think that
litigating these issues would be a waste of judicial resources. The fact that your
letter contemplates further meeting and conferring on these issues after the filing of
a motion to compel supports our view that the parties are not at an impasse and that
the disputes set forth in your letter are manufactured. Google remains willing to
meet and confer.
Plaintiffs did not meet and confer and instead filed this Motion. Page 8 PageID#
ARGUMENT
Given the schedule, and the DOJ’s commitment, made before this Court, to focus its
discovery requests on a refresh of the productions made during its investigation, the DOJ should
not be permitted now to leverage the speed and efficiency of this Court, impose burdensome
requests, and then prematurely request judicial intervention. Google has agreed to produce
responsive information consistent with the Court’s discovery deadlines, just not on the timetable
demanded.
Plaintiffs’ request for the additional successor custodians should be denied because
Plaintiffs have not tethered the roles of post-2020 successors to the pre-2020 conduct alleged in
the Complaint. Google has already agreed to add 31 more custodians at Plaintiffs’ specific
request when Plaintiffs were able to explain the relevance of the requested custodians. Plaintiffs
have already amassed nearly 3 million of Google’s documents and Google has agreed to review
over 2 million more drawn from the files of 150 custodians. Plaintiffs’ unsubstantiated assertion
that every possible successor must be “key” and added as a custodian is insufficient to show that
the discovery they demand is “proportional to the needs of the case.” Fed. R. Civ. P. 26. What
the DOJ described to this Court as a mere “refresh” cannot amount to detailed discovery on a
scale equal to or exceeding the 3-year pre-filing review already conducted.
Plaintiffs’ request for source code should also be denied. Source code is extremely
sensitive, and Plaintiffs are required to show that its production is both relevant and necessary to
the claims and defenses in the case. Notwithstanding this standard, Google has agreed to make
available source code for the specific features and algorithms that Plaintiffs have requested.
Google has also proposed comprehensive searches to capture the related explanatory material
sought by Plaintiffs, and Plaintiffs have not challenged the sufficiency of those searches. The Page 9 PageID#
only remaining categories of source code sought by Plaintiffs’ Motion are an ill-defined selection
of catch-alls as to which Plaintiffs made no effort to discharge their burden of showing relevance
and necessity. The Court should deny Plaintiffs’ Motion as it pertains to those sundry categories.
In filing their Motion two hours after Google sent a letter intended to put to rest all of
these issues, Plaintiffs have acted prematurely. This Court has set an “aggressive” schedule in
this case and encouraged the parties “make good use of the time that [they] have” and to “focus
[their] attention on certain things” because it will not be possible to “turn over every grain of
sand.” March 24, 2023 Tr. at 18:6-10, Dkt. 70. Plaintiffs’ Motion ignores that guidance.
I.
Plaintiffs Have Not Met Their Burden to Compel the Addition of Successor
Custodians
In its Discovery Proposal, Google identified 119 custodians that would form the basis of
its document review responding to Plaintiffs’ Requests.6 Those 119 custodians comprised over
100 that were negotiated and agreed with the DOJ during the course of its investigation, plus
additional custodians that Google identified to respond to Plaintiffs’ Requests. As noted above,
through the meet and confer process, Google has conditionally agreed to add—at Plaintiffs’
explicit request—31 further custodians for which Plaintiffs were able to articulate a concrete
nexus to the claims and defenses in the case. For any additional custodians, the burden rests with
It is a basic tenet of discovery that the responding party—here, Google—bears the initial
responsibility for determining which custodians should be searched for relevant documents. This
Court, like most around the United States, has endorsed the Sedona Principles. See Brink’s Co. v.
Chubb Eur. Grp. Ltd., No. 3:20-cv-520, 2021 WL 5083335, at *4 n.2 (E.D. Va. Feb. 24, 2021)
(citing the Sedona principles favorably in discussing Rule 26’s proportionality standard). Under
Sedona Principle 6, the “responding party is entitled to select the custodians most likely to
possess responsive information and to search the files of those individuals.” Mortg. Resol.
Servicing, LLC v. JPMorgan Chase Bank, N.A., No. 15 Civ. 0293 (LTS) (JCF), 2017 WL
2305398, at *2 (S.D.N.Y. May 18, 2017); see also F.D.I.C. v. Baldini, No. 12-7050, 2014 WL
1302479, at *8 (S.D.W. Va. Mar. 28, 2014) (Noting that Rule 34 “plainly contemplates that [the
responding party] must take[] steps initially to locate ESI responsive to each of [the] requests”
where the requesting party has “no familiarity or working knowledge respecting the ESI’s
responsiveness”). Page 10 PageID#
Plaintiffs to “demonstrate that the additional requested custodians would provide unique relevant
information by providing . . . evidence that there are unique responsive documents being missed
in the current search scheme that would justify the inclusion of additional custodians.” Coventry
Capital US LLC v. EEA Life Settlements Inc., No. 17 Civ. 7417 (VM) (SLC), 2020 WL 7383940,
at *6 (S.D.N.Y. Dec. 16, 2020) (internal quotation marks and citation omitted); accord In re
Facebook, Inc. Consumer Privacy User Profile Litig., No. 3:18-MD-02843-VC-JSC, 2021 WL
10282213, at *1 (N.D. Cal. Nov. 14, 2021); In re 3M Combat Arms Earplug Prod. Liab. Litig.,
No. 3:19-md-2885, 2020 WL 4501794, at *1 (N.D. Fla. Aug. 5, 2020).
Plaintiffs have not met that burden.
Their Motion does nothing to tie the
successor-custodians to the allegations in their Complaint beyond the blanket assertion that they
must be “key” simply because they took over a role that may have been relevant to an allegation
of historical conduct. See Mtn at 2. That is insufficient under Federal Rule of Civil Procedure
26(b)(1), because “[r]elevance is . . . the foundation for any request for production.” Cook v.
Howard, 484 Fed. App’x. 805, 812 (4th Cir. 2012); see also Blankenship v. Fox News Network,
No. 19-cv-00236, 2020 WL 9718873, at *18 (S.D. W.Va. Sept. 21, 2020) (denying in part
plaintiff’s motion to compel discovery from individuals organized by job titles because plaintiff
failed to demonstrate relevance of those individuals to their claims).
II.
Google Had Already Agreed to Search For and Provide All Relevant Source Code
Prior to Plaintiffs Filing Their Motion
Plaintiffs’ Motion should be denied as moot to the extent that it seeks to compel
production of source code regarding specific features and algorithms referenced in the
Complaint. That is because Google had (prior to Plaintiffs filing their Motion) agreed to search
for and make available that source code to the extent that it exists. Plaintiffs’ Motion should be
denied to the extent that it seeks other broad and ill-defined categories of source code because Page 11 PageID#
Plaintiffs have failed to make the required showing of necessity (in addition to relevance) for
those broad, catch-all requests.
Federal Rule of Civil Procedure 26(c)(1)(G) allows the court, for good cause, to issue an
order “requiring that a trade secret or other confidential research, development, or commercial
information not be revealed or be revealed only in a specified way.” In order for the court to
apply the rule, two criteria must exist. First, the material sought to be protected must be “a trade
secret or other confidential research, development, or commercial information.”
Id. That
condition is satisfied here because it is well settled that source code constitutes a trade secret.
See, e.g., SDSE Networks, Inc. v. Mathur, 2022 WL 18539944, at *4-5 (E.D. Va. Dec. 28, 2022)
(finding that source code qualifies as a trade secret); Trandes Corp. v. Guy F. Atkinson Co., F.2d 655, 663 (4th Cir. 1993) (“[S]ource code can and does qualify as a trade secret.”). Google’s
source code is no exception.
Second, there must be a “good cause” basis for granting the request for disclosure. Fed.
R. Civ. P. 26. “[D]isclosure [of trade secrets] will be required” if “the information sought is
relevant and necessary.” Duplan Corp. v. Deering Milliken, Inc., 397 F. Supp. 1146, 1185-(D.S.C. 1975) (noting that “courts are loath to order disclosure of trade secrets absent a clear
showing of an immediate need”); see also Viacom Int’l Inc. v. Youtube Inc., 253 F.R.D. 256, (S.D.N.Y. 2008) (discussing the required showing by plaintiffs seeking production of source
code “before disclosure of so valuable and vulnerable an asset is compelled”). If confidential or
trade secret information is not clearly relevant, it is not an abuse of discretion to deny access
altogether. See, e.g., Keyes v. Lenoir Rhyne Coll., 552 F.2d 579 (4th Cir. 1977) (affirming denial
of access and noting that the district court must use discretion to balance one party’s need for
discovery against the other party’s need for confidentiality). Page 12 PageID#
Plaintiffs’ Motion seeks to compel production of documents, including source code,
responsive to the following Requests.
Plaintiffs’ Requests 39, 40, and Requests 39 seeks “[s]ource code repositories and related repositories, including training
data, launch or release dates, pseudocode, and the implementation status or roll-out associated
with that source code,” relating to twelve purported “features and any successors or updates to
those features.” Mtn at 4.
Of those twelve “features” listed in Request 39, the first nine
implicate specific features or algorithms identified in Plaintiffs’ Complaint, namely: “(1) Project
Bernanke (including Global Bernanke and Project Bell); (2) Project Poirot (also known as
optimized fixed CPM bidding and Automated Bidding); (3) AdX’s (sellside) dynamic revenue
share; (4) AdX’s predicted highest other bidder (pHOB) algorithm; (5) Smart Bidding,
including any inputs to and outputs from the feature or program, including but not limited to
minimum bid to win; (6) Dynamic Allocation; (7) Enhanced Dynamic Allocation; (8) Project
Bell; (9) Reserve Price Optimization (RPO).” Id. (emphasis added).Source Code
Prior to Plaintiffs filing their Motion, Google informed Plaintiffs that it would agree to
search for and make available source code sufficient to describe the following features, to the
extent it exists and can be located: Bernanke (including Global Bernanke and Bell), Poirot,
Dynamic Revenue Share, Minimum Bid to Win aka Highest Other Bid, Smart Bidding,
Dynamic Allocation, Enhanced Dynamic Allocation, and Reserve Price Optimization.
Plaintiffs’ motion should be denied as moot as to those categories because Google has agreed to
search for and make available this source code to the extent that it exists. See Romfo v.
Google does not understand the difference between Plaintiffs’ categories (1) and (8) to the
extent that they both refer to “Project Bell.” As set forth below, Google has agreed to search for
and provide source code for “Bell” to the extent that it exists. Page 13 PageID#
Scottsdale Ins. Co., No. 5:17-CV-422-D, 2019 WL 1210105, at *3 (E.D.N.C. Mar. 13, 2019)
(denying in part a motion to compel as moot where the responding party had “agreed to produce”
responsive documents “but had not done so at the time the motion was filed”).
By contrast, the remaining three categories of “features” in Plaintiffs’ Request 39 are not
features, but are ill-defined catch-alls that do not pertain to any specific feature or algorithm.
Rather, they seek “(10) any curation function for the Related Products that determine which
exchanges, DSPs, and Ad Networks to call; (11) any bid, price floor, or auction optimization
algorithm, product, or feature; and (12) any automated means or manner of bidding, whether on
individual bids or on a campaign-level basis.” Mtn. at 4. Plaintiffs have never articulated what,
specifically, they are seeking through these catch-all requests and their Motion provides no
further clues.As to these three remaining categories, Plaintiffs’ Motion fails because Plaintiffs have not
made the requisite showing of relevance and necessity. See, e.g., Keyes, 552 F.2d at 581 (noting
that district court must use discretion to balance one party’s need for discovery against the other
party’s need for confidentiality); see also In re Apple & AT & TM Antitrust Litig., No.
C–07–05152 JW (PVT), 2010 WL 1240295, at *3 (N.D. Cal. Mar. 26, 2010) (denying motion to
compel source code where plaintiffs had not established that the source code they sought was
“relevant and necessary” and observing that “[p]laintiffs only speculate that the additional source
code may be relevant”). What Google has already offered to produce is more than sufficient for
Plaintiffs’ purposes—Plaintiffs are “not entitled to the whole universe of source code . . . at issue
because . . . there are many functions within the relevant [programs] which have no bearing on
Plaintiffs assert that Google agreed to produce “certain but not all of the source code requested
by Plaintiffs.” Mtn. at 6. It is unclear what Plaintiffs believe they are missing, but to the extent
they believe something was omitted from the first nine categories of specific features or
algorithms, Google remains willing to meet and confer regarding the production of specific
identifiable source code in response to Plaintiffs’ requests. Page 14 PageID#
this case.”
Microsoft Corp. v. Multi-Tech Sys., Inc., No. 00-1412 (ADM/RLE), 2001 WL
37130770, at *7 (D. Minn. Dec. 14, 2001).
The need to make a clear showing of relevance is not relaxed, as Plaintiffs suggest,
simply because a protective order has been entered. See Mtn. at 10. A “protective order is not a
substitute for establishing relevance or need. Its purpose here is to prevent harm by limiting
disclosure of relevant and necessary information.” Micro Motion, Inc. v. Kane Steel Co., F.2d 1318, 1325 (Fed. Cir. 1990). As courts have recognized in analyzing the special protections
afforded to source code, “the protections set forth in [a] stipulated confidentiality order are
careful and extensive, but nevertheless not as safe as nondisclosure.” Viacom, 253 F.R.D. at 260.
Other Explanatory Materials
In addition to source code repositories, Request 39 also seeks “related repositories,
including training data, launch or release dates, pseudocode, and the implementation status or
roll-out associated with that source code.” Mtn. at 4. Similarly, Requests 40 and 41 seek the
following by reference to Request 39:
Request 40: All documents created since January 1, 2008, related to language
explanations, pseudocode, code repositories, design documents, A/B or comparison
testing, and any other manuals or resources explaining the features, functions, and
operation of the source code used to operate the features delineated in Request No.
39 and any updates to or successors of those features.
Request 41: Documents sufficient to show the inputs and outputs, including but not
limited to a list of any variable fields, categories of data, or Product Requirements
Documents (PRDs), to the source code and algorithms that are used in operating the
features delineated in Request No. 39 or successors or updates to those features.
Google has already agreed to conduct comprehensive searches to identify explanatory
documents responsive to Plaintiffs’ Requests 39-41. As summarized in the tables below, Google
has already proposed (and Plaintiffs have agreed to) expansive searches intended to capture
explanatory materials and other documents that relate to these Requests. In addition to the Page 15 PageID#
comprehensive search terms originally proposed by Google, Plaintiffs proposed 23 additional
search strings, and Google agreed to all of them.9 The tables below summarize agreed-upon
searches that address Plaintiffs’ Requests 39-41.Subcategory of Plaintiffs’ Requests 39-
Searches Capturing Plaintiffs’ Request
(1) Project Bernanke (including Global
Bernanke and Project Bell)
All documents that hit on “Bernanke” and
“Project Bell.”
(2) Project Poirot (also known as optimized
fixed CPM bidding and Automated
Bidding)
All documents that hit on “Poirot” or the
misspelling “Poriot.”
(3) AdX’s (sellside) dynamic revenue
share
All documents that hit on “DRS” as well as
documents concerning Dynamic Revenue
Share with appropriate modifiers.
(4) AdX’s predicted highest other bidder
(pHOB) algorithm
All documents that hit on “pHOB” or “HOB.”
(5) Smart Bidding, including any inputs to
and outputs from the feature or program,
including but not limited to minimum bid to
win
All documents that hit on “SmartBid” or
similar variations.
(6) Dynamic Allocation
Documents concerning dynamic allocation
with appropriate modifiers.
(7) Enhanced Dynamic Allocation
Documents concerning enhanced dynamic
allocation with appropriate modifiers.
(8) Project Bell
All documents that hit on “Project Bell.”
(9) Reserve Price Optimization (RPO)
Documents concerning RPO with appropriate
modifiers.
In its Objections to Requests 39-40, Google stated that Plaintiffs had failed to show that
production of source code was necessary “particularly given the extensive document production
made and contemplated in the MDL Expanded Investigation Materials describing the operation
of Google’s ad tech products.” In its Objections to Request 41, Google also pointed Plaintiffs to
its Discovery Proposal, which contemplates extensive custodial searches that will capture these
related materials.
The agreed searches are voluminous and we have summarized them here for the Court’s
convenience. Google can provide full text of the search terms if the Court would find them
helpful. Page 16 PageID#
Subcategory of Plaintiffs’ Requests 39-
Searches Capturing Plaintiffs’ Request
(10) Any curation function for the Related
Products that determine which exchanges,
DSPs, and Ad Networks to call
Documents concerning which products bid into
auctions, including which exchanges, DSPs, and
ad networks publishers could choose to call.
(11) Any bid, price floor, or auction
optimization algorithm, product, or feature
Documents concerning the design or effect of
bid, price floor, or auction optimizations,
including documents on Poirot, Elmo, Marple,
DA, EDA, DRS, Bernanke, Alchemist, Bell,
RPO, UPR, Smart Bidding, and pHOB.
(12) Any automated means or manner of
bidding, whether on individual bids or on a
campaign-level basis
Documents concerning bidding algorithms,
including documents on Poirot, Elmo, Marple,
Bernanke, Alchemist, Smart Bidding, and
pHOB.
Plaintiffs’ Motion should be denied as to explanatory materials because Plaintiffs’
Motion does not challenge the sufficiency of Google’s expansive proposed searches that target
the specific materials sought in Requests 39-41.
Plaintiffs’ Request Plaintiffs’ Motion should also be denied as to Request 42 for the same reasons. Like
Request 39, Request 42 seeks production of “Source code repositories and related repositories,
including training data, launch or release dates, pseudocode, and the implementation status or
roll-out associated with that source code.”
Mtn. at 4-5. But in contrast to the first nine
subcategories of Request 39, Request 42 does not target any specific feature, but is instead an
ambiguous, catch-all request that seeks production of “source code . . . related to: (1) how bids
are selected and submitted in any product that generates, alters, or submits a bid to purchase an
impression in any Relevant Product or in AdMeld; (2) the way or method by which Relevant
Products and AdMeld determine or determined the winners of any real-time auction, waterfall
bidding, and any sales of impressions.” Id. As to source code, for the same reasons as noted Page 17 PageID#
above in connection with Request 39, Plaintiffs have failed to make the requisite showing of
relevance and necessity. Supra at 13-14.
With respect to explanatory materials, notwithstanding its objections, Google’s Discovery
Proposal already includes the following extremely broad searches targeting documents
responsive to Request 42:
Subcategory of Plaintiffs’ Request
Searches Capturing Plaintiffs’ Request
(1) how bids are selected and submitted in
any product that generates, alters, or
submits a bid to purchase an impression in
any Relevant Product or in AdMeld
Documents concerning bidding algorithms, bid
selection, and bid submission by exchanges,
DSPs, or ad networks (i.e., including
documents on Awbid, Poirot, Elmo, Marple,
Bernanke, Alchemist, Smart Bidding, and
pHOB).
(2) the way or method by which Relevant
Products and AdMeld determine or
determined the winners of any real-time
auction, waterfall bidding, and any sales of
impressions.
Documents concerning AdMeld, and
documents concerning the method of sales of
impressions in Relevant Products and Admeld,
including direct sales and sales by real-time
auction, Exchange Bidding, header bidding, and
waterfall bidding.
Plaintiffs’ Motion should be denied as to non-source code explanatory materials because
Plaintiffs’ Motion does not challenge the sufficiency of Google’s expansive proposed searches
that target the specific materials sought in Request 42.
Plaintiffs’ Request Plaintiffs’ Motion fails with respect to Request 43 for the same reasons as the prior
Requests. Request 43 seeks production of “All documents created since January 1, 2005, related
to plain language explanations, code repositories, pseudocode, design documents, A/B or
comparison testing, and any other manuals or resources explaining the features, functions, and
operation of the source code used to: (1) determine the selection of bids into or sent from any Page 18 PageID#
Relevant Product or in AdMeld; and (2) determine the winners of any real-time auction,
waterfall bidding, and any sales of impressions.”
With respect to source code, Request 43 does not identify any specific feature or
algorithm, and so Plaintiffs have failed to show that production of source code is both relevant
and necessary. Supra at 13-14. As to searches for related documents, Google objected to this
Request because it “purports to require Google to produce information from Ariane and RASTA
beyond the approximately 40,000 documents from Ariane that were produced to the U.S.
Department of Justice, many of which include summaries of the results of RASTA experiments.”
Plaintiffs’ Motion does not address those comprehensive productions of responsive material.
With respect to explanatory materials, notwithstanding its Objections, Google’s
Discovery Proposal contemplates extensive custodial searches that will capture responsive
information. Specifically, Google’s Discovery Proposal already includes the following searches
targeting responsive materials:
Subcategory of Plaintiffs’ Request
Searches Capturing Plaintiffs’ Request
Documents concerning bidding algorithms, bid
(1) determine the selection of bids into or selection, and bid submission by exchanges,
sent from any Relevant Product or in DSPs, or ad networks (i.e., including documents
AdMeld;
on Awbid, Poirot, Elmo, Marple, Bernanke,
Alchemist, Smart Bidding, and pHOB).
Documents concerning the determination of
(2) determine the winners of any real-time winners of sales of impressions, including
auction, waterfall bidding, and any sales of winners of direct sales and sales by real-time
impressions.
auction, Exchange Bidding, header bidding, and
waterfall bidding.
Plaintiffs’ Motion should be denied as to non-source code explanatory materials because
Plaintiffs’ Motion does not challenge the sufficiency of Google’s expansive proposed searches
that target the specific materials sought in Request 43. Page 19 PageID#
* * *
Plaintiffs filed this Motion prematurely, after they were on notice that Google had agreed
to provide almost all of the relevant information that they were seeking, and without making any
attempt to justify their Requests for the limited information that Google did not agree to provide.
The Federal Rules of Civil Procedure and this Court’s Local Civil Rules require parties, through
counsel, to meaningfully meet and confer before filing a discovery motion in an effort to resolve
disputes without judicial intervention. See Fed. R. Civ. P. 37(a)(1). And Local Civil Rule 37(E)
requires that counsel “confer to decrease, in every way possible the filing of unnecessary
discovery motions.”
Plaintiffs’ Motion is a waste of judicial resources.
Google has complied with its
responsibility under the rules, having met and conferred with Plaintiffs in good faith on multiple
occasions to resolve any disputes regarding successor custodians and source code.
CONCLUSION
For the foregoing reasons, the Court should deny Plaintiffs’ Motion.
Dated: May 31,
Respectfully submitted,
/s/ Craig C. Reilly
CRAIG C. REILLY (VSB # 20942)
209 Madison Street
Alexandria, VA Telephone: (703) 549-Facsimile: (703) 549-Email: Craig.reilly@ccreillylaw.com
Eric Mahr (pro hac vice)
Julie S. Elmer (pro hac vice)
Andrew J. Ewalt (pro hac vice)
Lauren Kaplin (pro hac vice)
Jeanette Bayoumi (pro hac vice)
Sara Salem (pro hac vice)
FRESHFIELDS BRUCKHAUS Page 20 PageID#
DERINGER US LLP
700 13th Street NW, 10th Floor
Washington, DC Telephone: (202) 777-Facsimile: (202) 777-Email: eric.mahr@freshfields.com
Karen L. Dunn (admitted pro hac vice)
Jeannie Rhee (admitted pro hac vice)
William Isaacson (admitted pro hac vice)
Joseph Bial (admitted pro hac vice)
Byron Becker (VSB # 93384)
PAUL, WEISS, RIFKIND, WHARTON &
GARRISON LLP
2001 K Street NW
Washington, DC Telephone: (202) 223-Facsimile: (202) 223-kdunn@paulweiss.com
jrhee@paulweiss.com
wisaacson@paulweiss.com
jbial@paulweiss.com
bpbecker@paulweiss.com
Meredith Dearborn (admitted pro hac vice)
PAUL, WEISS, RIFKIND, WHARTON &
GARRISON LLP
535 Mission Street, 24th Floor
San Francisco, CA Telephone: (646) 432-Facsimile: (202) 330-mdearnborn@paulweiss.com
Daniel S. Bitton (pro hac vice)
AXINN, VELTROP & HARKRIDER LLP
55 2nd Street
San Francisco, CA Telephone: (415) 490-Facsimile: (415) 490-dbitton@axinn.com
Bradley Justus
Koren Wong-Ervin (pro hac vice)
AXINN, VELTROP & HARKRIDER LLP
1901 L Street NW
Washington, DC Page 21 PageID#
Telephone: (202) 912-Facsimile: (202) 912-bjustus@axinn.com
kwongervin@axinn.com
Counsel for Google LLC
PDF Page 1
PlainSite Cover Page
PDF Page 2
Case 1:23-cv-00108-LMB-JFA Document 238 Filed 05/31/23 Page 1 of 21 PageID# 1983
UNITED STATES DISTRICT COURT
EASTERN DISTRICT OF VIRGINIA
ALEXANDRIA DIVISION
UNITED STATES, et al.,
Plaintiffs,
vs.
No: 1:23-cv-00108-LMB-JFA
GOOGLE LLC,
Defendant.
GOOGLE LLC’S MEMORANDUM OF LAW IN OPPOSITION
TO PLAINTIFFS’ MOTION TO COMPEL PRODUCTION OF
SUCCESSOR CUSTODIAN DOCUMENTS AND SOURCE CODE
The Court should deny Plaintiffs’ Motion to Compel (the “Motion”) production of
documents that they now describe as key to their claims but which the DOJ failed to request
during its three-year-plus investigation of Google. Having failed to seek these documents when
they had the opportunity to do so, Plaintiffs now demand them within the compressed timeframe
for fact discovery. Plaintiffs should not be allowed to use this Court’s well-earned reputation for
speed and efficiency to unduly burden Google with discovery concerning information they did
not see fit to seek during their investigation.
Plaintiffs’ request that Google be compelled to add successor custodians in addition to
the 31 custodians that Google has already agreed to conditionally add should be denied.1
Plaintiffs represented to the Court that they “don’t intend to duplicate the efforts that have
1
As described below, Google’s agreement to add the 31 custodians was conditional on the
collection process resulting in a reasonable and proportionate universe of additional documents
to review.
1
PDF Page 3
Case 1:23-cv-00108-LMB-JFA Document 238 Filed 05/31/23 Page 2 of 21 PageID# 1984
already occurred during the investigation phase, and we will be as efficient as possible in seeking
third-party requests and seeking a moderate amount of refreshing of data from Google as well,”
March 24, 2023 Tr. at 6:7-11, Dkt. 70, but Plaintiffs’ requests to date require Google to review
two million additional documents beyond what was already reviewed and produced in Plaintiffs’
pre-complaint investigation.
For the 31 custodians Google has voluntarily agreed to add,
Plaintiffs were able to articulate their connection to the claims and defenses in the case. But for
the successor custodians, Plaintiffs have not articulated how these individuals are relevant or
proportional to the needs of the case. Plaintiffs’ request should therefore be denied.
As for Plaintiffs’ motion to compel source code, Google has already agreed to search for
and make available (to the extent it exists) source code for the specific features and algorithms at
issue. Specifically, Google has agreed to make available nine of the twelve categories of source
code requested.2 Google has also proposed to conduct broad searches for technical explanatory
materials that Plaintiffs seek, and Plaintiffs have not challenged the sufficiency of those
proposals. The remaining three categories of source code identified by Plaintiffs do not relate to
any specific feature or algorithm. They are ill-defined, catch-all categories of source code that
are not susceptible to a reasonable search, and Plaintiffs have not explained how they are
relevant or necessary—much less proportional to the needs of the case.
BACKGROUND
This case comes on the heels of an expansive three-year-plus investigation into several of
Google’s businesses—including digital advertising—led by the United States Department of
Justice Antitrust Division (“DOJ”). From August 2019 to October 2020, the DOJ issued five
Civil Investigative Demands (“CID”) containing 73 requests. Following its filing of a lawsuit
2
Google has agreed to make source code available pursuant to the procedures described in
Appendix C of the Protective Order, Dkt. 203.
2
PDF Page 4
Case 1:23-cv-00108-LMB-JFA Document 238 Filed 05/31/23 Page 3 of 21 PageID# 1985
regarding Google’s Search business in October 2020,3 DOJ continued its ad tech investigation
with another CID containing 37 broad requests related to Google’s advertising technology
products.
During that time, the DOJ had tremendous leverage over Google to demand
documents and information. In response to the DOJ’s CIDs, Google produced nearly 3 million
documents and the DOJ deposed at least 30 Google employees.
Plaintiffs filed their lawsuit in this case on January 24, 2023. On March 27, 2023,
Plaintiffs issued their First Set of Requests for Production (the “Requests”). Google timely
served its Objections on April 11, 2023 and its Responses on April 26, 2023. The parties have
since met and conferred on six separate occasions to discuss Plaintiffs’ Requests, including
Plaintiffs’ Requests for successor custodians and source code.
Plaintiffs’ Request for Successor Custodians
Along with its Objections, on April 11, 2023, Google provided Plaintiffs with a
comprehensive plan to meet its discovery obligations (the “Discovery Proposal”). Google’s
Discovery Proposal involved searching the files of 119 custodians over a ten year period—more
than had been negotiated and agreed to during the course of the DOJ’s expansive ad tech
investigation. As a result, Google is in the process of reviewing more than 2 million additional
documents to respond to Plaintiffs’ Requests.
On April 14, 2023, Plaintiffs responded to Google’s Discovery Proposal and requested
that Google: (1) add 15 identified custodians for specific reasons that DOJ provided, (2) identify
and add custodians with information relevant to Plaintiffs’ Federal Agency Advertiser claims, (3)
identify and add custodians from Google’s publisher partnership group, and (4) identify
successors to all of the 119 custodians proposed by Google.
See Compl., United States v. Google LLC, No. 1:20-cv-03010-APM (D.D.C. Oct 20, 2020),
ECF No. 1.
3
3
PDF Page 5
Case 1:23-cv-00108-LMB-JFA Document 238 Filed 05/31/23 Page 4 of 21 PageID# 1986
Google agreed to the vast majority of Plaintiffs’ requests and, as a result, is currently in
the process of adding more than 30 custodians in response to Plaintiffs’ specific requests.
Google made clear to Plaintiffs that the addition of each incremental custodian is a
time-consuming process that can materially increase the population of over 2 million documents
already being reviewed by Google. But despite that burden, Google: (1) conditionally agreed to
add all 15 of the custodians specifically identified by Plaintiffs,4 (2) identified and conditionally
agreed to add 14 custodians likely to have relevant information on the sale of digital advertising
to Federal Agency Advertisers, and (3) identified and conditionally agreed to add 5 custodians
from Google’s publisher partnership group. In each case, Google’s agreement to add the
identified custodians was conditional on the collection process resulting in a reasonable and
proportionate universe of additional documents to review.
Given Google’s agreement to make a substantial custodial production, that leaves only
Plaintiffs’ fourth request—that Google identify post-October-2020 successors to all 119
custodians included in Google’s Discovery Proposal—which is the subject of the instant Motion.
Google told Plaintiffs that their request to identify 119 successors is burdensome because Google
does not systematically maintain information on successors in the ordinary course of business,
and so the only way to address the request is through individualized manual inquiries. That is
because roles and responsibilities at Google are not static, reorganizations happen with some
frequency, and teams often redistribute work or direct reports as employees change teams or
depart the company.
Given this fluidity, there is no centralized system tracking successor
information.
Google intends to add one of those custodians for a limited time period because he left his role
in the pertinent business in 2014.
4
4
PDF Page 6
Case 1:23-cv-00108-LMB-JFA Document 238 Filed 05/31/23 Page 5 of 21 PageID# 1987
In light of the burdensome nature of this request and the fact that the conduct at issue
occurred over a long period of time before 2020, Google asked Plaintiffs to narrow their request
and identify a priority set of custodians who are likely to have relevant information post
October-2020. Plaintiffs responded with a list of 63 custodians, but still did not explain why
post-October 2020 documents for these 63 individuals would be relevant to Plaintiffs’ claims.
Despite Plaintiffs’ failure to address relevance and the burden involved in responding, on
May 24, Google provided successor information for six senior employees that were the focus of
Plaintiffs’ queries, and stated that it would provide further information by the end of the week.
As promised, on May 26, Google provided additional information for about 39 of the 63 Google
employees identified in Plaintiffs’ May 10 email.5 This included confirming that either no
equivalent successor existed or the successor was already captured in the existing custodian list
for the vast majority of employees. Google also confirmed that it was continuing to investigate
other possible successors.
Just two hours later, and without further meet and confer, Plaintiffs filed the instant
Motion to “compel Google to complete its search for and produce relevant materials from the
custodial files of successor employees . . . or otherwise confirm that the employees are already
included on Google’s custodian list.” Mtn. at 3-4. The Motion does not assess relevance or
narrow the relief sought in light of the successor information that Google had already provided
prior to the Motion being filed.
Plaintiffs’ Requests for Source Code
Plaintiffs also requested source code and associated explanatory materials, which they
now characterize as being “at the Heart of This Case” and “crucial to allow Plaintiffs a fair
5
An anonymized version of the successor information Google provided on May 26 is provided
in Exhibit A to protect the privacy of non-party employees. Google has already provided
Plaintiffs with the non-anonymized information in its May 26, 2023 letter to K. Garcia.
5
PDF Page 7
Case 1:23-cv-00108-LMB-JFA Document 238 Filed 05/31/23 Page 6 of 21 PageID# 1988
opportunity to prosecute their case and understand Google’s potential defenses.” Mtn. at 4-5.
But the DOJ never included similar requests for this “crucial” information during its
three-year-plus investigation.
Google timely served its Objections on April 11, 2023 and its Responses on April 26,
2023.
Notwithstanding its Objections, Google agreed to meet and confer with Plaintiffs
regarding all five Requests that are the subject of the instant Motion.
At the outset of the meet and confer process, Google told Plaintiffs that it was
considering Plaintiffs’ belated demands for source code in good faith, but explained that
Google’s source code is extremely sensitive, and that identifying code potentially responsive to
the Plaintiffs’ requests, some of which spans over 15 years, would be a burdensome process that
could take weeks to resolve. Google also explained that the process of identifying potentially
responsive lines of code (out of a much larger universe of source code) involved speaking with
multiple employees across Google.
To complicate matters further, Plaintiffs’ Requests for source code have been a moving
target.
After initially demanding both source code and pseudocode, on May 5 Plaintiffs
proposed that, in lieu of a production of source code, Google could instead produce a “complete
repository” of pseudocode, which Plaintiffs proposed to evaluate over a two-week period and
then inform Google whether (and to what extent) they would continue to press their request for
source code.
Google worked diligently to address Plaintiffs’ newly-revised Requests. On May 23
Google informed Plaintiffs that it had been unable to identify any such “complete repository” of
pseudocode requested by Plaintiffs. In the same meet and confer, Google told Plaintiffs that it
was nearing the end of its inquiries regarding production of the source code itself and anticipated
6
PDF Page 8
Case 1:23-cv-00108-LMB-JFA Document 238 Filed 05/31/23 Page 7 of 21 PageID# 1989
giving Plaintiffs a substantive response by May 30, 2023. Plaintiffs said that they would not wait
and planned to file a motion to compel.
Google asked that Plaintiffs refrain from filing any motion because the parties were not at
an impasse. Providing further context, Google explained to Plaintiffs that May 30 was the
deadline for substantial completion of production in the parallel proceedings in the Southern
District of New York and that Google was working diligently to respond in tandem to similar
requests for source code made by the MDL plaintiffs. Google specifically asked that—consistent
with the Coordination Order negotiated by the parties across the two cases (and agreed to by the
DOJ)—it be allowed to respond to these requests at the same time to make the process more
efficient. Plaintiffs refused.
On May 26, 2023, Google sent Plaintiffs a letter intended to resolve all remaining
disputes as to source code. Google explained that, subject to the provisions of the governing
Protective Order and associated Source Code Protocol, Google is prepared to search for and
make available source code sufficient to describe the following features, to the extent it exists
and can be located: Dynamic Allocation, Enhanced Dynamic Allocation, Reserve Price
Optimization, Dynamic Revenue Share, Minimum Bid to Win aka Highest Other Bid, Smart
Bidding, Poirot, and Bernanke (including Global Bernanke and Bell).
That letter concluded:
In light of Google’s extensive good faith cooperation in discovery, we think that
litigating these issues would be a waste of judicial resources. The fact that your
letter contemplates further meeting and conferring on these issues after the filing of
a motion to compel supports our view that the parties are not at an impasse and that
the disputes set forth in your letter are manufactured. Google remains willing to
meet and confer.
Plaintiffs did not meet and confer and instead filed this Motion.
7
PDF Page 9
Case 1:23-cv-00108-LMB-JFA Document 238 Filed 05/31/23 Page 8 of 21 PageID# 1990
ARGUMENT
Given the schedule, and the DOJ’s commitment, made before this Court, to focus its
discovery requests on a refresh of the productions made during its investigation, the DOJ should
not be permitted now to leverage the speed and efficiency of this Court, impose burdensome
requests, and then prematurely request judicial intervention. Google has agreed to produce
responsive information consistent with the Court’s discovery deadlines, just not on the timetable
demanded.
Plaintiffs’ request for the additional successor custodians should be denied because
Plaintiffs have not tethered the roles of post-2020 successors to the pre-2020 conduct alleged in
the Complaint. Google has already agreed to add 31 more custodians at Plaintiffs’ specific
request when Plaintiffs were able to explain the relevance of the requested custodians. Plaintiffs
have already amassed nearly 3 million of Google’s documents and Google has agreed to review
over 2 million more drawn from the files of 150 custodians. Plaintiffs’ unsubstantiated assertion
that every possible successor must be “key” and added as a custodian is insufficient to show that
the discovery they demand is “proportional to the needs of the case.” Fed. R. Civ. P. 26. What
the DOJ described to this Court as a mere “refresh” cannot amount to detailed discovery on a
scale equal to or exceeding the 3-year pre-filing review already conducted.
Plaintiffs’ request for source code should also be denied. Source code is extremely
sensitive, and Plaintiffs are required to show that its production is both relevant and necessary to
the claims and defenses in the case. Notwithstanding this standard, Google has agreed to make
available source code for the specific features and algorithms that Plaintiffs have requested.
Google has also proposed comprehensive searches to capture the related explanatory material
sought by Plaintiffs, and Plaintiffs have not challenged the sufficiency of those searches. The
8
PDF Page 10
Case 1:23-cv-00108-LMB-JFA Document 238 Filed 05/31/23 Page 9 of 21 PageID# 1991
only remaining categories of source code sought by Plaintiffs’ Motion are an ill-defined selection
of catch-alls as to which Plaintiffs made no effort to discharge their burden of showing relevance
and necessity. The Court should deny Plaintiffs’ Motion as it pertains to those sundry categories.
In filing their Motion two hours after Google sent a letter intended to put to rest all of
these issues, Plaintiffs have acted prematurely. This Court has set an “aggressive” schedule in
this case and encouraged the parties “make good use of the time that [they] have” and to “focus
[their] attention on certain things” because it will not be possible to “turn over every grain of
sand.” March 24, 2023 Tr. at 18:6-10, Dkt. 70. Plaintiffs’ Motion ignores that guidance.
I.
Plaintiffs Have Not Met Their Burden to Compel the Addition of Successor
Custodians
In its Discovery Proposal, Google identified 119 custodians that would form the basis of
its document review responding to Plaintiffs’ Requests.6 Those 119 custodians comprised over
100 that were negotiated and agreed with the DOJ during the course of its investigation, plus
additional custodians that Google identified to respond to Plaintiffs’ Requests. As noted above,
through the meet and confer process, Google has conditionally agreed to add—at Plaintiffs’
explicit request—31 further custodians for which Plaintiffs were able to articulate a concrete
nexus to the claims and defenses in the case. For any additional custodians, the burden rests with
6
It is a basic tenet of discovery that the responding party—here, Google—bears the initial
responsibility for determining which custodians should be searched for relevant documents. This
Court, like most around the United States, has endorsed the Sedona Principles. See Brink’s Co. v.
Chubb Eur. Grp. Ltd., No. 3:20-cv-520, 2021 WL 5083335, at *4 n.2 (E.D. Va. Feb. 24, 2021)
(citing the Sedona principles favorably in discussing Rule 26’s proportionality standard). Under
Sedona Principle 6, the “responding party is entitled to select the custodians most likely to
possess responsive information and to search the files of those individuals.” Mortg. Resol.
Servicing, LLC v. JPMorgan Chase Bank, N.A., No. 15 Civ. 0293 (LTS) (JCF), 2017 WL
2305398, at *2 (S.D.N.Y. May 18, 2017); see also F.D.I.C. v. Baldini, No. 12-7050, 2014 WL
1302479, at *8 (S.D.W. Va. Mar. 28, 2014) (Noting that Rule 34 “plainly contemplates that [the
responding party] must take[] steps initially to locate ESI responsive to each of [the] requests”
where the requesting party has “no familiarity or working knowledge respecting the ESI’s
responsiveness”).
9
PDF Page 11
Case 1:23-cv-00108-LMB-JFA Document 238 Filed 05/31/23 Page 10 of 21 PageID# 1992
Plaintiffs to “demonstrate that the additional requested custodians would provide unique relevant
information by providing . . . evidence that there are unique responsive documents being missed
in the current search scheme that would justify the inclusion of additional custodians.” Coventry
Capital US LLC v. EEA Life Settlements Inc., No. 17 Civ. 7417 (VM) (SLC), 2020 WL 7383940,
at *6 (S.D.N.Y. Dec. 16, 2020) (internal quotation marks and citation omitted); accord In re
Facebook, Inc. Consumer Privacy User Profile Litig., No. 3:18-MD-02843-VC-JSC, 2021 WL
10282213, at *1 (N.D. Cal. Nov. 14, 2021); In re 3M Combat Arms Earplug Prod. Liab. Litig.,
No. 3:19-md-2885, 2020 WL 4501794, at *1 (N.D. Fla. Aug. 5, 2020).
Plaintiffs have not met that burden.
Their Motion does nothing to tie the
successor-custodians to the allegations in their Complaint beyond the blanket assertion that they
must be “key” simply because they took over a role that may have been relevant to an allegation
of historical conduct. See Mtn at 2. That is insufficient under Federal Rule of Civil Procedure
26(b)(1), because “[r]elevance is . . . the foundation for any request for production.” Cook v.
Howard, 484 Fed. App’x. 805, 812 (4th Cir. 2012); see also Blankenship v. Fox News Network,
No. 19-cv-00236, 2020 WL 9718873, at *18 (S.D. W.Va. Sept. 21, 2020) (denying in part
plaintiff’s motion to compel discovery from individuals organized by job titles because plaintiff
failed to demonstrate relevance of those individuals to their claims).
II.
Google Had Already Agreed to Search For and Provide All Relevant Source Code
Prior to Plaintiffs Filing Their Motion
Plaintiffs’ Motion should be denied as moot to the extent that it seeks to compel
production of source code regarding specific features and algorithms referenced in the
Complaint. That is because Google had (prior to Plaintiffs filing their Motion) agreed to search
for and make available that source code to the extent that it exists. Plaintiffs’ Motion should be
denied to the extent that it seeks other broad and ill-defined categories of source code because
10
PDF Page 12
Case 1:23-cv-00108-LMB-JFA Document 238 Filed 05/31/23 Page 11 of 21 PageID# 1993
Plaintiffs have failed to make the required showing of necessity (in addition to relevance) for
those broad, catch-all requests.
Federal Rule of Civil Procedure 26(c)(1)(G) allows the court, for good cause, to issue an
order “requiring that a trade secret or other confidential research, development, or commercial
information not be revealed or be revealed only in a specified way.” In order for the court to
apply the rule, two criteria must exist. First, the material sought to be protected must be “a trade
secret or other confidential research, development, or commercial information.”
Id. That
condition is satisfied here because it is well settled that source code constitutes a trade secret.
See, e.g., SDSE Networks, Inc. v. Mathur, 2022 WL 18539944, at *4-5 (E.D. Va. Dec. 28, 2022)
(finding that source code qualifies as a trade secret); Trandes Corp. v. Guy F. Atkinson Co., 996
F.2d 655, 663 (4th Cir. 1993) (“[S]ource code can and does qualify as a trade secret.”). Google’s
source code is no exception.
Second, there must be a “good cause” basis for granting the request for disclosure. Fed.
R. Civ. P. 26. “[D]isclosure [of trade secrets] will be required” if “the information sought is
relevant and necessary.” Duplan Corp. v. Deering Milliken, Inc., 397 F. Supp. 1146, 1185-87
(D.S.C. 1975) (noting that “courts are loath to order disclosure of trade secrets absent a clear
showing of an immediate need”); see also Viacom Int’l Inc. v. Youtube Inc., 253 F.R.D. 256, 260
(S.D.N.Y. 2008) (discussing the required showing by plaintiffs seeking production of source
code “before disclosure of so valuable and vulnerable an asset is compelled”). If confidential or
trade secret information is not clearly relevant, it is not an abuse of discretion to deny access
altogether. See, e.g., Keyes v. Lenoir Rhyne Coll., 552 F.2d 579 (4th Cir. 1977) (affirming denial
of access and noting that the district court must use discretion to balance one party’s need for
discovery against the other party’s need for confidentiality).
11
PDF Page 13
Case 1:23-cv-00108-LMB-JFA Document 238 Filed 05/31/23 Page 12 of 21 PageID# 1994
Plaintiffs’ Motion seeks to compel production of documents, including source code,
responsive to the following Requests.
Plaintiffs’ Requests 39, 40, and 41
Requests 39 seeks “[s]ource code repositories and related repositories, including training
data, launch or release dates, pseudocode, and the implementation status or roll-out associated
with that source code,” relating to twelve purported “features and any successors or updates to
those features.” Mtn at 4.
Of those twelve “features” listed in Request 39, the first nine
implicate specific features or algorithms identified in Plaintiffs’ Complaint, namely: “(1) Project
Bernanke (including Global Bernanke and Project Bell); (2) Project Poirot (also known as
optimized fixed CPM bidding and Automated Bidding); (3) AdX’s (sellside) dynamic revenue
share; (4) AdX’s predicted highest other bidder (pHOB) algorithm; (5) Smart Bidding,
including any inputs to and outputs from the feature or program, including but not limited to
minimum bid to win; (6) Dynamic Allocation; (7) Enhanced Dynamic Allocation; (8) Project
Bell; (9) Reserve Price Optimization (RPO).” Id. (emphasis added).7
Source Code
Prior to Plaintiffs filing their Motion, Google informed Plaintiffs that it would agree to
search for and make available source code sufficient to describe the following features, to the
extent it exists and can be located: Bernanke (including Global Bernanke and Bell), Poirot,
Dynamic Revenue Share, Minimum Bid to Win aka Highest Other Bid, Smart Bidding,
Dynamic Allocation, Enhanced Dynamic Allocation, and Reserve Price Optimization.
Plaintiffs’ motion should be denied as moot as to those categories because Google has agreed to
search for and make available this source code to the extent that it exists. See Romfo v.
7
Google does not understand the difference between Plaintiffs’ categories (1) and (8) to the
extent that they both refer to “Project Bell.” As set forth below, Google has agreed to search for
and provide source code for “Bell” to the extent that it exists.
12
PDF Page 14
Case 1:23-cv-00108-LMB-JFA Document 238 Filed 05/31/23 Page 13 of 21 PageID# 1995
Scottsdale Ins. Co., No. 5:17-CV-422-D, 2019 WL 1210105, at *3 (E.D.N.C. Mar. 13, 2019)
(denying in part a motion to compel as moot where the responding party had “agreed to produce”
responsive documents “but had not done so at the time the motion was filed”).
By contrast, the remaining three categories of “features” in Plaintiffs’ Request 39 are not
features, but are ill-defined catch-alls that do not pertain to any specific feature or algorithm.
Rather, they seek “(10) any curation function for the Related Products that determine which
exchanges, DSPs, and Ad Networks to call; (11) any bid, price floor, or auction optimization
algorithm, product, or feature; and (12) any automated means or manner of bidding, whether on
individual bids or on a campaign-level basis.” Mtn. at 4. Plaintiffs have never articulated what,
specifically, they are seeking through these catch-all requests and their Motion provides no
further clues.8
As to these three remaining categories, Plaintiffs’ Motion fails because Plaintiffs have not
made the requisite showing of relevance and necessity. See, e.g., Keyes, 552 F.2d at 581 (noting
that district court must use discretion to balance one party’s need for discovery against the other
party’s need for confidentiality); see also In re Apple & AT & TM Antitrust Litig., No.
C–07–05152 JW (PVT), 2010 WL 1240295, at *3 (N.D. Cal. Mar. 26, 2010) (denying motion to
compel source code where plaintiffs had not established that the source code they sought was
“relevant and necessary” and observing that “[p]laintiffs only speculate that the additional source
code may be relevant”). What Google has already offered to produce is more than sufficient for
Plaintiffs’ purposes—Plaintiffs are “not entitled to the whole universe of source code . . . at issue
because . . . there are many functions within the relevant [programs] which have no bearing on
8
Plaintiffs assert that Google agreed to produce “certain but not all of the source code requested
by Plaintiffs.” Mtn. at 6. It is unclear what Plaintiffs believe they are missing, but to the extent
they believe something was omitted from the first nine categories of specific features or
algorithms, Google remains willing to meet and confer regarding the production of specific
identifiable source code in response to Plaintiffs’ requests.
13
PDF Page 15
Case 1:23-cv-00108-LMB-JFA Document 238 Filed 05/31/23 Page 14 of 21 PageID# 1996
this case.”
Microsoft Corp. v. Multi-Tech Sys., Inc., No. 00-1412 (ADM/RLE), 2001 WL
37130770, at *7 (D. Minn. Dec. 14, 2001).
The need to make a clear showing of relevance is not relaxed, as Plaintiffs suggest,
simply because a protective order has been entered. See Mtn. at 10. A “protective order is not a
substitute for establishing relevance or need. Its purpose here is to prevent harm by limiting
disclosure of relevant and necessary information.” Micro Motion, Inc. v. Kane Steel Co., 894
F.2d 1318, 1325 (Fed. Cir. 1990). As courts have recognized in analyzing the special protections
afforded to source code, “the protections set forth in [a] stipulated confidentiality order are
careful and extensive, but nevertheless not as safe as nondisclosure.” Viacom, 253 F.R.D. at 260.
Other Explanatory Materials
In addition to source code repositories, Request 39 also seeks “related repositories,
including training data, launch or release dates, pseudocode, and the implementation status or
roll-out associated with that source code.” Mtn. at 4. Similarly, Requests 40 and 41 seek the
following by reference to Request 39:
Request 40: All documents created since January 1, 2008, related to language
explanations, pseudocode, code repositories, design documents, A/B or comparison
testing, and any other manuals or resources explaining the features, functions, and
operation of the source code used to operate the features delineated in Request No.
39 and any updates to or successors of those features.
Request 41: Documents sufficient to show the inputs and outputs, including but not
limited to a list of any variable fields, categories of data, or Product Requirements
Documents (PRDs), to the source code and algorithms that are used in operating the
features delineated in Request No. 39 or successors or updates to those features.
Google has already agreed to conduct comprehensive searches to identify explanatory
documents responsive to Plaintiffs’ Requests 39-41. As summarized in the tables below, Google
has already proposed (and Plaintiffs have agreed to) expansive searches intended to capture
explanatory materials and other documents that relate to these Requests. In addition to the 80
14
PDF Page 16
Case 1:23-cv-00108-LMB-JFA Document 238 Filed 05/31/23 Page 15 of 21 PageID# 1997
comprehensive search terms originally proposed by Google, Plaintiffs proposed 23 additional
search strings, and Google agreed to all of them.9 The tables below summarize agreed-upon
searches that address Plaintiffs’ Requests 39-41.10
Subcategory of Plaintiffs’ Requests 39-41
Searches Capturing Plaintiffs’ Request
(1) Project Bernanke (including Global
Bernanke and Project Bell)
All documents that hit on “Bernanke” and
“Project Bell.”
(2) Project Poirot (also known as optimized
fixed CPM bidding and Automated
Bidding)
All documents that hit on “Poirot” or the
misspelling “Poriot.”
(3) AdX’s (sellside) dynamic revenue
share
All documents that hit on “DRS” as well as
documents concerning Dynamic Revenue
Share with appropriate modifiers.
(4) AdX’s predicted highest other bidder
(pHOB) algorithm
All documents that hit on “pHOB” or “HOB.”
(5) Smart Bidding, including any inputs to
and outputs from the feature or program,
including but not limited to minimum bid to
win
All documents that hit on “SmartBid” or
similar variations.
(6) Dynamic Allocation
Documents concerning dynamic allocation
with appropriate modifiers.
(7) Enhanced Dynamic Allocation
Documents concerning enhanced dynamic
allocation with appropriate modifiers.
(8) Project Bell
All documents that hit on “Project Bell.”
(9) Reserve Price Optimization (RPO)
Documents concerning RPO with appropriate
modifiers.
In its Objections to Requests 39-40, Google stated that Plaintiffs had failed to show that
production of source code was necessary “particularly given the extensive document production
made and contemplated in the MDL Expanded Investigation Materials describing the operation
of Google’s ad tech products.” In its Objections to Request 41, Google also pointed Plaintiffs to
its Discovery Proposal, which contemplates extensive custodial searches that will capture these
related materials.
10
The agreed searches are voluminous and we have summarized them here for the Court’s
convenience. Google can provide full text of the search terms if the Court would find them
helpful.
9
15
PDF Page 17
Case 1:23-cv-00108-LMB-JFA Document 238 Filed 05/31/23 Page 16 of 21 PageID# 1998
Subcategory of Plaintiffs’ Requests 39-41
Searches Capturing Plaintiffs’ Request
(10) Any curation function for the Related
Products that determine which exchanges,
DSPs, and Ad Networks to call
Documents concerning which products bid into
auctions, including which exchanges, DSPs, and
ad networks publishers could choose to call.
(11) Any bid, price floor, or auction
optimization algorithm, product, or feature
Documents concerning the design or effect of
bid, price floor, or auction optimizations,
including documents on Poirot, Elmo, Marple,
DA, EDA, DRS, Bernanke, Alchemist, Bell,
RPO, UPR, Smart Bidding, and pHOB.
(12) Any automated means or manner of
bidding, whether on individual bids or on a
campaign-level basis
Documents concerning bidding algorithms,
including documents on Poirot, Elmo, Marple,
Bernanke, Alchemist, Smart Bidding, and
pHOB.
Plaintiffs’ Motion should be denied as to explanatory materials because Plaintiffs’
Motion does not challenge the sufficiency of Google’s expansive proposed searches that target
the specific materials sought in Requests 39-41.
Plaintiffs’ Request 42
Plaintiffs’ Motion should also be denied as to Request 42 for the same reasons. Like
Request 39, Request 42 seeks production of “Source code repositories and related repositories,
including training data, launch or release dates, pseudocode, and the implementation status or
roll-out associated with that source code.”
Mtn. at 4-5. But in contrast to the first nine
subcategories of Request 39, Request 42 does not target any specific feature, but is instead an
ambiguous, catch-all request that seeks production of “source code . . . related to: (1) how bids
are selected and submitted in any product that generates, alters, or submits a bid to purchase an
impression in any Relevant Product or in AdMeld; (2) the way or method by which Relevant
Products and AdMeld determine or determined the winners of any real-time auction, waterfall
bidding, and any sales of impressions.” Id. As to source code, for the same reasons as noted
16
PDF Page 18
Case 1:23-cv-00108-LMB-JFA Document 238 Filed 05/31/23 Page 17 of 21 PageID# 1999
above in connection with Request 39, Plaintiffs have failed to make the requisite showing of
relevance and necessity. Supra at 13-14.
With respect to explanatory materials, notwithstanding its objections, Google’s Discovery
Proposal already includes the following extremely broad searches targeting documents
responsive to Request 42:
Subcategory of Plaintiffs’ Request 42
Searches Capturing Plaintiffs’ Request
(1) how bids are selected and submitted in
any product that generates, alters, or
submits a bid to purchase an impression in
any Relevant Product or in AdMeld
Documents concerning bidding algorithms, bid
selection, and bid submission by exchanges,
DSPs, or ad networks (i.e., including
documents on Awbid, Poirot, Elmo, Marple,
Bernanke, Alchemist, Smart Bidding, and
pHOB).
(2) the way or method by which Relevant
Products and AdMeld determine or
determined the winners of any real-time
auction, waterfall bidding, and any sales of
impressions.
Documents concerning AdMeld, and
documents concerning the method of sales of
impressions in Relevant Products and Admeld,
including direct sales and sales by real-time
auction, Exchange Bidding, header bidding, and
waterfall bidding.
Plaintiffs’ Motion should be denied as to non-source code explanatory materials because
Plaintiffs’ Motion does not challenge the sufficiency of Google’s expansive proposed searches
that target the specific materials sought in Request 42.
Plaintiffs’ Request 43
Plaintiffs’ Motion fails with respect to Request 43 for the same reasons as the prior
Requests. Request 43 seeks production of “All documents created since January 1, 2005, related
to plain language explanations, code repositories, pseudocode, design documents, A/B or
comparison testing, and any other manuals or resources explaining the features, functions, and
operation of the source code used to: (1) determine the selection of bids into or sent from any
17
PDF Page 19
Case 1:23-cv-00108-LMB-JFA Document 238 Filed 05/31/23 Page 18 of 21 PageID# 2000
Relevant Product or in AdMeld; and (2) determine the winners of any real-time auction,
waterfall bidding, and any sales of impressions.”
With respect to source code, Request 43 does not identify any specific feature or
algorithm, and so Plaintiffs have failed to show that production of source code is both relevant
and necessary. Supra at 13-14. As to searches for related documents, Google objected to this
Request because it “purports to require Google to produce information from Ariane and RASTA
beyond the approximately 40,000 documents from Ariane that were produced to the U.S.
Department of Justice, many of which include summaries of the results of RASTA experiments.”
Plaintiffs’ Motion does not address those comprehensive productions of responsive material.
With respect to explanatory materials, notwithstanding its Objections, Google’s
Discovery Proposal contemplates extensive custodial searches that will capture responsive
information. Specifically, Google’s Discovery Proposal already includes the following searches
targeting responsive materials:
Subcategory of Plaintiffs’ Request 43
Searches Capturing Plaintiffs’ Request
Documents concerning bidding algorithms, bid
(1) determine the selection of bids into or selection, and bid submission by exchanges,
sent from any Relevant Product or in DSPs, or ad networks (i.e., including documents
AdMeld;
on Awbid, Poirot, Elmo, Marple, Bernanke,
Alchemist, Smart Bidding, and pHOB).
Documents concerning the determination of
(2) determine the winners of any real-time winners of sales of impressions, including
auction, waterfall bidding, and any sales of winners of direct sales and sales by real-time
impressions.
auction, Exchange Bidding, header bidding, and
waterfall bidding.
Plaintiffs’ Motion should be denied as to non-source code explanatory materials because
Plaintiffs’ Motion does not challenge the sufficiency of Google’s expansive proposed searches
that target the specific materials sought in Request 43.
18
PDF Page 20
Case 1:23-cv-00108-LMB-JFA Document 238 Filed 05/31/23 Page 19 of 21 PageID# 2001
* * *
Plaintiffs filed this Motion prematurely, after they were on notice that Google had agreed
to provide almost all of the relevant information that they were seeking, and without making any
attempt to justify their Requests for the limited information that Google did not agree to provide.
The Federal Rules of Civil Procedure and this Court’s Local Civil Rules require parties, through
counsel, to meaningfully meet and confer before filing a discovery motion in an effort to resolve
disputes without judicial intervention. See Fed. R. Civ. P. 37(a)(1). And Local Civil Rule 37(E)
requires that counsel “confer to decrease, in every way possible the filing of unnecessary
discovery motions.”
Plaintiffs’ Motion is a waste of judicial resources.
Google has complied with its
responsibility under the rules, having met and conferred with Plaintiffs in good faith on multiple
occasions to resolve any disputes regarding successor custodians and source code.
CONCLUSION
For the foregoing reasons, the Court should deny Plaintiffs’ Motion.
Dated: May 31, 2023
Respectfully submitted,
/s/ Craig C. Reilly
CRAIG C. REILLY (VSB # 20942)
209 Madison Street
Alexandria, VA 22314
Telephone: (703) 549-5354
Facsimile: (703) 549-5355
Email: Craig.reilly@ccreillylaw.com
Eric Mahr (pro hac vice)
Julie S. Elmer (pro hac vice)
Andrew J. Ewalt (pro hac vice)
Lauren Kaplin (pro hac vice)
Jeanette Bayoumi (pro hac vice)
Sara Salem (pro hac vice)
FRESHFIELDS BRUCKHAUS
19
PDF Page 21
Case 1:23-cv-00108-LMB-JFA Document 238 Filed 05/31/23 Page 20 of 21 PageID# 2002
DERINGER US LLP
700 13th Street NW, 10th Floor
Washington, DC 20005
Telephone: (202) 777-4500
Facsimile: (202) 777-4555
Email: eric.mahr@freshfields.com
Karen L. Dunn (admitted pro hac vice)
Jeannie Rhee (admitted pro hac vice)
William Isaacson (admitted pro hac vice)
Joseph Bial (admitted pro hac vice)
Byron Becker (VSB # 93384)
PAUL, WEISS, RIFKIND, WHARTON &
GARRISON LLP
2001 K Street NW
Washington, DC 20006
Telephone: (202) 223-7300
Facsimile: (202) 223-7420
kdunn@paulweiss.com
jrhee@paulweiss.com
wisaacson@paulweiss.com
jbial@paulweiss.com
bpbecker@paulweiss.com
Meredith Dearborn (admitted pro hac vice)
PAUL, WEISS, RIFKIND, WHARTON &
GARRISON LLP
535 Mission Street, 24th Floor
San Francisco, CA 94105
Telephone: (646) 432-5100
Facsimile: (202) 330-5908
mdearnborn@paulweiss.com
Daniel S. Bitton (pro hac vice)
AXINN, VELTROP & HARKRIDER LLP
55 2nd Street
San Francisco, CA 94105
Telephone: (415) 490-2000
Facsimile: (415) 490-2001
dbitton@axinn.com
Bradley Justus
Koren Wong-Ervin (pro hac vice)
AXINN, VELTROP & HARKRIDER LLP
1901 L Street NW
Washington, DC 20036
20
PDF Page 22
Case 1:23-cv-00108-LMB-JFA Document 238 Filed 05/31/23 Page 21 of 21 PageID# 2003
Telephone: (202) 912-4700
Facsimile: (202) 912-4701
bjustus@axinn.com
kwongervin@axinn.com
Counsel for Google LLC
21
Back to 
