`
` Pages 1 - 46
`
`UNITED STATES DISTRICT COURT
`
`NORTHERN DISTRICT OF CALIFORNIA
`
`BEFORE THE HONORABLE THOMAS S. HIXSON, MAGISTRATE JUDGE
`
`
`IN RE APPLE IPHONE ANTITRUST )
`LITIGATION. ) No. 11-cv-06714-YGR (TSH)
` )
`DONALD R. CAMERON, et al., )
` )
` Plaintiffs, )
` )
` VS. ) No. 19-cv-03074-YGR (TSH)
` )
`APPLE INC.,
`)
` )
` Defendant.
`)
` )
`EPIC GAMES, INC.,
`)
` )
` Plaintiff/ )
` Counter-defendant, )
` )
` VS. ) No. 20-cv-05640-YGR (TSH)
` )
`APPLE INC.,
`)
` )
` Defendant/ )
` Counterclaimant. )
` )
` San Francisco, California
` Wednesday, December 9, 2020
`
`
`
`TRANSCRIPT OF PROCEEDINGS VIA ZOOM WEBINAR
`
`
`
`(Appearances on next page)
`
`
`
`Reported by: Katherine Powell Sullivan, CSR #5812, CRR, RMR
` Official Reporter - U.S. District Court
`
`
`
`Case 4:19-cv-03074-YGR Document 183 Filed 12/10/20 Page 2 of 46
`
` 2
`
`APPEARANCES: (via Zoom Webinar)
`
`For Plaintiffs in In re Apple iPhone Antitrust Litigation,
`11-cv-06714-YGR:
` WOLF HALDENSTEIN ADLER FREEMAN & HERZ LLP
` 750 B Street, Suite 1820
` San Diego, California 92101
` BY: RACHELE R. BYRD, ESQ.
`
`For Plaintiffs in Cameron, et. al v. Apple Inc.,
`19-cv-03074-YGR:
` HAGENS BERMAN SOBOL SHAPIRO LLP
` 1301 Second Avenue, Suite 2000
` Seattle, Washington 98101
` BY: ROBERT F. LOPEZ, ESQ.
`
`For Plaintiff Epic Games, Inc.:
` CRAVATH, SWAINE & MOORE LLP
` 825 Eighth Avenue
` New York, New York 10019
` BY: LAUREN A. MOSKOWITZ, ESQ.
`
`For Defendant Apple Inc.:
` GIBSON, DUNN & CRUTCHER LLP
` 555 Mission Street
` San Francisco, California 94105-0921
` BY: ETHAN D. DETTMER, ESQ.
`
` 1
`
` 2
`
` 3
`
` 4
`
` 5
`
` 6
`
` 7
`
` 8
`
` 9
`
`10
`
`11
`
`12
`
`13
`
`14
`
`15
`
`16
`
`17
`
`18
`
`19
`
`20
`
`21
`
`22
`
`23
`
`24
`
`25
`
`
`
`Case 4:19-cv-03074-YGR Document 183 Filed 12/10/20 Page 3 of 46
`
` 3
`
` 1
`
` 2
`
` 3
`
` 4
`
` 5
`
` 6
`
` 7
`
` 8
`
` 9
`
`10
`
`11
`
`12
`
`13
`
`14
`
`15
`
`16
`
`17
`
`18
`
`19
`
`20
`
`21
`
`22
`
`23
`
`24
`
`25
`
`Wednesday - December 9, 2020
`
` 1:01 p.m.
`
`P R O C E E D I N G S
`
`---000---
`
`THE CLERK: All right, everyone. Good afternoon.
`
`Thank you for all joining us on the Zoom call.
`
`We're here in Civil Action 11-6714, In re Apple iPhone
`
`Antitrust Litigation, and in Case Number 19-3074, Cameron, et
`
`al. versus Apple Inc. And the last case is 20-5640, Epic
`
`Games, Inc. versus Apple Inc.
`
`Counsel, please state your appearances for the record.
`
`The Honorable Thomas S. Hixson, presiding.
`
`Let's start with the Apple Antitrust Litigation first.
`
`MS. BYRD: Good afternoon, Your Honor. This is
`
`Rachele Byrd, with Wolf Haldenstein Adler Freeman & Herz, on
`
`behalf of the plaintiffs.
`
`THE COURT: Good afternoon.
`
`THE CLERK: Thank you.
`
`And the Cameron v. Apple.
`
`MR. LOPEZ: Good afternoon, Your Honor. This is Rob
`
`Lopez, of Hagens Berman, for the developer plaintiffs in the
`
`Cameron matter.
`
`THE COURT: Good afternoon.
`
`THE CLERK: And Epic Games, Inc., versus Apple Inc.
`
`MS. MOSKOWITZ: Good afternoon, Your Honor. Lauren
`
`Moskowitz, from Cravath, Swaine & Moore, on behalf of Epic
`
`
`
`Case 4:19-cv-03074-YGR Document 183 Filed 12/10/20 Page 4 of 46
`
` 4
`
` 1
`
` 2
`
` 3
`
` 4
`
` 5
`
` 6
`
` 7
`
` 8
`
` 9
`
`10
`
`11
`
`12
`
`13
`
`14
`
`15
`
`16
`
`17
`
`18
`
`19
`
`20
`
`21
`
`22
`
`23
`
`24
`
`25
`
`Games.
`
`THE COURT: Good afternoon.
`
`MR. DETTMER: And, Your Honor, Ethan Dettmer, from
`
`Gibson Dunn & Crutcher, on behalf of Apple in all three
`
`matters.
`
`THE COURT: Good afternoon.
`
`MR. DETTMER: Good afternoon, Your Honor.
`
`THE COURT: So let me tell you what my agenda is for
`
`the hearing. I have some questions that I want to ask Epic
`
`Games and Apple, and then I have some thoughts that I want to
`
`discuss with those two parties.
`
`And then what I'm likely to do is tell you to go meet and
`
`confer a little bit more in light of my -- the ideas I discuss
`
`with the two parties.
`
`I know that you've met and conferred for several weeks
`
`already in a good-faith attempt to get to an answer, and I'm
`
`hoping that with some feedback from me we can get you over the
`
`finish line.
`
`So, anyway, there's not going to be an order coming out of
`
`this hearing. Or the order will be that you talk with each a
`
`little bit more.
`
`First, I want to ask Epic, Inc., just to make sure that I
`
`understand how you searched for documents, I understand that
`
`you had a set of search terms and that was shared with Apple.
`
`Is that correct?
`
`
`
`Case 4:19-cv-03074-YGR Document 183 Filed 12/10/20 Page 5 of 46
`
` 5
`
` 1
`
` 2
`
` 3
`
` 4
`
` 5
`
` 6
`
` 7
`
` 8
`
` 9
`
`10
`
`11
`
`12
`
`13
`
`14
`
`15
`
`16
`
`17
`
`18
`
`19
`
`20
`
`21
`
`22
`
`23
`
`24
`
`25
`
`MS. MOSKOWITZ: Yes, Your Honor. We had initially
`
`proposed a set of terms, and Apple responded with a lengthy set
`
`of proposed conditions that we met and conferred and reached
`
`agreement on.
`
`THE COURT: Okay. So there was -- at the start of the
`
`process, there was an agreement on a group of search terms so
`
`Apple would know if a document didn't have one of their search
`
`terms it wasn't going to come up. And if it did, then the
`
`search term would pull it out. Is that right?
`
`MS. MOSKOWITZ: Correct. Correct.
`
`THE COURT: Okay. Then after you ran the search
`
`terms, did you have a team of document reviewers do manual
`
`review to see which documents were, in fact, responsive?
`
`MS. MOSKOWITZ: That's correct, Your Honor. We are
`
`performing a linear review of all of the search term hit
`
`results, which is in the order of over 3.5 million documents.
`
`THE COURT: Wow. That sounds like a big task.
`
`MS. MOSKOWITZ: It is.
`
`THE COURT: Was there ever a time when you took a
`
`sample of the documents that hit on the search terms and then
`
`disclosed to Apple, here's how we're making responsiveness
`
`calls with respect to this set, and let us know what you --
`
`what you think?
`
`In other words, I knew that they had input on your search
`
`terms. What I'm wondering is whether they also had any input
`
`
`
`Case 4:19-cv-03074-YGR Document 183 Filed 12/10/20 Page 6 of 46
`
` 6
`
` 1
`
` 2
`
` 3
`
` 4
`
` 5
`
` 6
`
` 7
`
` 8
`
` 9
`
`10
`
`11
`
`12
`
`13
`
`14
`
`15
`
`16
`
`17
`
`18
`
`19
`
`20
`
`21
`
`22
`
`23
`
`24
`
`25
`
`into your responsiveness calls or if -- and I know this a term
`
`that no one -- it sounds pejorative, but whether what the
`
`manual reviewers were doing is not just a black box to Apple?
`
`MS. MOSKOWITZ: Your Honor, we have not disclosed a
`
`set of specific documents with the up or down determinations
`
`that we are making. We haven't discussed that with Apple.
`
`We are -- we can describe that we are taking quite a broad
`
`responsiveness on view, and we have disclosed that concept to
`
`Apple, that we are not making very fine responsiveness calls,
`
`but rather we are erring on the side of responsiveness.
`
`THE COURT: I see. So, presumably, there's some
`
`attorney-client privilege review memo, and your team are
`
`working off of that. Is that right?
`
`MS. MOSKOWITZ: A responsiveness protocol that
`
`includes both privilege calls and the types of documents that
`
`really shouldn't be marked "responsive," along the lines of
`
`technical documents, coding documents, that really are not
`
`relevant to the issues in the case. But separate from that,
`
`the instructions have been to mark "responsive."
`
`THE COURT: Okay. And then in your validation
`
`protocol, I think you had two aspects to it. One is you would
`
`validate that the search terms did, in fact, pull up a certain
`
`recall level. Is that the idea?
`
`MS. MOSKOWITZ: For step one, Your Honor, that's
`
`correct. What we had proposed to do upfront, so that we could
`
`
`
`Case 4:19-cv-03074-YGR Document 183 Filed 12/10/20 Page 7 of 46
`
` 7
`
` 1
`
` 2
`
` 3
`
` 4
`
` 5
`
` 6
`
` 7
`
` 8
`
` 9
`
`10
`
`11
`
`12
`
`13
`
`14
`
`15
`
`16
`
`17
`
`18
`
`19
`
`20
`
`21
`
`22
`
`23
`
`24
`
`25
`
`know with certainty the universe that we were going to be
`
`performing the linear review on, would be to take a sample of
`
`the overall population.
`
`And that's, I think, on the order of 6-and-a-half million
`
`documents after de-duplication, and to take a sample of 1500
`
`documents from that full universe, some of which would be from
`
`the hit results and some from the non-hit results, review those
`
`by an expert for responsiveness, and calculate a recall
`
`percentage based on the responsive hit in that sample
`
`population.
`
`And we had proposed to Apple that we would reach an
`
`80 percent recall on that such that 80 percent of the
`
`responsive documents were being returned by the search terms
`
`and that we would be prepared to move forward on that basis.
`
`THE COURT: Okay. So that's the first aspect.
`
`Then there was another aspect, and that was the -- to take
`
`a random sample of the ones coded "not responsive" to determine
`
`which ones should have been responsive, and you would calculate
`
`the elusion rate.
`
`And for the benefit of the court reporter, I'm saying
`
`e-l-u-s-i-o-n. I'm not saying i-l-l-u-s-i-o-n.
`
`So is that the other aspect of it?
`
`MS. MOSKOWITZ: That's correct, Your Honor. That is
`
`what our proposal was, as a compromise, given that we -- we
`
`personally have never seen anyone go that far on a search terms
`
`
`
`Case 4:19-cv-03074-YGR Document 183 Filed 12/10/20 Page 8 of 46
`
` 8
`
` 1
`
` 2
`
` 3
`
` 4
`
` 5
`
` 6
`
` 7
`
` 8
`
` 9
`
`10
`
`11
`
`12
`
`13
`
`14
`
`15
`
`16
`
`17
`
`18
`
`19
`
`20
`
`21
`
`22
`
`23
`
`24
`
`25
`
`protocol.
`
`I'm happy to give more background on how we ended up
`
`there, but that is the end result of what we were proposing to
`
`do.
`
`THE COURT: I see.
`
`So I had two concerns that maybe you can speak to. The
`
`one is that the recall rate that you're estimating is not a
`
`true recall rate because your doc review process had two steps.
`
`First you had search terms, and then we had the manual
`
`review. And so this 80 percent, it doesn't capture the manual
`
`review part, and that's kind of what I'm struggling with.
`
`I understand that your elusion rate is designed to deal
`
`with the manual review problem, but the concern there -- I
`
`think Apple's right, that if the search terms got in a lot of
`
`nonresponsive documents, then you could pass the elusion rate
`
`test even if actual recall wasn't that great.
`
`So I'm wondering if you could speak to those concerns.
`
`MS. MOSKOWITZ: Yes, Your Honor. I think, just from
`
`a -- backing up to first principles on what you're trying to
`
`achieve in discovery, I think that we have a disagreement on
`
`what that is.
`
`And we don't view the exercise of discovery, especially in
`
`cases with just millions and millions and millions and millions
`
`of documents, that you are trying to find every responsive
`
`document.
`
`
`
`Case 4:19-cv-03074-YGR Document 183 Filed 12/10/20 Page 9 of 46
`
` 9
`
` 1
`
` 2
`
` 3
`
` 4
`
` 5
`
` 6
`
` 7
`
` 8
`
` 9
`
`10
`
`11
`
`12
`
`13
`
`14
`
`15
`
`16
`
`17
`
`18
`
`19
`
`20
`
`21
`
`22
`
`23
`
`24
`
`25
`
`We've all served bilateral requests that, if read
`
`literally, call for every document in our files. And the real
`
`idea is to get not just responsive documents but what's the
`
`meat of the documents in this case.
`
`An exhibit list in this trial is going to be a fraction of
`
`a fraction of the documents produced. And so search terms are
`
`not just targeted at getting responsive documents but getting
`
`what the critical documents are.
`
`Apple got to say, we think that the documents that you
`
`have, that we really care about, are going to say the following
`
`types of words, and go look for those documents, and then
`
`review those for responsiveness and produce them.
`
`That is the whole idea behind search terms. It's not a
`
`blind responsiveness review but, rather, a targeted meet and
`
`confer on targeted terms that Apple specifically can request us
`
`to look at to get the documents that they think they need in
`
`the case.
`
`And so the -- the 80 percent offer that we did was
`
`really -- it does look at responsiveness, but it's really just
`
`a check on is there a huge number of responsive documents that
`
`aren't hitting on the terms?
`
`And that gives you the comfort that that's not happening,
`
`that only 20 percent of technically responsive documents are
`
`being left behind.
`
`But Apple has already told us the specific documents they
`
`
`
`Case 4:19-cv-03074-YGR Document 183 Filed 12/10/20 Page 10 of 46
`
` 10
`
` 1
`
` 2
`
` 3
`
` 4
`
` 5
`
` 6
`
` 7
`
` 8
`
` 9
`
`10
`
`11
`
`12
`
`13
`
`14
`
`15
`
`16
`
`17
`
`18
`
`19
`
`20
`
`21
`
`22
`
`23
`
`24
`
`25
`
`want by virtue of the search terms. So it's really the search
`
`terms themselves are the validation of our protocol. And the
`
`80 percent is just a double check or a triple check because
`
`they really do have so much involvement and had so much
`
`involvement in search terms themselves.
`
`So that's why we proposed that on the search terms piece.
`
`And I can certainly speak to the second aspect of it, but I may
`
`have lost your question if I haven't yet answered it.
`
`THE COURT: I think you have. I think the concept
`
`here, that you're getting at, is sometimes referred to as
`
`precision. The only way you can get every relevant document is
`
`you produce all the documents. But no one wants all the
`
`documents, they want the good stuff. And so okay.
`
`All right. Let me -- let me ask Apple some questions.
`
`If I understand the way you did TAR, at the beginning you
`
`would have to -- I mean, the machine learns. So you start by
`
`making some responsiveness calls so the machine can learn from
`
`that and it can find documents that are similar to the ones
`
`that are responsive and not so much of the nonresponsive ones.
`
`Let me ask you a version of the same question I asked Epic
`
`Games, which is: Was there a time early on when you had a
`
`sample set of documents, you made responsiveness calls, and you
`
`shared those with Epic Games to get their input?
`
`MR. DETTMER: No, Your Honor.
`
`First of all, we obviously started this review long before
`
`
`
`Case 4:19-cv-03074-YGR Document 183 Filed 12/10/20 Page 11 of 46
`
` 11
`
` 1
`
` 2
`
` 3
`
` 4
`
` 5
`
` 6
`
` 7
`
` 8
`
` 9
`
`10
`
`11
`
`12
`
`13
`
`14
`
`15
`
`16
`
`17
`
`18
`
`19
`
`20
`
`21
`
`22
`
`23
`
`24
`
`25
`
`Epic filed its lawsuit. You know, so there's two reviews,
`
`right? There's the App Store antitrust class action reviews
`
`that have been going on, you know, for better part of a year
`
`now, and then the new Epic ones, which are separate but
`
`related.
`
`But in direct answer to your question, the way this
`
`particular TAR system works is it does not operate, you know,
`
`with any kind of seed set or anything like that. It's a
`
`continuous active learning-type system where the review starts
`
`and each and every call that a reviewer makes then looks at
`
`each document, looks at the -- you know, the characteristics of
`
`that document, and looks for other documents in the whole
`
`collection that are either like or unlike it, and then re-sorts
`
`the stack of documents to put the more likely to be a
`
`responsive documents toward the top.
`
`But that system keeps going throughout the whole review,
`
`and you keep doing it until you get to a point where, you know,
`
`the richness of the documents you're reviewing gets to a very
`
`low point. And then, you know, you do some validation and you
`
`stop because there aren't enough documents left.
`
`So this system is great in that it makes the review more
`
`efficient by pushing the more likely to be responsive documents
`
`toward the top, but also has the advantage of allowing you to
`
`review, you know, each and every document as you go along until
`
`you get to a point where, you know, there just aren't many
`
`
`
`Case 4:19-cv-03074-YGR Document 183 Filed 12/10/20 Page 12 of 46
`
` 12
`
` 1
`
` 2
`
` 3
`
` 4
`
` 5
`
` 6
`
` 7
`
` 8
`
` 9
`
`10
`
`11
`
`12
`
`13
`
`14
`
`15
`
`16
`
`17
`
`18
`
`19
`
`20
`
`21
`
`22
`
`23
`
`24
`
`25
`
`responsive documents left in what you're reviewing.
`
`Now, I will just -- and sorry to drag this out a little
`
`bit. I'll give you a caveat in that, as Your Honor probably
`
`knows, we have a much more aggressive timeline in place right
`
`now than we had before the Epic case started.
`
`So we have made a slight change to that now where we're
`
`actually going to use kind of a hybrid system of what I just
`
`described and a more traditional TAR 1 system where, you know,
`
`the seed set, so to speak, is what we've done already, which is
`
`obviously millions of documents reviewed, and hundreds of
`
`thousands in the Epic case, which then has taught the system.
`
`So it's complicated, but that's the -- that's the high
`
`level. And I will say we have described this in detail to the
`
`plaintiffs, including, you know, allowing them to talk with our
`
`consultant about it.
`
`THE COURT: Okay. Thank you. That's helpful.
`
`And then, as I understand your validation proposal,
`
`there's a symmetry with Epic Games in the sense that what you
`
`want to validate is 75 percent recall of what you think is
`
`responsive, but they don't really know what that is.
`
`Like, just as the -- they haven't -- you don't have any
`
`input into what their manual reviewers are saying, responsive
`
`or not responsive. The calls that Apple has been making, you
`
`know, you've been making, but they don't get to see that and
`
`they wouldn't as part of the validation. Is that right?
`
`
`
`Case 4:19-cv-03074-YGR Document 183 Filed 12/10/20 Page 13 of 46
`
` 13
`
` 1
`
` 2
`
` 3
`
` 4
`
` 5
`
` 6
`
` 7
`
` 8
`
` 9
`
`10
`
`11
`
`12
`
`13
`
`14
`
`15
`
`16
`
`17
`
`18
`
`19
`
`20
`
`21
`
`22
`
`23
`
`24
`
`25
`
`MR. DETTMER: Well, yes and no, Your Honor. I mean,
`
`we did produce to Epic, you know, as soon as -- well, shortly
`
`after they started the case, we produced to them the
`
`approximately 3.7 million documents that we produced in the
`
`class actions to date.
`
`So they have that production and, obviously, know that
`
`those documents were coded as responsive. So to that extent,
`
`they do know what we are coding is responsive. And as Your
`
`Honor knows, they take issue with at least part of that. And,
`
`obviously, I'm happy to talk about that further.
`
`So I suppose, in a sense, they do know at least --
`
`THE COURT: Well, I mean, you both know that, what the
`
`other side has produced. I understand that. But how the
`
`sausage got made, you don't know. You know the output of it
`
`and they know the output of it.
`
`There are some times with TAR where people do exchange
`
`seed sets, and they do, as part of validation, exchange all
`
`that. I took from the letter brief that that didn't happen
`
`here. I'm just asking the question to confirm that that
`
`understanding is correct.
`
`MR. DETTMER: That's correct, Your Honor. And, as I
`
`described, we don't really have a seed set in the first case.
`
`THE COURT: You started it before there was an Epic,
`
`Inc. lawsuit, so I guess they -- they really couldn't have had
`
`much input.
`
`
`
`Case 4:19-cv-03074-YGR Document 183 Filed 12/10/20 Page 14 of 46
`
` 14
`
` 1
`
` 2
`
` 3
`
` 4
`
` 5
`
` 6
`
` 7
`
` 8
`
` 9
`
`10
`
`11
`
`12
`
`13
`
`14
`
`15
`
`16
`
`17
`
`18
`
`19
`
`20
`
`21
`
`22
`
`23
`
`24
`
`25
`
`MR. DETTMER: Right.
`
`THE COURT: Okay. So here's a thought that I have.
`
`At a high level, I -- I don't think search terms are any better
`
`than TAR. They both have pluses and minuses. They have, you
`
`know, pros and cons depending on the kind of case, but there's
`
`no reason why one of them is inherently better than the other.
`
`So I want to set a recall burden that's roughly similar for the
`
`two sides because I do think that's appropriate.
`
`Apple's proposal of -- I think it was 75 percent? Is that
`
`what you were looking at?
`
`MR. DETTMER: Yes, Your Honor.
`
`THE COURT: That seems right to me.
`
`There is this tension between recall and precision that if
`
`you go up too high then you start producing a lot of junk,
`
`which, you know, no one wants.
`
`But here's a thought that I have, is that maybe we need to
`
`measure recall differently for the two sides. And let me ask
`
`Apple first.
`
`So with the search terms it seems a little unfair to
`
`measure recall against the population of documents that were
`
`collected, because you know and Epic Games knows that
`
`everything that didn't have a search term is just never going
`
`to get reviewed.
`
`And so what I'm thinking is that maybe the real issue is
`
`making sure that their manual review was consistent and found
`
`
`
`Case 4:19-cv-03074-YGR Document 183 Filed 12/10/20 Page 15 of 46
`
` 15
`
` 1
`
` 2
`
` 3
`
` 4
`
` 5
`
` 6
`
` 7
`
` 8
`
` 9
`
`10
`
`11
`
`12
`
`13
`
`14
`
`15
`
`16
`
`17
`
`18
`
`19
`
`20
`
`21
`
`22
`
`23
`
`24
`
`25
`
`the stuff that it's supposed to find.
`
`And so I'm thinking that maybe we -- for Epic Games, we
`
`measure recall by the manual review as compared to the
`
`documents that hit on the search terms, and we want 75 percent
`
`of recall in that universe.
`
`Another way of doing it is to compare what the manual
`
`review says is responsive against not just what the search
`
`terms found but what they were run against and didn't pull up.
`
`But I'm wondering if that might be too big a burden on Epic
`
`Games.
`
`What are your thoughts?
`
`MS. MOSKOWITZ: Your Honor, I do --
`
`THE COURT: First I was asking Apple --
`
`MS. MOSKOWITZ: Oh. Pardon me.
`
`THE COURT: -- because I'm proposing something more
`
`favorable to you, and I want to hear their strenuous objections
`
`to it before --
`
`MR. DETTMER: Well, I guess -- I guess a concern, Your
`
`Honor, is twofold. One, you know, the -- the idea of search
`
`terms, as Your Honor pointed out, there are pluses and minuses.
`
`One of the minuses, particularly in a case like this one,
`
`where, you know, we obviously did have some input into what the
`
`search terms were going to be, you know, at the outset we
`
`obviously didn't have any idea what their documents were at the
`
`time.
`
`
`
`Case 4:19-cv-03074-YGR Document 183 Filed 12/10/20 Page 16 of 46
`
` 16
`
` 1
`
` 2
`
` 3
`
` 4
`
` 5
`
` 6
`
` 7
`
` 8
`
` 9
`
`10
`
`11
`
`12
`
`13
`
`14
`
`15
`
`16
`
`17
`
`18
`
`19
`
`20
`
`21
`
`22
`
`23
`
`24
`
`25
`
`So, you know, while we can make a, you know, not
`
`particularly well-educated guess about what the good search
`
`terms might be, there's a lot of blind spots that we had in
`
`giving our input into those search terms.
`
`So I think, you know, I would have a lot of concern with
`
`that, that there could be a lot of responsive documents that
`
`were left out of the search term call. I mean, in, you know, a
`
`lot of these cases, as I'm sure Your Honor knows, there's an
`
`iterative search process where you have search terms, you get
`
`documents back, you look at the documents, you do another round
`
`of that where you have sort of more knowledge about what the
`
`documents are so that you can then, you know, say something
`
`more intelligent about -- about the search terms.
`
`And, frankly, we just don't have time for that here given
`
`the really aggressive schedule in the case.
`
`THE COURT: Let me sort of short-circuit this a little
`
`bit by saying, I understand and agree with all the things
`
`you're saying. It's always frustrating when one side says,
`
`"What search terms do you want us to run?" You know, I don't
`
`know, I have your documents. That's a logical thing to say.
`
`I guess I'm more asking you about here we are getting
`
`toward the end, and you knew that they were using search terms.
`
`You had some limited ability to thoughtfully comment on them,
`
`and you did what you could. But the known risk is that if a
`
`document didn't hit on a search term, they weren't going to
`
`
`
`Case 4:19-cv-03074-YGR Document 183 Filed 12/10/20 Page 17 of 46
`
` 17
`
` 1
`
` 2
`
` 3
`
` 4
`
` 5
`
` 6
`
` 7
`
` 8
`
` 9
`
`10
`
`11
`
`12
`
`13
`
`14
`
`15
`
`16
`
`17
`
`18
`
`19
`
`20
`
`21
`
`22
`
`23
`
`24
`
`25
`
`find it. So that was -- you knew that at the beginning. And
`
`so I'm wondering if it's fair to hold you to that.
`
`The thing you had no ability to evaluate was their manual
`
`review. That really was a black box. And I'm wondering if --
`
`should we at least hold Epic Games to the burden to prove that
`
`they found the things they were trying to find?
`
`That's really all your validation proposal is showing, as
`
`well, that you found the things that you think were responsive.
`
`And so I'm wondering if just because you walked into this
`
`known risk -- like, at the end of the document production we
`
`can't have Epic Games redo all their search terms. If they
`
`weren't any good, that problem existed from the beginning. But
`
`we just don't have time to do that now.
`
`And so that's why I'm wondering if recall should be
`
`measured against what hit on the search terms. Can you respond
`
`more to that, rather than addressing why search terms aren't
`
`great, which I know already.
`
`MR. DETTMER: Well, Your Honor, I didn't mean to
`
`sidetrack. I guess the reason I raise that is because that
`
`creates, at the end of the day, a validation protocol that
`
`really treats the two parties very differently.
`
`And I understand we're using different methodologies, but
`
`at the end of the day our belief -- Apple's belief is that,
`
`regardless of whether we're using search terms or TAR, each
`
`party should -- and, frankly, the lawyers should be held to the
`
`
`
`Case 4:19-cv-03074-YGR Document 183 Filed 12/10/20 Page 18 of 46
`
` 18
`
` 1
`
` 2
`
` 3
`
` 4
`
` 5
`
` 6
`
` 7
`
` 8
`
` 9
`
`10
`
`11
`
`12
`
`13
`
`14
`
`15
`
`16
`
`17
`
`18
`
`19
`
`20
`
`21
`
`22
`
`23
`
`24
`
`25
`
`same ultimate recall standard.
`
`And I think if Apple is being held to a recall standard
`
`against all the documents that are sort of in the search
`
`universe and Epic is being held to one that's, you know, a much
`
`narrower universe of documents that have already been searched
`
`and winnowed, that's really not a -- you know, a real
`
`comparison. And Epic is being held to a much lower standard
`
`under that different protocol than we are.
`
`THE COURT: Well, that's true in one sense. They're
`
`certainly being held to a different standard. But I guess at
`
`the beginning Apple said, we're running TAR against all of the
`
`documents that we collect.
`
`And Epic Games has said, we're not going to do a manual
`
`review of all the documents, we're only going to do it with the
`
`ones that come up with search terms.
`
`So the way that these are symmetrical is that we're trying
`
`to decide how good were you at finding the things you wanted to
`
`find in the places you were looking? And Apple and Epic were
`
`looking at different places.
`
`But, anyway, let me hear Epic Games' view on that.
`
`MS. MOSKOWITZ: Thank you, Your Honor.
`
`We do think that what Your Honor is proposing is better
`
`than what Apple is proposing, certainly. Though the way the
`
`math would work out, if we were including both the search terms
`
`recall, meaning hits and non-hits, and the recall on the
`
`
`
`Case 4:19-cv-03074-YGR Document 183 Filed 12/10/20 Page 19 of 46
`
` 19
`
` 1
`
` 2
`
` 3
`
` 4
`
` 5
`
` 6
`
` 7
`
` 8
`
` 9
`
`10
`
`11
`
`12
`
`13
`
`14
`
`15
`
`16
`
`17
`
`18
`
`19
`
`20
`
`21
`
`22
`
`23
`
`24
`
`25
`
`review, we would have to hit -- almost 97 out of every hundred
`
`needs to be in agreement. And I couldn't even agree with
`
`myself 97 out of a hundred times.
`
`And so we do -- we do think search terms are different, as
`
`Your Honor pointed out. We do think that the very nature of
`
`them allowed Apple to have input on what was going to be
`
`reviewed. And so we should move past that and then do the
`
`recall on the review.
`
`We had proposed elusion because, frankly, whether or not
`
`they got extra documents as responsive that maybe they would
`
`have thought were nonresponsive, so what.
`
`And, really, what we should be looking at is, were there
`
`really responsive documents in what we were not producing and
`
`marking "nonresponsive." That's what we thought, really, we
`
`should be trying to get at, because is anyone really going to
`
`complain from Epic producing a few documents that maybe could
`
`have been marked "nonresponsive" when people were being really
`
`careful and circumscribed in what we were calling responsive,
`
`which we're not doing.
`
`So that is why we proposed it. We thought that would give
`
`Apple the comfort that there weren't a whole batch of documents
`
`sitting as a nonresponsive code, that they wouldn't get. That
`
`is the thought behind it.
`
`But if Your Honor was focused on doing an overall recall
`
`number, we would be prepared to meet and confer with Apple and
`
`
`
`Case 4:19-cv-03074-YGR Document 183 Filed 12/10/20 Page 20 of 46
`
` 20
`
` 1
`
` 2
`
` 3
`
` 4
`
` 5
`
` 6
`
` 7
`
` 8
`
` 9
`
`10
`
`11
`
`12
`
`13
`
`14
`
`15
`
`16
`
`17
`
`18
`
`19
`
`20
`
`21
`
`22
`
`23
`
`24
`
`25
`
`to do that on the percentage of documents that hit on search
`
`terms and, thus, were actually linearly reviewed.
`
`THE COURT: All right. Thank you. That is what I'm
`
`thinking.
`
`Let me turn to the -- let me just look at my notes to make
`
`sure that I -- so now I have a question for Apple.
`
`Epic Games says that of the 3.8 million documents you
`
`produced, 2.2 were these bulk emails. If I were to say you
`
`have to exclude those before you calculate the recall rate, how
`
`would that affect -- how would that affect you?
`
`MR. DETTMER: It would be a dramatic problem in the
`
`whole review, Your Honor, for a host of reasons.
`
`One is, as Your Honor probably knows, Judge Gonzalez
`
`Rogers talked about how we should not be relitigating the
`
`earlier production and review, and this would be exactly that.
`
`You know, Epic said twice in their letter to you that
`
`these are responsive documents. They're certainly responsive
`
`documents. It was admitted twice in the letter. They claim
`
`that they're not relevant. That is just not true.
`
`I mean, these documents are part of what Apple does to
`
`review apps and app providers, developers in the App Store
`
`ecosystem. It does things -- these emails show things like
`
`approval of selling apps in different areas. It shows, you
`
`know, changes to apps that may be needed to be followed up for
`
`security or quality reasons. It shows, you know, when there
`
`
`
`Case 4:19-cv-03074-YGR Document 183 Filed 12/10/20 Page 21 of 46
`
` 21
`
` 1
`
` 2
`
` 3
`
` 4
`
` 5
`
` 6
`
` 7
`
` 8
`
` 9
`
`10
`
`11
`
`12
`
`13
`
`14
`
`15
`
`16
`
`17
`
`18
`
`19
`
`20
`
`21
`
`22
`
`23
`
`24
`
`25
`
`have been changes to how the apps are actually operating that
`
`Apple needs to follow up on.
`
` So these are all part of showing how Apple takes these
`
`dramatic steps and then puts all this investment into making
`
`the App Store ecosystem safer, better for developers and
`
`consumers. So these are, you know, very relevant to the claims
`
`and defenses in this case. And the fact that Epic doesn't
`
`think they are is just, A, not right. They are part of our
`
`defenses in the case and certainly relevant to those.
`
`So, you know, in calculating recall, which, as Your Honor
`
`knows, is just the percentage of documents that are responsive,
`
`they're responsive. I mean, it's admitted. It's just true.
`
`And if we had to go back and redo our production now
`
`because of sort of a second-guessing of, you know, whether
`
`these documents were relevant enough, which I don't even know
`
`how you impose that kind of standard, would --
`
`THE COURT: You're getting a little bit away from my
`
`question. I'm certainly not proposing that you redo your
`
`document production. And I understand the philosophy behind
`
`including those documents.
`
`The concern I had is that there -- it looked like most of
`
`your document production are these highly repetitive --
`
`relevant, yes, but also low-value documents. And I'm worried
`
`that -- that the inclusion of those could mean that if your
`
`recall rate was terrible on everything else, you would,
`
`
`
`Case 4:19-cv-03074-YGR Document 183 Filed 12/10/20 Page 22 of 46
`
` 22
`
` 1
`
` 2
`
` 3
`
` 4
`
` 5
`
` 6
`
` 7
`
` 8
`
` 9
`
`10
`
`11
`
`12
`
`13
`
`14
`
`15
`
`16
`
`17
`
`18
`
`19
`
`20
`
`21
`
`22
`
`23
`
`24
`
`25
`
`nonetheless, pass the 75 percent threshold.
`
`And so I just mean operationally, from a TAR validation
`
`standpoint, what if I said, look, knock out the 2.2 million.
`
`Can you describe -- would this have any -- I don't want to do
`
`something that would make it too difficult for Apple to satisfy
`
`the validation. That's the point I'm getting at. I don't want
`
`to tell them to produce something that's impossible or very
`
`difficult.
`
`And so I'm asking, as an operational method, if you had to
`
`block out the 2.2 million, those documents, what effect, if
`
`any, would that have on your ability to get a recall rate of
`
`75 percent?
`
`MR. DETTMER: It would -- I don't -- we haven't done
`
`it at this point, but it would certainly affect, probably very
`
`negatively, our ability to do that in a way that I think is
`
`both, you know, unfair -- I mean, look, Your Honor, we reviewed
`
`all these documents. I mean, these weren't bulk-coded
`
`documents. They weren't, you know, machine coded.
`
`People went