Privilege Logs, Null Sets, Search Strings, and Number of Custodians in One Decision

[EDRM Editor’s Note: The opinions and positions are those of Michael Berman.]

The decision in Cook v. Meta Platforms, Inc., 2024 WL 251942 (N.D. Cal. Jan. 21, 2025), packs a lot into very few pages.  In two instances, where Meta had offered a compromise solution, the court held Meta to that offer.

           I.     PRIVILEGE LOG DISPUTE

First, the court resolved a privilege log dispute under the ESI Order.  Meta had stored some of the documents in databases and “due to the records associated with those databases, it is sometimes difficult or impossible to identify (a) the beginning date range; (b) the end date range; (c) the names of the authors and recipients, including identification of any in-house or outside counsel.”  Plaintiff was “skeptical” and “thinks that any decent document management system should record this information.”

The Court’s view is that a privilege log should be robust, but it also shouldn’t be an overwhelming burden. The Court is not going to require Meta to spend many hours rummaging through deprecated databases to find every last person who may have contributed to or received a document, particularly as Meta represents that for some of these documents, it is likely impossible for Meta to come up with a complete answer. Of course, if Meta takes the position that it cannot definitively say who all the authors or recipients were of a given document, that could potentially have implications for whether that document is privileged. Plaintiff is quite candid that she is interested in making precisely that kind of challenge to Meta’s claims of privilege.

Cook v. Meta Platforms, Inc., 2024 WL 251942 (N.D. Cal. Jan. 21, 2025).

The Cook court ordered Meta to provide information that “it is able to discover through reasonable diligence.”  For any item that Meta deems unduly burdensome, “Meta must serve a declaration that explains why that is impossible or unduly burdensome and otherwise substantiating Meta’s claim of privilege under Rule 26(b)(5). This explanation must be made on a document by document basis, although multiple documents may be addressed in the same declaration.”

         II.     NULL SET DISPUTE

The ESI Order included a provision requiring a search of the null set – a/k/a the discard pile.  A null set consists of the documents that are not returned by keyword searches.  The concern is whether it contains “false negatives” that would impeach the accuracy of the keyword search. 

The ESI Order stated in part:

After applying its search terms to the targeted document set(s), that Producing Party will review a statistically valid (95% ±5), randomly generated sample set of the documents that do not hit upon any search terms (“Null Set Sample”) and produce to the Requesting Party the relevant, responsive, non-privileged documents [in that null set]. The Producing Party will develop appropriate new search term(s), if needed, to capture any relevant documents located within the Null Set Sample. The Producing Party will then apply the original and new search terms to the [full] targeted document set and disclose the original and new terms to the Requesting Party, along with a hit report.

Id.

The ESI Order then set out a four-step process that the requesting party could initiate.

A dispute arose over the application of the ESI Order and:

Meta wants to clarify that the requesting party’s opportunity to object to the sufficiency of the producing party’s original and new search terms is triggered only if the producing party identified responsive documents in the null set sample. The Court agrees and thinks that is already implied. The purpose of the null set sample is to test the quality of the original search terms. If responsive documents show up in the null set, that means the original search terms were imperfect, which gives rise to the obligation to develop new search terms “if needed.”  The ESI Order refers to new search terms as being “if needed” because if there are no responsive documents in the null set, no new search terms are needed, and if there are close to no responsive documents in the null set, perhaps the original search terms, though imperfect, are good enough. If responsive documents show up in the null set, the original search terms were to some degree flawed, and the new search terms reflet [sic – reflect]  the producing party’s effort (if any) to remedy that problem, and this gives rise to the requesting party’s opportunity to object to the original search terms (which have been shown to have at least some problem) and to the new search terms (which the requesting party might not think are a good enough solution). But it’s the presence of responsive documents in the null set that indicates there is any sort of problem with the original search terms that might need to be fixed. Meta is correct that the absence of responsive documents in the null set means that the remaining steps in paragraph 5 are not triggered. [emphasis added].

Id.

In other words, if there are few or no “false negatives” in the discard pile, the keyword search terms have been sufficiently accurate in retrieving responsive information.

      III.     THE SEARCH STRING DISPUTE OVER THE NULL SET

This dispute centered on seven search strings.  The court wrote:

Meta says that it ran a null set, which contained no responsive documents. Plaintiff later demanded that Meta run 77 additional search terms. Meta says that it agreed to adopt more than 50 of Plaintiff’s proposed additional search terms, then agreed to use six more, and drew the line at the seven now in dispute. Given Meta’s view (discussed above) that the producing party should not be able to object to search terms unless the null set has responsive documents, why did Meta agree to any of these additional search terms? The answer, as clarified at the hearing, is that Meta did not stand on that objection and was attempting to resolve this dispute. Accordingly, as Meta has not stood on that objection, the Court considers it waived for the present dispute. [emphasis added].

Id.

The Cook court then wrote:

The hit counts for search strings 3 (19,073 documents) and 4 (17,128 documents) are not overly burdensome, and the search strings seem aimed at relevant subjects. The Court ORDERS Meta to use search strings 3 and 4. The hit counts for search strings 1, 2, 5, 6 and 7 are excessive given the subject matter. The Court ORDERS Meta to counterpropose search strings that substantially reduce the hit count for those search strings.

Id.

The court also ordered the parties to meet and confer about the counterproposal.

      IV.     THE NUMBER OF CUSTODIANS DISPUTE

Plaintiff wanted Meta to increase the number of custodians from 10 to 21. The court denied the motion without prejudice.  It wrote: “More than doubling the number of Meta’s document custodians at this stage poses problems under the case schedule.”  Fact discovery was due to close in two months. The Cook court wrote:

If the Court granted Plaintiff’s request for more document custodians on the day it was filed (January 8) and ordered Meta to more than double the number of document custodians, Meta would have to scramble to substantially complete its custodial document productions by the closing days of fact discovery. Plaintiff would inevitably ask for an extension of the fact discovery cutoff, citing the need to take depositions after documents have been produced….  Plaintiff’s late-breaking request to more than double the number of Meta’s document custodians would probably blow up the fact discovery cutoff. Certainly, it is not compatible with the existing case schedule.

Id.

The court added: “In an attempt to resolve this dispute, Meta offered to add one or two additional custodians…. The Court will hold Meta to its offer, which Meta confirmed at the hearing it can accomplish by the current close of fact discovery.”

In a prior decision, Cook v. Meta Platforms, Inc., 2024 WL 4133811 (N.D. Cal. Sept. 9, 2024), the court wrote:

So, here is the deal with a court order, especially one that you stipulated to. It’s an order, so you have to comply with it. If you learn new information that makes it seem like it will be difficult or impossible for you to comply with the order, then you should move to modify the order with reasonable diligence, explaining why the order has become difficult or impossible for you to comply with. You might win your motion, or you might lose, depending on how good your argument is. But that’s what you’re supposed to do if you find yourself in that situation. What you’re not supposed to do is pretend like the order doesn’t exist, or pretend that your violations are somehow a form of compliance, and then hope your opponent doesn’t point out what you’re doing.

Cook v. Meta Platforms, Inc., 2024 WL 4133811 (N.D. Cal. Sept. 9, 2024).

See Is a Court-Ordered ESI Protocol a Trap? (Sep. 14, 2024).

Assisted by GAI and LLM Technologies per EDRM GAI and LLM Policy.