Slack, the collaborative workspace designed to streamline real-time communications between business teams, poses some challenges in the eDiscovery context (see our prior blog post). This blog post discusses a recent decision that explores the proper scope of discovery regarding Slack messages – specifically, whether each individual Slack message should be treated as a discrete document under Rule 34, or whether a court should view an entire Slack channel as a single, continuous document.

Background

In Lubrizol Corp. v. IBM Corp., plaintiff alleged defendant breached a contract between the parties and committed fraud and various torts in connection with a project to implement new enterprise resource planning software known as S/4HANA. (ECF Doc. No. 60).  Before the Magistrate was plaintiff (Lubrizol)’s motion to compel defendant International Business Machines Corporation (IBM) to produce Slack messages. (ECF No. 77).

Procedural History

On Feb. 1, 2023, IBM filed a letter requesting a telephonic discovery conference pursuant to Local Rule 37.1 regarding certain purported deficiencies in Lubrizol’s discovery responses. (ECF No. 62).

Shortly thereafter, Lubrizol filed a letter requesting a discovery conference regarding several alleged deficiencies in IBM’s discovery responses, including IBM’s alleged failure to produce complete conversations conducted through Slack. (ECF No. 70). Lubrizol asked the court to order IBM to produce the full Slack conversation for any Slack thread containing at least one responsive message. Id. The matter was referred to the Magistrate for purposes of resolving the discovery disputes. (See ECF non-document entry dated Feb. 10, 2023). 

On March 30, 2023, the parties appeared before the court for a telephonic discovery conference on all outstanding discovery issues and advised the court that some issues could be resolved without judicial intervention. In response, the court directed the parties to submit a joint status report advising which, if any, discovery issues remained unresolved despite the parties’ efforts. (See ECF non-document entry dated March 30, 2023).

The parties were able to resolve all outstanding issues except the dispute regarding IBM’s production of Slack messages. (ECF No. 74). The court directed the parties to submit written briefs setting forth their positions regarding the proper scope of discovery with respect to Slack messages. During the briefing, Lubrizol narrowed its position. Rather than request IBM be ordered to produce the entire conversation thread for any Slack thread containing at least one responsive message, it requested that (1) for any Slack conversation containing 20 total messages or fewer, IBM be required to produce the entire conversation thread, so long as the conversation contained at least one responsive message; and (2) for any Slack conversation thread containing more than 20 total messages, IBM be required to produce the 10 messages preceding and following any responsive message. (ECF No. 77).

IBM opposed Lubrizol’s application, advising it had reviewed all Slack messages that hit on any of the parties’ agreed search terms, as well as the 10 messages before and after any message that hit on a search term.[1](ECF No. 79). IBM further argued it produced any message in that window that provided context for the relevant discussion, even if the message did not hit on a particular search term. IBM argued that adopting Lubrizol’s proposal would require IBM to produce not relevant/responsive materials in violation of the Federal Rules of Civil Procedure and would impose an undue burden on IBM.[2]

Law & Analysis

Recognizing that courts have broad discretion in overseeing the scope of discovery and ruling on motions to compel, the court began its analysis by noting the requirements of relevance and proportionality set forth in Rule 26. It then noted the shifting burden on a motion to compel – that the party moving to compel “bears the burden of demonstrating [the] relevance” of the requested discovery (Rule 37 (a)(1)(B)) and if satisfied, “the burden shifts to the non-movant to show that to produce the information would be unduly burdensome.” Id.

The court then articulated: “[t]he sole issue … is whether IBM should be required to produce additional Slack messages to provide further context for the messages that IBM has already produced. IBM does not dispute that the materials in its possession, custody, and control include Slack messages that are relevant to the case. Nor does IBM argue that production of Slack messages generally is disproportionate to the needs of the case. Instead, IBM argues only that it has already produced all Slack messages that it is required to under the Federal Rules, and that the additional materials Lubrizol is seeking are irrelevant and would be unduly burdensome to produce.”  According to the court – the “parties’ dispute is centered around the question of whether each individual Slack message should be treated as a discrete document under Rule 34, or whether a court should view an entire Slack channel as a single, continuous document. IBM argues that each message is a discrete document because the messages are stored on the system as individual files known as “JSON” files. Lubrizol does not dispute that Slack messages are stored as individual JSON files, but it argues that a conversation should nonetheless be treated as a single document because the participants in a Slack conversation view the entire conversation at once, even if the conversation spans months or years.” Id. Emphasis added.

This issue of whether to treat Slack messages as individual documents whose relevance must be analyzed separately for discovery purposes was one of first impression for the court and evidently one of first impression in case law, too, as neither party cited to any cases on point.  So, in the absence of case law directly on point, both parties argued by analogy. “IBM asserts that, because Slack messages are stored as individual JSON files, Lubrizol’s proposal is the equivalent of requiring a party to produce irrelevant hard copy documents simply because they are [ ] in the same box as responsive documents. Lubrizol, on the other hand, argues that Slack messages are most analogous to emails or text messages, and further asserts that parties are regularly required to produce the entirety of an email or a text message conversation, even if portions of the document are irrelevant.”

The court rejected IBM’s position, noting hard copy documents “are not originally created in a single box and their subsequent placement in the same box may be the result of sheer happenstance.” In contrast to hard copy documents, which are scattered all over, the messages in a particular Slack channel appear in a single conversation and a participant in a conversation can view all of those messages at once, regardless of when they were originally sent.  And although the court determined Slack messages are most akin to text messages, that did not resolve the motion, as courts have adopted different approaches with respect to what portions of a text message must be produced.

The court then determined, “[g]iven the particular facts and circumstances here, …Lubrizol’s proposal is the most appropriate approach … and IBM should be required to produce surrounding messages even if … some or all of those messages are not relevant to this case. In reaching that conclusion, I am guided by several considerations.” The court then provided three considerations. First, there was a legitimate dispute regarding whether all of the messages IBM withheld were truly irrelevant. Some of the messages related to similar work performed by IBM for at least one other client, which Lubrizol argued were relevant to its arguments that IBM misled Lubrizol about how the project would be staffed, how IBM managed the project, and whether IBM had the expertise it claimed when it pitched the project. The court noted that it’s unclear what proportion of the messages encompassed by Lubrizol’s proposal related to other similar projects, but that it appeared to be undisputed that some of the messages withheld related to those projects, which weighed in favor of granting the motion to compel. Second, any of IBM’s concerns regarding production of reportedly irrelevant messages that might otherwise be sensitive were mitigated by the existence of a protective order. And third, IBM failed to establish it would be unduly burdensome to comply with Lubrizol’s proposal given the scope of this case, which the court noted, involved a substantial dispute between two large commercial parties. Therefore, the court found that the production of the additional Slack messages (within 28 days) was not unduly burdensome or disproportionate to the needs of this case.[3]

Conclusion

Of note, this decision is the first to recognize that Slack messages are not individual documents under Rule 34 comparable to a hard copy document, and instead Slack messages are most comparable to text messages and require context. Context could be more than or less than 20 messages, but it is important to consider this when assessing whether the production of full threads from Slack that hit on search terms will be required.

Regarding search terms – here the issue was one of context around search terms. And so, this case is a good reminder of the importance of identifying search terms based on the data and after having familiarity with the data

eDiscovery is very detailed and requires different considerations for each source of ESI. Lawyers need to understand these sources and the associated issues to argue effectively and draft ESI protocol (or subsequent agreements if generated separately) appropriately.

The final takeaway; get into the data. This applies whether you are the producing party or you have received a production. If you’re the producing party, get into your data and be prepared to show examples of it to bolster your argument. Here, IBM was not able to do that effectively for the court. The court doesn’t mention what they did provide, so we can’t know the full scope of what was presented. But it wasn’t enough. Lubrizol, on the other hand, argued effectively that data showing relevant messages existed beyond the context produced by IBM. Since it was IBM’s burden, that was enough to prevail.

Know your data. The receiving party should review the data for these issues as soon as it’s received. Waiting may prejudice your arguments with the court and/or mean the loss of valuable time with depositions and fact development. In motion practice it takes longer to get that information, and when you’ve already engaged in those depositions, or you’re getting close to the end of discovery and you’re just now getting that information. Waiting may prejudice the case or client. It’s expensive to go to motion practice, and it’s often unnecessary if you can demonstrate to the other side via data that your arguments will prevail. But if you can’t convince the other side, and you still have to go to motion practice, you will have done the bulk of the work to substantiate your arguments for the court.


[1] Although not articulated explicitly in the decision, the parties agreed upon search terms that IBM ran against its Slack messages and produced those responsive to the search terms. In prior posts I have discussed how Slack works in “channels” with two main ways to communicate. First, you can communicate via a channel where multiple people are members — perhaps to communicate about a specific topic — or you can use direct messages from one person to another within Slack. To understand the complexity of the tool, if using enterprise Slack, there could be thousands of channels and hundreds of thousands of different conversations. 

[2] Be mindful that in Slack, you can reply generally in a channel to someone, or you can reply in a thread to that specific message. Replying in the thread keeps the topics of that particular conversation together. But people use Slack differently, so understanding how one’s client/client employees use Slack is important to when determining what Slack data needs to be collected, reviewed, and produced. A lawyer’s duty of competence includes in this context understanding the data and how it’s organized.

[3] IBM requested that a similar ruling should apply to Lubrizol’s production of its messages from a different collaboration tool, which were produced individually without context. Lubrizol agreed to abide by the same protocol and that it would produce the 10 messages proceeding and following any message in its possession, custody, or control that was responsive to IBM’s discovery request or the entirety of any conversation containing 20 or fewer messages in total.