Work Co. d/b/a Toku v. Liquifi, Inc.

EFiled: Apr 21 2025 05: o1P EDT Lin
Transaction ID 76118162 “ify 4A")
Case No. 2024-1334-JTL 3 Nee ish

IN THE COURT OF CHANCERY OF THE STATE OF DELAWARE ey orbs mee

WORKCO, INC. d/b/a TOKU,
Plaintiff,

C.A. No. 2024-1334-JTL

Vv.

LIQUIFI, INC. and BENJAMIN SNIPES,

Defendants.

ORDER ADOPTIN
SPECIAL DISCOVERY MAGISTRATE’S REPORT

WHEREAS:

A. On February 6, 2025, WorkCo., Inc. d/b/a Toku (“Toku”) moved to
compel LiquiFi, Inc. to apply Toku’s proposed search terms to identify
responsive documents.

B. On March 7, 2025, the court appointed a Special Discovery
Magistrate.

C. On April 8, 2025, the Special Discovery Magistrate a report
recommending denying Toku’s Motion To Compel Defendants To Apply Search
Terms And Collect Documents From The Relevant Period.

D. No party has taken exceptions to the Special Discovery
Magistrate’s recommendation.

IT IS HEREBY ORDERED:

1. Toku’s motion is denied on the basis of the Special Magistrate's

report.
2. The court adopts the report, attached as Exhibit A, as a ruling of

the court.
3. The parties must continue to cooperate in identifying any

additional targeted searches that may become necessary as forensic analysis

progresses.

2 Chancellor Laster
| 21, 2025
Bled: Api 3b 2035 Ve aPMe

Taps apne A)
CaReNe, eee YP

IN THE COURT OF CHANCERY OF THE STATE OF DELAWARE

WORKCO, INC. d/b/a TOKU,

Plaintiff,

Vv.
C.A. No. 2024-1334-JTL

LIQUIFI, INC. and BENJAMIN
SNIPES,

Defendants.
SPECIAL DISCOVERY MAGISTRATE’S REPORT
RECOMMENDING DENYING
PLAINTIFF’S MOTION TO COMPEL DEFENDANTS TO APPLY SEARCH TERMS
AND COLLECT DOCUMENTS FROM THE RELEVANT PERIOD
Date Submitted: April 8, 2025

Tara Emory, Special Discovery Magistrate

I. INTRODUCTION AND PROCEDURAL BACKGROUND

Plaintiff WorkCo., Inc. d/b/a Toku (“Toku’”) has moved to compel Defendant LiquiFi, Inc.
(“LiquiFi”) to apply Toku’s proposed search terms to identify responsive documents. By Order
dated March 7, 2025, Vice Chancellor Laster appointed me as Special Discovery Magistrate to
facilitate resolution and submit this Report.

This discovery dispute, and the subsequent search process to resolve it, demonstrates an
evolution from ineffective “go fish” keyword searches to a methodical, collaborative, and
effective approach aligned with the case’s specific information needs.

The case involves claims of misappropriation of trade secrets and confidential
information. Toku alleges that its former in-house attorney Benjamin Snipes (“Snipes”) became

discontented in March 2024 and, before his departure in July 2024, downloaded over 25,000 files
containing Toku’s confidential information and trade secrets. Compl., D.I. 1, {9 3, 12, 60-66.
Snipes was hired by LiquiFi in September 2024. Toku claims he shared its intellectual property
with LiquiFi, including information relating to a real-time tax calculation engine for
cryptocurrency-based compensation and a token-based pricing model. Id. {J 12-14, 67-82. Of
particular significance is Toku’s allegation regarding a “prompt” questionnaire, such as “LiquiFi
x Ben Snipes Prompt,” that Toku claims LiquiFi used when hiring Snipes to determine whether it
would disclose Toku trade secrets. Id. Ff] 4-7, 13, 74-82. Toku alleges similar prompt documents
were used when it later hired another Toku employee, Kirk Alcock (“Alcock”). Id. ff] 83-89.
Toku further alleges that LiquiFi is using that information to compete unfairly and has made
untrue statements that disparage Toku. Id. J] 91-94.

On January 8, 2025, Toku served its first set of requests for production. The parties
attempted to negotiate but did not agree on culling search terms or the relevant time period for
the document search. LiquiFi applied its selected search terms, reviewed those documents for
relevance, and produced 332 documents. Plaintiff’s Motion to Compel Defendants to Apply
Search Terms and Collect Documents from the Relevant Period “Motion to Compel” or “Pl.’s
Mot.”), D.I. 74, 9] 2, 10. LiquiFi declined to apply Toku’s more extensive proposed search
terms, or to include documents from a broader time period identified as relevant by Toku.
Defendant LiquiFi, Inc.'s Opposition to Pl.’s Mot. (“Def.’s Opp.”), D.I. 82, Ff 8-10. On February
6, 2025, Toku filed the Motion to Compel, which sought to have LiquiFi run its requested search
terms, provide a hit report for those terms, and then, without any additional relevance review,
produce all documents that hit on the terms. Pl.’s Mot. J 21-25. The motion was resolved as to
Mr. Snipes on February 28, 2025, in an Order that withdrew Plaintiff’s Motion for Contempt of

the Status Quo Order Against Snipes. D.1.107.
This Report focuses on LiquiFi’s methodology for searching and identifying responsive
documents within its corporate data sources for the agreed categories of information. Some
aspects of discovery in this matter are proceeding through separate processes and are outside the
scope of this Report. These include forensic examinations of devices and systems to identify the
25,000 files allegedly taken by Snipes, forensic analysis to identify any deleted files, analysis of
Snipes’ personal devices, and certain categories of documents related to damages but not relevant
to these preliminary injunction proceedings.

Il. THE PRIOR SEARCHES AND APPOINTMENT OF SPECIAL DISCOVERY

MAGISTRATE

Throughout January and early February 2025, counsel for the parties met and conferred
regarding search terms. PIl.’s Mot. Exs. 3, 4, 5. LiquiFi’s counsel provided hit count reports for
dozens of search term strings. Negotiations focused on search term hit volumes rather than
search effectiveness. While this is a traditional approach to addressing eDiscovery searches, each
party’s proposed search terms were ultimately shown to be ineffective in finding a reasonable
amount of relevant documents. Notably absent from the parties’ discussions were considerations
of alternative search methodologies or validation of search effectiveness.

Applying its selected search terms and other targeted searches for documents dated from
September 1, 2024 through February 2025, LiquiFi reviewed in excess of 12,000 documents but
produced only 332, with a resulting responsiveness rate of under 3%. Def’s Opp. {| 10. For
documents in the March | to August 31, 2024 range, which Toku requested to be included,
LiquiFi sampled the set, with Toku’s proposed search terms applied, and reported that the results
were less than 1% responsive (though it did not provide a margin of error or size of its sample).

Id. § 12. It accordingly declined to search or produce from that set based on low responsiveness.
Based on statistical sampling, it has since been determined with 95% confidence that the
overall richness (responsiveness rate) of LiquiFi’s documents from March 1, 2024-February
2025 lies in a range from about 2.5% to 3%. In other words, both parties’ search terms performed
about as well as a random selection from the documents. As we also now know, LiquiFi’s
production represented less than 4% of the responsive documents in LiquiFi’s data set (a metric
referred to as “recall”). LiquiFi’s approach failed to recognize that even in document sets with
low richness, significant numbers of responsive documents may exist.

In its March 7, 2025 Order Appointing a Special Discovery Magistrate, the Court held
that the relevant time period for search should include March 1, 2024 to present D.I. 122, 7 1.
The Order further noted that it did not appear “that the defendants have made a reasonable effort
to identify responsive documents,” and that “[c]onducting a meaningful search for responsive
documents is not so simple as merely applying the plaintiff’s search terms. That is the equivalent
of ‘go fish’”! Id.

Following the briefing of Toku’s Motion to Compel and prior to my appointment as
Special Discovery Magistrate, LiquiFi retained Troutman Pepper Locke (“Troutman”) as new
counsel. Troutman implemented effective search methodologies, and together with Toku’s

counsel Quinn Emanuel, the parties cooperated to resolve the challenges to LiquiFi’s search

' For many years, initial search terms have been referred to in caselaw and the eDiscovery
industry as “Go Fish” searches, referring to the child’s game in which a player asks for a card,
hoping they will be lucky — often unsuccessfully, resulting in being told to “go fish” for another
card. See Da Silva Moore v. Publicis Groupe & MSL Group, 287 F.R.D. 182, 190-191 (S.D.N.Y.
2012) (S.D.N.Y. 2012) (“In too many cases, however, the way lawyers choose keywords is the
equivalent of the child’s game of ‘Go Fish””), citing Ralph C. Losey, “Child’s Game of ‘Go Fish’
is a Poor Model for e-Discovery Search,” Adventures in Electronic Discovery 209-10 (2011) and
citing William A. Gross Constr. Assocs., Inc. v. Am. Mfrs. Mut. Ins. Co., 256 F.R.D. 134, 136
(S.D.N.Y. 2009) (“This Opinion should serve as a wake-up call to the Bar in this District about
the need for careful thought, quality control, testing, and cooperation with opposing counsel in
designing search terms or ‘keywords’ to be used to produce emails or other electronically stored
information”).
raised in Toku’s Motion to Compel, and led to LiquiFi’s production of over 8,000 responsive

documents (over 9,500 total, when including not-responsive family documents).

Ill. LEGAL STANDARD

Under Court of Chancery Rule 26(b)(1), parties may obtain discovery regarding “any
non-privileged matter that is relevant to any party’s claim or defense and proportional to the
needs of the case.” The scope of discovery under this rule is broad and far-reaching. See Jn re
Oxbow Carbon LLC Unitholder Litig., 2017 WL 959396, at *1 (Del. Ch. 2017) (noting that
“(rjelevance must be viewed liberally” and discovery should be permitted if “there is any
possibility that the discovery will lead to relevant evidence”).

Established legal frameworks recognize a producing party’s discretion to choose their
own production methodologies. The Sedona Conference Principle 6 acknowledges that
“responding parties are best situated to evaluate the procedures, methodologies, and technologies
appropriate for preserving and producing their own electronically stored information.” The
Sedona Principles, Third Edition: Best Practices, Recommendations & Principles for Addressing
Electronic Document Production, 19 SEDONA CONF. J. 1, 118 (2018). This principle
recognizes that the producing party has unique access to and knowledge of its own data
environment.

This principle, however, does not relieve the producing party of its obligation to conduct
reasonable searches for responsive information. The corollary to The Sedona Conference
Principle 6 is that a responding party must take care that its process is reasonably designed to
lead to production of relevant evidence. Particularly where the parties have not agreed otherwise,
the responding party should not rely on a requesting patty’s proposed terms as the sole method

for finding responsive documents, nor should it simply reject those terms without evaluating
their reasonableness or alternative options. Even where a responding party has demonstrated that
the requesting party’s proposed search terms are ineffective, the responding is not relieved of its
obligation to develop search methods that are reasonably designed to identify responsive

information.

IV. ALIGNMENT OF SEARCH METHODOLOGY WITH SEARCH GOALS

A. Ineffective Initial Approaches

Before a responding party determines its search methodology, it should first analyze what
types of information it is seeking, and which search approaches are likely to be most appropriate
for locating that information. Here, both parties defaulted to empirically-blind search term
negotiations, resulting in an inefficient process. The parties debated about unproductive search
terms. LiquiFi then conducted an inefficient and ineffective review based on search terms.
Toku’s Motion to Compel has requested relief based on more inefficient search terms. Neither
party initially considered whether search terms were the optima! approach for the information
sought.

In their negotiations, Toku requested broader terms, believing, without evidence, that its
approach was reasonable to retrieve the relevant information it sought. Meanwhile, LiquiFi
interpreted low search term hit rates for Toku’s proposed search terms and time period as proof
of absence of responsive documents. In actuality, neither conclusion served the goal of
conducting a reasonably effective search.

As the Court held in the Order of my appointment, and evidenced by LiquiFi’s
subsequent search process, the initial production process was deficient. LiquiFi did not represent
that it had attempted to test, iteratively develop, or validate its search terms. The 3%

responsiveness rate resulting from the review of the documents hitting on LiquiFi’s initial search
demonstrated that the terms were at minimum inefficient, and there was no empirical evidence
that the terms were reasonably designed to find responsive documents. Further, LiquiFi did not
demonstrate that the burden of additional search would be disproportionate to the needs of the
case because it did not identify any specific burdens based on reasonable attempts to design
effective searches. Efforts spent on conducting unreasonable search and review processes do not
count in the consideration of burden to a party.

In addition, LiquiFi’s finding that Toku’s requested search terms were only 1%
responsive (precision rate) as applied to the March-August 2024 dates did not justify abandoning
the search effort. While a reasonable search would seemingly require LiquiFi to at least consider
Toku’s requested searches, Toku’s ineffective “go fish” search terms did not excuse LiquiFi from
its duty to produce responsive documents.

LiquiFi supplemented its existing data sets with additional collections from custodial

devices and noncustodial LiquiFi data, including Google shared drives, Telegram, Notion, and

Slack.
B. Cooperating on Scope of Search Topics

Through the Special Discovery Magistrate process, in a series of meetings, the parties
identified and agreed to the scope of categories of relevant information for which LiquiFi should
search. Framing the negotiation this way shifted the parties’ discussion from debating search
term syntax to focusing on substantive information needs.

As is common in discovery, while Toku’s Requests for Production were numerous and
detailed, the parties did not initially share a common understanding on the scope of relevant
topics. For example, Toku wanted information about LiquiFi’s hiring practices for positions other

than those filled by Snipes and Alcock. While that information potentially fell within several of
Toku’s requests, LiquiFi’s prior search and review had not considered that a relevant topic. By
focusing discussion on information categories rather than search terms, they developed a shared
understanding of concepts from which LiquiFi could develop targeted and appropriate searches.
The parties’ initial, mismatched expectations about the scope of relevance illustrates an

important factor in discovery search negotiations: the parties’ interpretations of responsiveness
are inherently subjective and may differ significantly, even when both parties are acting in good
faith. LiquiFi’s prior limited production may have reflected not just ineffective search
methodology, but also a genuine misunderstanding of what information Toku considered
responsive to its requests for production. By discussing topics rather than isolated search terms,
the parties established a clearer, shared understanding of what information would be relevant to
the claims and defenses in this case, avoiding potential future disputes over whether documents
were improperly withheld. See, e.g., Winfield v. City of New York, 2017 WL 5664852 at *9
(S.D.N.Y. 2017) (ordering Defendant to expand its search and production because its
interpretation of relevance had been too narrow).

Through cooperative discussion, the parties agreed the topics to be searched related to:

1. LiquiFi’s sending and receipt of competitive information from potential hires;

2. “Prompt” documents used in hiring processes (as explained above);

3. Real-time tax calculators; and

4. Disparagement of Toku.
C. Designing Effective Searches

In designing and executing a reasonable search, parties have several tools available to
them, beyond search terms. Different search approaches can work better for different types of

information. After the parties agreed on scope of relevant topics, LiquiFi thoughtfully designed
searches to align with the specific information needs, using a combination of search techniques
involving metadata, targeted and highly responsive search terms as determined through
sampling, and Technology-Assisted Review (TAR).

{. Communications with or involving specific individuals: Given the significance of
former Toku employees Snipes and Alcock in this matter, and the potentially different
nature of documents related to them from before versus after they worked at LiquiFi, the
parties agreed that LiquiFi’s search would include:

o Prior to their respective hiring dates, all communications involving LiquiFi and
Snipes or Alcock. This search was based on only email and phone metadata.

o Search terms for discussions related to Snipes and Alcock from prior to their
hiring dates, which were tested though sampling. Terms demonstrating low
success and high burden were abandoned or refined. Search terms with higher
success and lower burden were applied. For example, Kirk Alcock’s first name
was determined to be a reasonable search term and was applied, while Ben
Alcock’s first name was determined to be disproportionately burdensome and
unreasonable as a search term in this data set.

2. Other LiquiFi documents: With the search categories established, and with
consideration for input from Toku, LiquiFi developed the following search methods for
documents not addressed by the Snipes and Alcock searches:

a. Potential Hires: LiquiFi identified the names of individuals who were
considered as potential hires during the relevant time frame. Using their

names, contacts and dates, LiquiFi conducted targeted searches for this
information. In addition, several targeted searches for discussions related
to hiring were developed, tested and implemented.

b. “Prompt” Documents: Searches were tailored to include file name
searches for “prompt.” In addition, search terms were used, involving
quoted language from the known prompt documents. For any prompt
documents located, LiquiFi agreed to also determine whether prior
versions might also exist on its systems.

c. Real-time Tax Calculators: The search for information related to real-
time tax calculators involved using a combination of high-precision search
terms, and relied on TAR to find additional documents.

d. Disparagement: The parties agreed that searching for “Toku,” “WorkCo”
and “WorkDAO” would reasonably identify most documents that involved
disparagement of Toku or other relevant discussions.

e. TAR: As mentioned above, to complement the targeted searches, LiquiFi

used TAR to find additional responsive documents.

V. RESULTS AND VALIDATION

Given the expedited nature of this case, LiquiFi worked to produce documents daily until
the search and production process was complete. Its search strategy that combined targeted
searches with TAR demonstrates the value of tailoring search techniques to the categories of
information to be sought.

LiquiFi shared information about the number of documents produced for each topic,
which provided transparency, resulting in Toku’s agreement that the searches were reasonable. In

all, LiquiFi produced 9,426 documents, of which 8,267 were responsive (the remainder being

10
family documents that were not responsive). The responsive document production included
3,948 documents resulting from metadata and targeted search terms, and 4,319 from the use of
TAR.

Based on samples of the unreviewed population of documents, the recall of LiquiFi’s
search, which involved multiple search techniques, is estimated to range between 88% and 98%,
to a 95% confidence level (meaning we can be 95% certain the actual value falls within this
range). That is a stark contrast to the initial production, which was indeed deficient and
represented under 4% recall. In addition to sharing metrics underlying the recall calculation,
LiquiFi shared the number of produced documents produced relevant for each topic for which it
searched. The results reflect that significant volumes of documents were produced for each topic,
suggesting the reasonable effectiveness of the search strategies for those topics.

The validation of the searches with statistical sampling illustrates why such sampling is
essential to effective discovery. The initial approach, which relied on untested search terms,
missed approximately 96% of responsive documents. Without testing and iterative revision of
search terms, and validation through sampling, this deficiency would have remained unknown.
Even in low-richness data sets (here, likely under 3%), significant volumes of responsive

documents may exist, and may be identified through a thoughtful search strategy.

VI.CONCLUSION AND RECOMMENDATIONS

LiquiFi has now implemented a reasonable search protocol tailored to the information
sought and proportional to the case’s needs.

Effective searching in electronic discovery requires parties to focus on fishing where the
fish are, rather than simply trying to cast wider nets with more search terms. Toku correctly

identified deficiencies in LiquiFi’s initial production. However, its Motion to Compel, which

11
demanded that LiquiFi produce documents with Toku’s chosen search terms, reflected the same
misunderstanding as LiquiFi about search term effectiveness. This approach would have resulted
in an inefficient, burdensome process unlikely to identify substantially more responsive
documents.

Both parties’ cooperation in developing topic-based search strategies proved effective and
ensured that the parties shared a common understanding about what documents would be
responsive. By first aligning on the scope of discovery topics rather than search techniques, then
designing targeted methodologies appropriate to each information category, the discussions
moved beyond ineffective search terms, to achieve more comprehensive results.

As the parties proceed with other fact discovery, they should remain ready to address new
information that may require additional targeted searches and productions. This is particularly
important given the nature of the allegations in this case, where new information may lead to
discovery of other responsive materials that, even if not obviously relevant at this stage, could
become significant as the case progresses. For example, if the forensic analysis demonstrates that
certain Toku documents were in LiquiFi’s environment, then the parties should collaborate to
identify additional searches that may help further identify if that information was used by
LiquiFi.

Accordingly, I recommend that the Court:

l. Deny Toku’s Motion to Compel Defendants to Apply Search Terms and

Collect Documents from the Relevant Period; and

2. Direct the parties to continue their cooperation in identifying any
additional targeted searches that may become necessary based on new

information, as forensic analysis progresses.

12
13

Respectfully submitted,

/s/ Tara S. Emory

Tara Emory

Covington & Burling LLP
One CityCenter

850 Tenth Street NW
Washington, DC 20001

Special Discovery Magistrate

Date: April 8, 2025