ESI Protocol Best Practices for Search Terms and Production Format Before Discovery Starts
Learn how to negotiate ESI protocol search terms, production formats, metadata, and change procedures before discovery begins.
ESI protocol disputes rarely start with a disagreement about whether electronically stored information should be produced. Instead, the friction is typically over what gets searched, how the search is run, and whether the production arrives in a form counsel can actually use. These decisions are much easier to resolve before collection, review, and production are already in motion.
In federal litigation, Rule 26(f) requires the parties’ discovery plan to address issues involving electronically stored information, including the form or forms of production. Rule 34 allows a requesting party to specify a production form and set defaults when no form has been agreed. An ESI protocol is one way to turn those obligations into practical rules for the matter, although the Federal Rules do not require every case to use a document with that title.
However, timing matters — once a team has reviewed a large population under one search approach or begun producing data in an agreed format, changing course can create avoidable rework and new disputes. This guide focuses on two decisions worth settling early: search methodology and production format.
For broader lifecycle context, see our eDiscovery guide for law firms.
Key Takeaways:
- Negotiate the methodology, not just the keywords: A list of terms does not show how a search will be scoped, tested, or refined. Agreeing on the methodology upfront gives both sides a clearer basis for judging whether the search is reasonable and reduces the risk of expensive rework later.
- Test assumptions before they become protocol language: Hit counts alone cannot tell you whether a query is finding the right material or missing known responsive documents. Sampling and comparing results before the protocol is finalized can surface problems while the parties still have room to adjust.
- Choose production formats for how the evidence will be used: Native files, images, searchable text, and metadata preserve different kinds of context and functionality. Resolving those choices early makes the resulting production easier to review, analyze, cite, and defend.
- Build change into the protocol from the start: New custodians, data sources, terminology, or technical constraints often emerge after discovery begins. A defined process for revisiting search and production decisions keeps ordinary course corrections from turning into avoidable disputes.
An ESI Protocol Should Resolve the Hard Choices Before the Data Starts Moving
A useful ESI protocol reduces ambiguity before ambiguity becomes expensive. It should tell the parties enough about preservation, search, review mechanics, and production that each side understands the operating rules of discovery. It should also leave room for the facts to change.
Keep in mind that this doesn’t mean that every case needs the same level of detail. A protocol for a focused contract dispute might look very different from one involving dozens of custodians, collaboration platforms, databases, and years of communications. The goal is not to reproduce a model order. It’s to identify the decisions that could materially affect scope, burden, usability, or later motion practice and resolve as many of them as practical at the outset.
Courts may also expect counsel to address ESI early. For example, the Northern District of California’s ESI Guidelines and Checklist are meant to be consulted at the beginning of a case and specifically direct attention to preservation, search methodology, and production format. Local rules, standing orders, and judge-specific practices should therefore be part of the preparation, not an afterthought.
Build Your Negotiation Position From the Matter, Not From a Sample Protocol
Model protocols are useful because they show the categories of decisions counsel may need to make. They’re less useful when the language is copied before the team understands the systems, people, and issues that make this matter different. A stronger starting point comes from the record you have, the information environment you expect, and the authorities that govern the case.
Map the Likely Custodians, Systems, and Data Sources
Before negotiating search terms, identify the people and systems most likely to contain unique evidence. That usually means mapping likely custodians, relevant date ranges, and sources such as email, document management systems, shared drives, collaboration platforms, messaging tools, mobile data, cloud applications, or structured databases (where relevant). You don’t need a complete forensic collection map before the first conversation. However, you should have enough information to understand where responsive material is likely to live, which sources duplicate one another, and where a seemingly simple search request could become disproportionately burdensome.
Understand the Language of the Dispute Before Proposing Search Terms
Search terms should reflect how the people in the matter actually talked about the underlying facts. Build the vocabulary from pleadings, key documents, interviews, and known communications. Capture names, product or project terms, acronyms, internal shorthand, aliases, misspellings, code names, and language that changed over time. This is where generic search-term lists tend to break down. A term that looks precise in a complaint may never have been used by the business, while another term may be common enough to return thousands of irrelevant hits. Starting from matter-specific language makes the eventual negotiation more concrete.
Review Precedent, Local Rules, and Judge-Specific Expectations
Prior protocols can show which issues repeatedly matter to your team, but they should be treated as precedent, not boilerplate. Review local rules, model orders, standing orders, and relevant decisions before carrying old language forward. If a court has a preferred ESI checklist or model order, use it as a floor for the conversation. Legal AI can make that preparation faster without replacing the lawyer’s judgment. Harvey can research legal and regulatory questions against trusted knowledge sources, while agentic search across iManage, SharePoint, and Google Drive can help teams find prior protocols, briefs, and work product without knowing the exact filename or folder first.
Negotiate a Search Methodology, Not Just a List of Keywords
A list of search terms is only the visible part of the search process. The more important questions are how those terms were developed, what population they run against, how results are tested, and what happens if the first approach is plainly overinclusive or misses known responsive material.
Build Proposed Terms Around the Issues You Actually Need to Find
We recommend starting with the factual issues and information you’re trying to locate, then work backward to candidate terms. That might mean combining names, phrases, acronyms, and synonyms with date, custodian, source, Boolean, or proximity restrictions. The right construction depends on the language and systems in the case. For example, a broad word such as “termination” may be almost useless across a company’s email. The same concept paired with a product code, project name, relevant custodian group, or proximity to issue-specific language will probably produce a much more workable result. The point is not to make every query complex, it’s to make the search logic explainable.
Test the Search Before Treating the Terms as Final
Hit counts are useful, but they aren’t validation. Review a representative sample of results and ask yourself two questions: What is the query retrieving that clearly does not belong, and what responsive material do we already know about that the query fails to find? Both questions expose problems a raw count can hide. Testing could also reveal that different data sources need different treatment. For example, a phrase that works well in email may perform poorly in short messages or a structured system. Where practical, compare alternative constructions before hard-coding one set of terms into the protocol.
Decide How Technology-Assisted Methods Fit Into the Protocol
Keyword searching may be one part of a broader identification and review strategy. Depending on the matter, teams may also use analytics, clustering, classification, technology-assisted review, or generative AI (GenAI). The protocol should be clear about the method the parties are actually agreeing to, without assuming one technique is categorically more defensible than another. How much methodology must be disclosed or validated will depend on the applicable rules, orders, agreements, and circumstances of the case. Avoid drafting a protocol that imposes technical requirements neither side can explain or that freezes the matter into a process before the underlying data is understood.
Agree on What Happens When the Original Terms Don’t Work
Search assumptions change. A new custodian may surface, an internal acronym may appear for the first time, or an agreed term may return far more data than anyone expected. The protocol should make those developments manageable instead of turning each one into a new fight. At minimum, define the triggers for another discussion, who will raise the issue, what information the parties will exchange, and how quickly they’ll meet and confer. A simple modification process gives counsel a way to refine search terms or scope without pretending the first set of assumptions will survive the entire case unchanged.
Choose Production Format Based on How the Evidence Will Actually Be Used
Production format is not just an eDiscovery vendor setting. It affects whether lawyers can search, sort, reconstruct, and understand the evidence they receive. Rule 34’s production-form provisions recognize that different types of ESI may call for different forms and that, absent agreement or court order, electronically stored information must be produced as ordinarily maintained or in a reasonably usable form.
Decide When Native Files or Image Productions Make Sense
Native production can preserve functionality that an image can’t. Spreadsheets often contain formulas, filters, hidden rows, and multiple tabs. Presentations might include speaker notes and embedded objects. Documents could contain comments or tracked changes. In those situations, flattening the file can strip away information that matters to review. That doesn’t mean that every file should be produced natively. Image-based production can make sense for many document types, especially when Bates labeling, redactions, or consistent rendering are important. The protocol should identify the exceptions rather than force every type of ESI into one format.
Specify the Text and Metadata Needed to Make the Production Usable
Agree on the text and metadata that will make the production workable in practice. Depending on the matter, that might include extracted text, OCR for image-only material, custodian information, file paths, sender and recipient fields, dates, time zones, document family relationships, or attachment information. However, more metadata is not automatically better. Focus on fields that serve a review or evidentiary purpose. If a field will be important for filtering, deduplication, threading, chronology, or understanding document relationships, address it before production begins.
Address Bates Numbering, Redactions, and Exceptions
The protocol should also explain how native files will be referenced, how redacted documents will be handled, and what happens when a file cannot sensibly be produced in the standard format. For native productions, that can include a placeholder or slip sheet tied to a Bates number. For redacted material, it may require a different production approach from the unredacted original. The exact mechanics will vary by platform and case, but the important point is to remove surprises before the first production. This way, it minimizes the risk of discovering midstream that the agreed format does not support the way the parties need to cite or review the material.
Don’t Force Modern Data Into Assumptions Built for Email
Email is no longer the only model for ESI. Short-message platforms, collaborative documents, linked files, structured data, audio, video, and mobile content can carry relationships that disappear when the information is exported as disconnected pages. This doesn’t require native production of everything, but it does require counsel to ask what context would be lost by the proposed form. For chats, that may be conversation sequence and participant data. For a linked document, it may be the relationship between the message and the version of the file that existed at the relevant time.
Search Methodology and Production Format Need to Work Together
Search and production are often negotiated in separate sections, but they’re still connected. Search methodology determines why a document or communication enters the review set. Production format determines whether the information that made it meaningful survives when the other side receives it.
Suppose a search identifies a short-message thread because several participants discuss a shared spreadsheet. If the final production separates the messages from the linked file, strips the thread sequence, or loses the version relationship, the receiving team may get the words without the context. The same problem can happen when a spreadsheet is identified as responsive but produced only as static images that hide formulas or structure.
The protocol should therefore connect the two decisions. Ask whether a search can find responsive information, as well as whether the agreed production will preserve the relationships lawyers need to evaluate it. That’s the difference between producing data and producing usable evidence.
Build Flexibility Into the Protocol Before Something Changes
No protocol can predict every source, custodian, or technical problem a matter will uncover. But a good one makes change easier to manage by defining when the parties should revisit an assumption and how that conversation will work.
Define the Events That Should Trigger Another Conversation
Common triggers include a newly identified custodian or system, new terminology, search results that are far broader or narrower than expected, a production format that proves unusable for a particular evidence type, or facts that materially change the reasonable scope of discovery. The protocol does not need an exhaustive list, but it should make clear that material new information can justify revisiting the original approach.
Agree on a Process for Resolving Disagreements
Set a practical meet-and-confer process before there’s a dispute. Identify what information a party should provide when requesting a change, how quickly the other side should respond, and when an issue should be escalated. A defined process can keep technical disagreements from drifting into motion practice simply because no one agreed on the next step.
Preserve an Audit Trail of What Changed and Why
Keep a record of material changes to search terms, custodians, date ranges, production exceptions, and other protocol decisions. The record doesn’t have to be elaborate, but it should be clear enough that the team can later explain what changed, why it changed, and whether the change was agreed, ordered, or implemented under an existing provision of the protocol.
Use This ESI Protocol Checklist Before You Agree to the Final Language
A model protocol can help you remember the categories, but the final language should reflect the matter in front of you. Before signing off, use the following questions as a negotiation checklist rather than as a substitute for case-specific legal judgment:
Search Methodology
- Which custodians, systems, and date ranges are likely to matter?
- What factual issues and matter-specific vocabulary informed the proposed search terms?
- How were the terms or other search methods tested against known material?
- Will Boolean, proximity, analytics, TAR, or other techniques play a role?
- What results or new facts would trigger refinement of the methodology?
- What information will the parties exchange when proposing a change?
Production Format
- Which file types should be produced natively and which can be imaged?
- What searchable text and metadata fields are actually needed?
- How will attachments, document families, and linked files be preserved?
- How will chat, structured data, audio, video, or other non-email sources be handled?
- How will Bates references, redactions, and production exceptions work?
- How will the parties document changes to the agreed production process?
Once the protocol is settled, the same matter-specific discipline should carry into the requests themselves. Our guide to drafting discovery requests and responses covers how to ground and verify that work before service.
Harvey Helps Litigation Teams Prepare for the Protocol and Use the Discovery That Follows
Harvey is not a forensic collection or production-processing engine. Its role is to help lawyers find, research, analyze, and organize the information that informs their decisions before the protocol is negotiated and after discovery starts arriving.
Prepare for the Negotiation Using Your Own Knowledge and Matter Materials
Before a meet-and-confer, teams can use connected matter content to locate prior protocols, playbooks, briefs, and correspondence, then use Harvey knowledge sources to research governing authority and firm-specific context. Harvey Agents can synthesize large matter records, compare proposed positions, and return cited work for lawyer review. These capabilities fit naturally into Harvey for Litigation, where discovery, analysis, drafting, and case strategy sit in the same broader workflow.
Carry Those Decisions Into the Discovery Work That Follows
Once productions begin, Harvey Vault can help litigation teams organize discovery materials and turn sprawling requests, responses, and emails into structured review tables. Legal Work Orchestration can also help teams assign and track litigation research and drafting tasks while keeping lawyer review in the loop. This keeps the protocol connected to the work it was meant to govern. If a new source appears or a production raises an issue, the team can work from the same matter context when deciding what needs to change.
Better Protocol Decisions Make the Rest of Discovery Easier to Manage
The best time to resolve uncertainty about search and production is before your team spends weeks working under the wrong assumptions. A strong ESI protocol gives both sides a clear starting point, preserves the context that makes evidence useful, and creates a practical way to adjust when the matter changes.
For litigation teams, that means less time correcting avoidable process problems and more time evaluating the record and shaping strategy. Explore Harvey for law firms to see how connected research, document analysis, agents, and matter workflows support that work, or request a demo to see Harvey on your own legal workflows.
