eDiscovery for law firms, including the small ones

Electronic discovery is often presented as a technology purchase, but the more consequential decision is operational: how much of the process will the firm control? A small firm handling occasional, modest-volume matters does not need the staffing model of an enterprise litigation-support function. It does need reliable preservation, a defensible collection scope, and enough technical judgment to test what a provider delivers before production.

Three operating models are practical. A firm can refer nearly everything to a provider, maintain a light in-house capability for smaller matters and refer larger ones, or build a sustained litigation-support capability. Each model has a different cost structure, and each fails in a different place.

Choose an operating model before choosing tools#

A provider-led model sends collection, processing, review, and production to an outside provider, leaving the firm as the legal decision-maker and client contact. The provider becomes the operational owner of much of the workflow. The firm still controls the litigation strategy, the scope of preservation, and the final legal judgments.

This model works best when matters are irregular or technically varied and the firm does not have enough consistent work to justify a dedicated support person. Its principal cost is not the technology itself but the coordination required to translate legal objectives into provider instructions, answer technical questions, and review the output. A firm can also inherit problems when responsibilities are divided too sharply: the provider understands the dataset, while the client assumes the production is sound.

A hybrid model keeps basic tasks inside the firm and refers work that requires greater scale or specialized expertise. The firm might manage its own legal holds, collect routine custodial data, and perform an initial review, while sending large-volume processing, complex analytics, or intensive quality control to a provider. This approach can offer more control over ordinary matters without creating a fixed, fully staffed litigation-support function.

The hybrid model breaks down when the dividing line is unclear. If counsel decides which matters are “small” without considering data volume, custodian complexity, document formats, review requirements, or production demands, the firm may discover too late that a nominally routine matter requires capabilities it does not have. Clear intake criteria are more useful than a volume threshold alone.

A full in-house model makes sense when the firm handles enough discovery work to keep the required people and tools occupied. It gives the firm tighter control over workflow, staffing, quality assurance, and client communication. It also exposes the firm to fixed personnel, training, storage, software, and infrastructure costs that continue when matters slow down.

Full in-house capability can become strained when several matters arrive together, each involving different technical requirements. The bottleneck may not be software. It may be the people who understand the load file, coordinate custodians, track production status, or investigate an error before documents leave the firm. A provider relationship can remain useful even in a technically sophisticated firm, particularly for surge support, specialist processing, or independent quality control.

What drives the economics#

Provider-led economics are driven by provider labor, data volume, processing choices, custodians, review scope, and the number of production cycles. Technology-assisted review can affect the shape of those costs, but a low document count does not necessarily mean low labor if the matter has many custodians, complicated metadata, privacy issues, or multiple legal theories.

Fixed costs also sit behind the provider relationship. The firm must identify the right provider, explain the matter’s requirements, establish access to the evidence, specify the work, and review results. Those tasks consume attorney and staff time even when the provider performs the underlying processing. A model that appears to outsource the entire process may simply move costs into instructions, supervision, and quality review.

A hybrid model shifts the cost balance by retaining labor-intensive legal and quality-control work while referring processing-intensive work. It can create duplicated effort if the firm and provider both perform overlapping intake, issue tracking, or status reporting. The handoff should identify who owns each dataset, who approved each processing decision, and who will correct an error.

Full in-house economics are driven primarily by sustained personnel capability. Reviewers, litigation-support staff, technical specialists, and supervising attorneys have different time demands. Software and storage are only part of the cost. Training, matter-specific knowledge, workflow design, and experienced quality assurance add costs that a narrow hourly or device-based view can miss.

The model also changes the risk profile. Referring the entire workflow can reduce operational dependence but increase dependence on the provider’s service quality and responsiveness. Retaining more work increases control but also requires the firm to recognize and escalate problems. The appropriate model is therefore matter-specific rather than a permanent statement about the firm’s identity.

Why smaller firms need a different approach#

Most matters handled by a smaller firm are not naturally suited to an enterprise platform staffed around continuous, high-volume discovery. They may involve a limited number of custodians, a confined issue, and a modest collection. In that setting, the difficulty is often not the sophistication of the search method. It is obtaining complete data from the right systems, preserving it properly, and producing it in a usable format.

Tool selection should follow the work. A smaller firm may need a provider offering managed services, a focused collection application, a familiar review interface, and secure transfer rather than a platform designed for broad, highly customized operations. The relevant questions are whether the tool supports the required custodian sources, preserves metadata, creates intelligible productions, and produces records that can be audited.

The central technical risk for a small matter is also often not the legal defensibility of technology-assisted review. It is the less visible risk of losing information before collection, collecting incomplete data, mishandling attachments, or delivering a production that cannot be mapped back to its source. Those failures affect every review method, whether the search is Boolean, keyword-based, predictive, or based on continuous active learning.

That distinction matters when evaluating a provider or software product. TAR defensibility questions focus on court acceptance, the process for validating results, and the documentation of review decisions. A small firm must first establish a simpler foundation: legal holds that reach the right custodians, a collection that captures the relevant sources, and a production that can be read and checked.

Understand the governing framework#

The electronic-discovery amendments to the Federal Rules of Civil Procedure took effect on 1 December 2015 and govern cases pending on, and commenced after, that date. They place proportionality within the scope of discovery itself under Rule 26(b)(1), including the parties’ relative access to relevant information. That factor is directly relevant when one party controls much of the relevant data and the other party cannot readily obtain it.

The amendments also changed how objections to requests for production are stated. Under Rule 34(b)(2)(B), a responding party must state its grounds with specificity. Under Rule 34(b)(2)(C), an objection must identify whether responsive materials are being withheld on that basis. These provisions favor clear issue definition before the parties begin expensive collection or review.

Rule 26(f) requires the parties to confer before the scheduling conference, and Rule 16(b) governs the scheduling order. Those requirements make early technology planning part of case management rather than a task to revisit after the first production dispute. Scope, custodians, search methods, production formats, and review responsibilities should be addressed during that conference when appropriate.

Privilege and work-product protection also require attention to procedure. Federal Rule of Evidence 502(d) permits a federal court to order that protection is not waived by disclosure in the proceeding before it. Where that mechanism is available, a production can include documents whose protection remains in place subject to the court’s order, with the privilege log identifying the basis for protection.

Technology-assisted review and the responding party#

Technology-assisted review encompasses methods such as predictive coding and continuous active learning. The judicial position is generally favorable to its use in appropriate cases, but the procedural record matters.

In Da Silva Moore v. Publicis Groupe (S.D.N.Y. 2012), Magistrate Judge Andrew Peck held that computer-assisted review was an acceptable way to search for relevant electronically stored information in appropriate cases. The parties had agreed to its use, and District Judge Andrew Carter affirmed in April 2012. The opinion is widely described as the first judicial opinion approving predictive coding.

In Rio Tinto PLC v. Vale S.A. (S.D.N.Y. 2015), Judge Peck stated that courts will permit a producing party to use technology-assisted review, describing the proposition as black-letter law. Maura Grossman was later appointed special master in the case to resolve disputes over TAR. That dispute-resolution role illustrates why method documentation, validation, and clear allocation of responsibilities remain important even when the general use of TAR is well understood.

Hyles v. City of New York (S.D.N.Y., 1 August 2016) presents the other side of the discretion question. Judge Peck held that a court would not force a producing party to use TAR when it preferred keyword searching. The decision aligns with Sedona Principle 6, which places the responding party in the best position to decide how to search for and produce its own electronically stored information.

Some language in Rio Tinto and Hyles is dicta rather than holding. Those portions signal judicial attitude rather than a binding rule, and they do not eliminate matter-specific analysis. The practical lesson is that producing parties retain meaningful discretion over search and review methods, while the record should explain the method, its validation, and any significant implementation choices.

Competence does not require a specialist on staff#

Lawyers are expected to understand the technology relevant to their matters. That duty does not mean every lawyer should design a TAR workflow or administer a processing system. It means the lawyer directing a matter should be able to evaluate the relevant technological assumptions, communicate them to support personnel or a provider, and recognize when specialist input is required.

For a small firm, this competence can be divided between a few people. One lawyer may own legal strategy and privilege decisions. A support employee may administer the hold, collection, and production workflow. A provider may handle processing and review at scale. The division will fail if nobody can connect the legal theory to the search terms, the relevance judgments, the custodian scope, and the documents produced.

Useful questions are concrete. What data is being preserved? Which systems are in scope? How will duplicates and near-duplicates be treated? Which fields will be searchable? Will email threading affect the production count? How will confidentiality and personal information be handled? What will the receiving party receive, and how can the firm trace that production back to the collected data?

A law firm does not need to explain the statistical operation of every tool. It does need enough fluency to ask whether the output answers the legal question. If no one can explain what the relevance ranking represents, how it was tested, or what population was used for validation, the firm has a governance gap rather than merely a software preference.

Capabilities the firm must retain#

Even a firm that refers the entire workflow needs four internal capabilities. Each creates a legal control point that cannot be delegated without instructions.

Issue a defensible hold#

A legal hold should identify the matter’s issues, the custodians and data sources likely to contain relevant information, the preservation steps required, and the person responsible for tracking compliance. It should also be reviewed as facts change. Custodians leave, systems change, and new claims emerge. The hold record should show when notices were sent, acknowledged, supplemented, and released.

Referring collection does not transfer the duty to decide what must be preserved. The provider can execute a scope, but the firm must determine that scope. Failure to preserve electronically stored information triggers a separate inquiry under rewritten Rule 37(e). Its two-tier framework addresses lost information that should have been preserved. Under subsection (e)(1), a finding of prejudice permits measures no greater than necessary to cure the prejudice. Under subsection (e)(2), an intent to deprive another party of the information is required before a court may presume that the information was unfavorable, instruct the jury that it may make that presumption, or dismiss the action or enter default judgment.

Scope the collection#

The collection plan should connect legal issues to custodians, sources, date ranges, and data types. It should also address data held in systems that are easy to overlook, including email archives, messaging platforms, shared drives, local devices, and application data. Where relevant information may be accessible through one party but not the other, proportionality under Rule 26(b)(1) includes their relative access to that information.

Scope should be recorded before collection begins. A later change should be treated as a documented decision rather than an undocumented instruction. This allows the firm to assess the burden and relevance of each addition and to reconstruct what was searched.

Read the production protocol#

A production protocol defines what the producing party will deliver and how. It should address the production format, such as native files, TIFF, or PDF with or without load files, as well as redactions, privilege treatment, confidentiality designations, and document metadata. It should also explain how family relationships, email threading, duplicates, and near-duplicates will be treated.

The protocol is a coordination document, not a formality. Counsel should understand whether the chosen format preserves the information needed for review, whether redactions remain legible, and whether the production can be reconciled to the load file and source collection.

Check the production#

A pre-production check should sample documents across custodians, date ranges, file types, and responsive categories. The reviewer should confirm that text is searchable, images are legible, redactions appear correctly, attachments are present, metadata is readable, and the production totals reconcile with the expected population.

Testing should not be confused with claiming that a product was benchmarked or independently verified. The point is control: the firm checks the actual output before it reaches the other side. If a provider performs the check, the firm should know the sampling method, exceptions recorded, and unresolved issues.

Use the reference model as a map, not a sequence#

The Electronic Discovery Reference Model, or EDRM, provides a shared vocabulary for discussing discovery work. Its stages are information governance, identification, preservation, collection, processing, review, analysis, production, and presentation.

EDRM is a reference model, not a mandatory sequence. Real matters loop between stages. New information may alter preservation, which may require another collection. Production issues may send documents back to processing or review. Analysis may identify a new theory and reopen the source collection.

For a smaller firm, the framework is most useful as a coverage check. It helps the team ask whether information governance has addressed retention and defensible disposal; whether preservation reaches the right sources; whether collection is documented; whether processing decisions are recorded; and whether production quality has been checked. It also helps define a provider’s role without confusing provider labor with firm judgment.

Decide who controls the client relationship#

At the outset, the firm should determine who holds the data, who holds the vendor contract, and what happens at the end of the matter. These are practical control questions, not merely procurement details.

The client should know which party has custody of the collected information, where copies reside, who may access them, and what restrictions apply. The vendor contract should establish responsibilities for confidentiality, security, subcontractors, incident reporting, return or deletion of data, and assistance with defects in a production. Counsel should also understand whether the provider can support the client’s position if a production is challenged.

Exit planning matters as much as acquisition. The firm should know how production files, load files, logs, review records, and correspondence will be returned, transferred, or destroyed; how access ends; and what record the firm must retain for its own file. If the provider also possesses the authoritative data, the client should not assume that ending the engagement makes the material immediately available elsewhere.

The right operating model is the one that preserves informed judgment as matter complexity changes. Referral can be sound without surrendering legal control. A light internal function can be effective if it retains preservation, scoping, protocol, and quality checks. Full in-house capability can justify its cost only when the work justifies the continuing commitment. In each model, the essential test is whether the firm can explain what was preserved, what was searched, what was withheld, and what was produced.