Managed document review services: what you are buying when you outsource review

Managed document review services: what you are buying when you outsource review

Managed review is the labor-intensive center of an outsourced discovery operation. A provider supplies people and a managed process to examine documents, identify responsiveness, flag privilege, apply agreed redaction instructions, and report on the work. Some engagements also include technology-assisted review, or TAR, under reviewer supervision; others use Boolean or keyword search.

The buyer is purchasing more than reviewer hours. The service combines staffing, workflow, quality control, reporting, and institutional knowledge into an accountable production system. Its value lies in sustained throughput without losing control of the review protocol, privilege decisions, or production sign-off.

That distinction matters because an outsourced review can be inexpensive per document yet expensive if errors require extensive re-review. Conversely, a more elaborate review may be justified where the document population is voluminous, the issues are sensitive, or the consequences of an overlooked document extend beyond the review itself. The appropriate comparison is cost, speed, and quality together, not document price alone.

What a managed review provider supplies#

A complete managed review engagement normally includes a review manager, trained reviewers, a workflow for assigning and escalating documents, and a reporting structure. The provider may supply its own platform or configure access to the purchaser’s platform. That choice affects how work is assigned, how reviewer decisions are logged, how quality control is performed, and what the purchaser can audit.

The review manager is the operational point of accountability. That role normally involves translating the review protocol into instructions, monitoring progress, answering questions, managing staffing changes, and reporting exceptions. A reviewer who encounters an ambiguous privilege or responsiveness issue needs a defined path to obtain an answer without stopping the entire review.

Quality control is another core service, not an optional report generated after the work. It can include sampling, calibration, second-pass review, targeted testing, and escalation of inconsistent categories. The provider should identify which activities are part of ordinary quality control and which consume separate project hours or incur additional charges.

Reporting should give the purchaser more than a count of completed documents. Useful reporting can cover production progress, staffing, backlogs, open decisions, quality-control results, and changes in the population under review. It should also permit reconciliation between the reviewed population, the produced population, and any materials withheld or omitted under an objection.

Technology may support that operation, but it does not replace judgment. TAR, predictive coding, and continuous active learning, or CAL, can assist with prioritization or classification when the engagement is designed around them. Boolean or keyword search remains a legitimate method, and de-duplication, email threading, near-duplicate detection, processing, and load-file creation can support the review even when no predictive system is used.

Technology-assisted review and the allocation of discretion#

Judicial decisions have treated TAR as an acceptable method in appropriate cases while recognizing that responding parties often are best positioned to select their own search and review method. In Da Silva Moore v. Publicis Groupe, the court accepted computer-assisted review as an appropriate way to search for relevant ESI, with the parties’ agreement playing an important role. In Hyles v. City of New York, the court declined to force a producing party that preferred keyword searching to use TAR.

Some language in the Rio Tinto and Hyles decisions is dicta rather than holding, so those decisions signal judicial attitudes rather than establish a binding rule. The practical lesson is that a managed-review proposal should identify its method and explain how reviewers validate it, rather than treating “TAR” as a quality claim by itself.

For a purchaser, the relevant questions include whether the technology assists rather than obscures reviewer judgment, what population is used for training or validation, how disagreements are handled, and whether a later change in methodology requires re-review. Method choice belongs within the review protocol and the contractual quality framework.

Staffing determines the actual service#

The provider’s proposal should identify who will perform the work and what qualification process they complete before assignment. The relevant background may vary by project: some documents may require domain knowledge, while others primarily require consistent application of defined criteria. Reviewers also need instruction in the project’s privilege, confidentiality, redaction, and escalation rules.

Experience alone does not establish fitness for a particular engagement. The provider should explain how reviewers are recruited, trained, tested, and corrected. It should also explain whether assignments are matched by subject matter and whether reviewers receive feedback on calibration or production results.

The proposal should also explain how instructions reach reviewers. A complex protocol may require written guidance, decision examples, live training, and periodic clarification. The review manager should maintain a controlled version of those instructions and be able to identify which reviewers worked under each version.

Quality problems often arise from supervision that exists only in name. Contract language should connect staffing levels to named responsibilities, define supervisory review, and require the provider to report departures from the staffing plan. If surge reviewers join the project, the provider should explain how they are trained and calibrated before handling substantive documents.

Changes during the review#

A review protocol is not static. Fact development may change the meaning of a search term, an issue may expand, or the producing party may narrow a category. The contract should define how proposed changes are documented, approved, and implemented.

The key distinction is between clarification and substantive revision. Clarifying an existing instruction may permit immediate application. Changing the substantive review standard may require notice, impact assessment, additional guidance, and review of work already completed. The provider should report the affected population and estimated rework rather than leaving the purchaser to infer the effect from the final count.

Instruction changes should preserve an audit trail. The record should connect each protocol version to the reviewers who received it, the dates it became effective, and any decision or re-review that followed. That trail supports later explanations of why a document received a particular classification.

How managed review is priced#

Managed review is commonly priced by document, by reviewer hour, or through a combination that reflects different phases. A per-document arrangement assigns the provider’s commercial incentive to the size of the reviewed population. A per-hour arrangement assigns it to the labor consumed, including effort caused by unclear instructions, complex issues, rework, or reporting changes.

The unit of price does not determine care by itself. It determines which behavior the contract makes visible. A per-document price may encourage efficiency but can leave the purchaser exposed if a provider avoids difficult documents without accurate coding. An hourly price may preserve flexibility for changing instructions but can make ordinary overhead and rework harder to distinguish from necessary labor.

The statement of work should define the unit of measure before negotiation reaches the price. That definition should address search, processing, de-duplication, email threading, near-duplicate detection, TAR or CAL use, responsiveness review, privilege review, redaction, privilege logging, QC, reporting, and production support. It should also identify which expenses are included, such as data transfer, platform configuration, overtime, surge staffing, travel, export, and corrective review.

Overtime and surge capacity need separate treatment. The proposal should identify the trigger for additional staffing, who authorizes it, and whether it affects deliverables or deadlines. Surge personnel should enter the workflow through the same training and calibration process as the established team. Work performed at premium rates should be reported separately enough to show why the capacity was added.

Quality terms belong in the statement of work#

A quality assurance plan is useful only when its requirements are operational. The purchaser and provider should agree on the sampling regime, which documents enter the quality-control sample, how reviewers are selected, and what constitutes an error. The sample should cover more than obvious responsiveness calls if privilege, redaction, and issue coding are part of the engagement.

Accuracy expectations should be tied to defined review tasks. “High quality” does not provide a usable benchmark. The agreement should instead state what the reviewer is deciding, what evidence is necessary, how ambiguous decisions are escalated, and how the result is measured. Testing can include blinded comparison with an agreed reference set, targeted tests of known documents, and review by a second qualified reviewer.

The measurement method should also fit the task. Agreement between reviewers can reveal inconsistency, but a sample may need additional review to determine whether the inconsistency reflects an unclear instruction, a judgment difference, or a processing error. The protocol should identify who makes the final call when the provider’s reviewers and purchaser disagree.

A failed quality-control sample should have a defined consequence. The statement of work can require broader sampling, additional training, re-review of an identified population, recalibration, or replacement of personnel. The provision should distinguish a documented individual error from a pattern that calls for wider corrective action.

Responsibility for re-review should be explicit. The agreement can state that re-review performed because the provider departed from the agreed protocol is included, while re-review requested because the purchaser changes the protocol is billed separately. Defining that boundary in advance reduces disputes after a failure and connects corrective work to the party that caused it.

The statement of work should also cover quality assurance for privilege logs, redactions, production formats, and any later production correction. Native, TIFF, or PDF output and the presence or absence of load files should be specified as part of production requirements. A reviewer decision can be correct while the produced document is incomplete, missing, or inconsistent with the logged action.

Confidentiality, reviewer location, and data movement#

The review environment should be described as precisely as the data itself. The contract and security materials should identify where reviewers work, whether access is through managed devices, whether documents can be downloaded or printed, and what restrictions apply to local storage, browser use, screenshots, personal accounts, email, chat, and removable media.

Export controls are especially important because a document can leave the managed workflow through ordinary actions. The provider should explain what a reviewer can export, in which formats, and whether exports are technically blocked, logged, or merely prohibited by policy. Reviewers should not be able to route document content through unapproved collaboration tools or consumer services as a practical convenience.

Termination controls should address both access and information. The provider should explain how credentials and tokens are revoked, whether cached content is removed, how in-progress work is returned, what remains in audit logs, and whether any certification follows. A signed confidentiality agreement alone does not answer those operational questions.

Security commitments should also cover subcontractors, off-site support, system administration, and incident response. The provider should identify who may access the data for troubleshooting and how access is recorded. The purchase agreement should make confidentiality obligations applicable to personnel who handle the matter, not merely to the contracting entity.

What the purchaser cannot outsource#

The review protocol remains the purchaser’s responsibility. That protocol should connect the responsive population to the claims and defenses, define search and review criteria, identify privilege categories, explain treatment of near-duplicate and family documents, and state how changes are handled. A provider can apply the protocol and recommend revisions, but it cannot supply the legal judgment that determines what the protocol should require.

Privilege calls require particular attention. The provider can identify candidates and present them in a privilege log, but the governing criteria should come from the purchaser. The team should decide who resolves difficult calls, whether subject-matter specialists are involved, and whether a disputed issue goes to counsel. Privilege review is not reducible to coding consistency.

The same principle applies to redaction and production decisions. Reviewers may apply approved redaction instructions and document the basis for each action, while counsel determines whether a legal conclusion supports the redaction and whether the resulting production is acceptable.

The production sign-off belongs at the end of that accountability chain. Before release, the purchaser should reconcile the review population, produced population, privilege log, redactions, load files, and production specifications. Any Rule 34(b)(2)(B) objection should state its grounds with specificity, and Rule 34(b)(2)(C) should address whether responsive material is being withheld on that basis.

Where clawback is part of the process, Federal Rule of Evidence 502(d) permits a federal court to order that privilege or work-product protection is not waived by disclosure in the proceeding before it. That possibility does not replace a considered decision about which material is produced, reviewed, or logged.

EDRM provides a useful way to examine the engagement because it treats electronic discovery as an operational model rather than a strictly linear sequence. Its stages are information governance, identification, preservation, collection, processing, review, analysis, production, and presentation. Real matters loop between those stages, which is why a managed-review statement of work should describe feedback and revision rather than promise a one-pass workflow.

The most defensible purchase is therefore not simply the lowest-cost reviewer channel. It is the arrangement that makes throughput accountable: qualified reviewers, effective supervision, documented instructions, measured quality, transparent pricing, controlled data access, and a clear record of who made each consequential decision. The provider can carry much of the workload. The purchaser still owns the protocol, the privilege judgment, and the final production decision.