What recall and precision rates does a court expect from a TAR review to be considered defensible?

There is no fixed recall or precision rate that courts treat as a magic number for defensibility. Technology-assisted review (TAR) is evaluated by the reasonableness of the overall process, not by hitting a single statistical threshold. Understanding what these two metrics measure helps explain why.

What recall and precision measure

• Recall is the share of all truly relevant documents that the review actually found. High recall means few responsive documents were missed.
• Precision is the share of documents the process flagged as relevant that actually are relevant. High precision means less time wasted reviewing non-responsive material.

In practice, recall is the metric courts and producing parties care about most, because it speaks directly to whether a production is reasonably complete. Precision relates more to efficiency and cost.

What courts actually expect

US federal courts generally apply a standard of reasonableness and proportionality rather than perfection. Discovery rules require parties to take reasonable steps to locate and produce responsive material in a manner proportional to the needs of the case; they do not mandate a perfect or exhaustive search. Many practitioners and commentators cite recall in the 70-80 percent or higher range as commonly defensible, but this is a practical benchmark, not a legal rule. The right target depends on the case, the data volume, the issues at stake, and what the parties agree to.

Defensibility typically rests on:

• A documented, repeatable workflow with consistent relevance criteria.
• Validation of results, often through statistically sound sampling that estimates recall with a stated confidence level and margin of error.
• Transparency and cooperation, including discussing methodology and validation with the opposing party where appropriate.

Practical takeaways

Treat recall and precision as evidence that your process was reasonable, not as scores to chase in isolation. Agree on validation methods early, measure against a defensible sample, and keep records of your decisions. Standards differ by jurisdiction, so confirm the expectations of state courts and non-US forums before relying on any benchmark.

For related concepts, see e-discovery.