Metrics

Yield

Definition

The number or proportion of responsive documents recovered by a step in the workflow.

Why it matters

It helps you judge whether a culling or prioritization move actually captured the responsive material.

In practice

Yield is how many responsive documents a step produced: a review batch, a search term, or a culling decision. It can be a count or a rate.

Comparing a batch yield with the collection richness shows whether prioritization is working. If the collection is 4% responsive but the first review batches are 30% responsive, the model is surfacing the right documents.

In a continuous review, falling batch yield over time is expected and is one of the signals teams watch when deciding whether to stop.

Worked example

Reviewers code a batch of 5,000 top-ranked documents and find 900 responsive, a batch yield of 18%, against an estimated collection richness of 4%.

Illustrative numbers, not from a real matter.

Compare with

  • Recall: Yield counts what a step found. Recall compares everything found with everything that exists.
  • Richness: Richness is a property of the collection; yield is a property of a step in the workflow.

Work it out

← Back to glossary