Core Concepts

Richness

Definition

The share of documents in a population that are actually responsive.

Why it matters

Low richness changes sample sizes, expectations, and review cost.

In practice

Richness is the share of a collection that is responsive. It is measured early, from a random sample, and it shapes almost every later decision.

Low richness is common in large collections, and it makes everything harder to measure. With few responsive documents in any sample, estimates of recall and elusion come with wide intervals unless the samples get bigger.

Richness also sets expectations for cost: it tells you roughly how many responsive documents exist to find.

Formula

Richness = responsive documents in the sample ÷ documents in the sample

Worked example

A random sample of 1,500 documents from a 400,000-document collection finds 60 responsive, so richness is 4%. That projects to about 16,000 responsive documents, with a 95% interval of roughly 3% to 5%, or 12,000 to 20,000 documents.

Illustrative numbers, not from a real matter.

Compare with

  • Prevalence: The same quantity. Richness is the usual eDiscovery word; prevalence is the statistics word.
  • Yield: Yield describes what a review step recovered; richness describes the collection before review.

Work it out

← Back to glossary