Precision

Of the results returned, what fraction are relevant — the “quality, not quantity” IR metric, complementary to recall.

Precision = (relevant retrieved) / (total retrieved). High precision means almost everything you returned is on-topic. Low precision means lots of noise. For ecommerce, precision matters most on the first page — shoppers don’t scroll past noise to find the good stuff.

Precision@K (truncated to the top K results) is the practical version: P@10 measures how many of your first ten results are relevant. P@10 = 0.7 means seven of ten are good, three are noise. That’s a typical bar for a healthy ecommerce search experience.

Precision and recall trade off. A strict ranker that only returns clear matches has high precision but may miss relevant items (low recall). A loose ranker returns more items, recovering the misses, but introduces noise (lower precision). NDCG and F1-score combine both into a single number.

Related terms

Recall

Of all the relevant items in the catalog, what fraction did the search return — the coverage metric, complementary to precision.

NDCG

Normalized Discounted Cumulative Gain — a ranking quality metric that rewards relevant results highly placed in the result list and discounts those buried below the fold.

Mean Reciprocal Rank (MRR)

A ranking metric that averages 1/(position of first relevant result) across queries — rewarding rankers that put the right answer first, indifferent to deeper ranking quality.

Browse all glossary terms