Glossary · Business Analytics

Peeking Problem

also: optional stopping · sequential testing

Definition

The peeking problem is the inflation of false-positive rates that occurs when a frequentist A/B test is repeatedly evaluated before reaching its pre-registered sample size. A nominal 5% false-positive rate can become 20–30% under daily peeking. Bayesian testing and sequential-analysis methods eliminate the problem.

Frequentist null-hypothesis testing guarantees its error rates only at the pre-specified sample size. Each interim look at the data re-rolls the dice: with 10 looks at alpha = 0.05, the probability of at least one false positive rises above 20%. Bonferroni correction or alpha-spending functions solve the problem at a power cost. Bayesian A/B testing sidesteps the issue entirely — posterior probabilities remain calibrated regardless of how many times the test is inspected.

Essays on this concept