Straightforward Bayesian A/B testing with Dirichlet posteriors

Dustin Hayden, Tom Armitage · 2025

Read Straightforward Bayesian A/B testing with Dirichlet posteriors by Dustin Hayden, Tom Armitage (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2508-08077.txt.

Description

Dustin Hayden and Thomas Armitage offer a simple Bayesian way to run A/B tests on web platforms. A Bayesian test compares two or more test groups through their joint posterior distribution. Building that distribution by hand for each test needs expert time and does not scale when firms run many tests each month.

Their fix approximates the joint posterior with a Dirichlet and categorical model that works across many test types. No expert needs to pick priors and likelihoods for each new test. In simulations and in real platform use the approximation works well enough to serve as the standard method. The trade off is clear: a hand tuned model can be better, but the shared method is fast and robust.

Catalog record

CREATORDustin Hayden, Tom Armitage
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE62 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2508-08077

Browse the full library · About the Tower of Babel