A controlled comparison of two or more variants (creative, landing page, audience, bid strategy) shown to randomly assigned groups, with the difference in outcome measured. Platforms offer built-in experiments (Google Ads experiments, Meta A/B tests) that avoid audience overlap. A test needs a pre-set sample size and a single primary metric to mean anything.
Related terms
Used in The Signal
Cite this entry:
Agency Echelon. "A/B Test." Agency Echelon Reference, reviewed September 2026. https://agencyechelon.com/reference/glossary/a-b-test/