A/B test

An A/B test (or split test) randomly serves two versions of a creative to equivalent audience segments and compares performance. The variable being tested could be the hook, the image, the offer, the CTA, or any single element. The methodology requires a clean isolated variable. Testing ‘everything different’ produces a verdict but no learning.

Classic A/B tests need statistical significance to be useful, typically 1,000+ conversions per cell over the test window. At lower volumes the result is only suggestive. Most performance teams skip statistical rigor and treat A/B as ‘ship the obvious winner after 7 days’ pragmatism. That's not formally correct but works at the testing volumes paid social demands.

A/B tests are best for testing significant changes (new hook angle, new offer, new format) rather than micro-tweaks (different shade of green button). The cost of running the test eats into otherwise-productive spend; reserve it for high-stakes variables.