Definition
Split testing, usually called A/B testing, is comparing two or more variants of something — a creative, a landing page, a headline, a call to action, an email subject, a funnel step — by showing each variant to a randomly assigned portion of traffic and measuring which produces the better result on a defined metric. Randomisation is what makes it valid: because visitors are assigned to variants by chance, differences in outcome can be attributed to the variant rather than to differences in who saw it.
A sound test changes one meaningful thing at a time, runs until it has enough conversions per variant to distinguish a real difference from noise, and is judged on the downstream metric that matters — cost per qualified action, not just click-through. Calling a winner early, on too few conversions, is the most common mistake and repeatedly locks in variants that were merely ahead by chance.
In context
In iGaming media buying, split testing is the core loop: test creative angles against each other, test prelanders, test landing-page layouts and form lengths, test CTA wording. The discipline is to hold everything else constant, give each variant a pre-decided impression or conversion budget, and only then compare.
Because iGaming conversions (qualified FTDs) are relatively rare per click, tests on downstream metrics need substantial traffic to reach confidence, so buyers often test upstream proxies (CTR, click-to-registration) first and reserve full FTD tests for the shortlisted variants.
The common pitfalls are worth naming: testing several changes at once so a winner cannot be attributed to any one of them; stopping the moment a variant looks ahead (early results swing wildly); ignoring that a creative winner today fatigues in two weeks; and optimising a shallow metric (a page that wins on registrations may lose on deposits). Well-run testing also documents results, so the team builds a library of what has worked by GEO and offer rather than re-testing the same ideas.
Worked example
A buyer A/B tests two casino landing pages: one with a six-field registration form, one with three fields and progressive profiling. Over 2,000 registrations per variant, the three-field page converts clicks to registrations at 19% versus 14%, but its deposit-to-registration rate is 4 points lower, so on cost per qualified FTD the two are within noise.
The buyer keeps the shorter form for its lower support load and moves on.
Related terms
Frequently asked questions
Browse the full iGaming & affiliate glossary — hundreds of EN/RU terms with examples.
← Back to glossary