A/B Test
An A/B test splits traffic between two or more variants to measure which one performs better on a defined outcome.
In an A/B test, shoppers are randomly assigned to a variant — commonly a control (A) and one or more alternatives (B, C, and so on) — and their behavior is compared on a specific metric decided before the test starts, such as conversion rate or added revenue.
Random assignment is what makes the comparison valid. If shoppers weren't assigned randomly, any difference between variants could be caused by the kind of shopper each variant happened to attract, rather than by the variant itself.
A holdout group is a special case of an A/B test where one variant is "no intervention at all" — used specifically to measure incremental lift, rather than to compare two different versions of an intervention against each other.
See it in the product
How NavonaAI implements this, in detail.
Related terms
Looking for another term?
Browse the full glossary, or see what NavonaAI adds back to your store.