A/B testing
Splitting the traffic of one short link between destination variants in order to compare them on a metric such as conversion.
Updated August 14, 2026
A/B testing on a short link means giving one address two or more destinations, assigning each a share of traffic, and judging them by what happens afterwards. The link platform handles the split; the comparison is yours to make on a metric that reflects the goal — a conversion, a signup, revenue — because clicks measure the link, not the page behind it.
Two mechanics decide whether the numbers can be trusted. The assignment must be stable, so a visitor who returns tomorrow sees the same variant rather than a different page, and the variant served must be recorded with the click so later events can be attributed back to it. Weights must also add up to the whole of the traffic, otherwise part of it goes somewhere unmeasured.
The usual errors are impatience and drift. Small samples wander, so an early lead is often noise; decide in advance how much traffic the test gets. Changing the weights mid-test mixes two different populations in one report. And a variant that differs in a dozen ways tells you that something worked, not what.