Reading your results
Impressions, conversion rate, and when to trust a winner.
Written by Aleksandar Simov · Updated · 3 min read
Each variant shows visitors, conversions and its rate; the leader is tagged once every variant has 100 visitors. Free reads "Collecting data" until then and "No clear winner yet" after. On every plan a broken traffic split comes first, as a warning. Pro adds significance, probability to beat and an uplift interval, and calls winners only past its thresholds.
1. What each variant shows
Each variant shows visitors, conversions and conversion rate, as cards or as a table that adds each version's change against the original, and the view you pick is remembered. While a variant is under the minimum (100 visitors by default) it shows its progress, a rate marked "so far" and a Collecting tag, and nobody is called Leading; past it, the best rate is tagged Leading, which is never called a winner. In Free the verdict reads "Collecting data" until the weakest variant reaches the minimum and "No clear winner yet" after that; nothing on Free can be read as significance. A lead is not a verdict: early numbers swing wildly, which is why the planner asks you to set a sample size before you start.
2. When to decide
Run until you hit the planned sample, in whole weeks so weekday and weekend visitors are both represented, then decide. Pro computes statistical significance (a two-sided two-proportion z-test), Bayesian probability-to-beat and an uplift confidence interval. On every plan a traffic split that is off comes first: the top of the page becomes a warning with the planned and the counted split, the likely causes and one fix, and nobody is called Leading until the split is even again. Pro calls a winner only when both variants have at least 100 visitors, there are 25 conversions between them, the confidence is at or above your threshold (95% by default) and no traffic split warning is active.
3. Declaring a winner
Declaring a winner is always your choice: pick the variant in the Winner list, and Declare winner says the test stops. One click then keeps showing it to every visitor, with no further tracking, until you have edited the real page. The day-by-day chart reads with the pointer or the arrow keys, or as a table. If the result is inconclusive at your planned sample, that is an answer too: keep the simpler version and test something bolder.
Want the longer read? See how long to run a test on the blog, or browse all docs.
Common questions
When does Pro declare a winner?
When a variant beats the original, both have at least 100 visitors, there are 25 conversions between them, confidence is at or above your threshold (95% by default, raised automatically when a test has more than two variants) and no traffic split warning is active.
What does "No clear winner yet" mean in Free?
Free does not compute significance. It shows the leader and the counts; the verdict text only tells you that nothing on Free can be read as significance.
What happens when I declare a winner?
One click keeps showing the winner to every visitor, with no further tracking, until you have edited the real page.
What do visitors see while a test is paused or stopped?
The original. While an element test is a draft, paused, or stopped without a kept winner, visitors see the original version and the other versions stay hidden.
Was this page helpful? Yes · No
Edit or report a problem