Experiment Engine
Read results: chance to win, likely range
What the verdict, Chance to beat original and Likely range mean, and when a version is called a likely winner.
The Results tab compares each version with the original on your main goal. A headline gives one of these states.
| Headline | Meaning |
|---|---|
| Waiting for visitors | The test is live and has none yet. |
| Not enough data yet | A version is below the minimum visitors or orders. |
| Too early to call | Results swing a lot in the first week. |
| Keep running | No clear winner yet; an estimate says how many more days. |
| Version B is leaning ahead | Promising but not certain. |
| Version B is a likely winner | A high chance to earn more, with a likely range. |
| Version B is likely worse | Ending the test and keeping the original is recommended. |
| No clear difference | The versions perform about the same. |
| Mixed result | A version earns more but a watch-out metric dropped. |
Chance to beat original
It is the probability that a version really does better than the original, given all the data so far. It never shows 0% or 100%. It does not mean the version is that much better. Select How we calculate this for the explanation.
Likely range
The true change is most likely inside this range. The wider the bar, the less sure we are, and it narrows as visitors arrive. If the range spans zero, the difference may be nothing.