Tracking & results
Split testing & reading results
Run A/B tests between variations of a page, choose what to measure, and read results without fooling yourself.
A split test sends visitors to two or more variations of the same page URL so you can see which one performs better. Everything happens on one address, so your ads and links don’t change.
Setting up a test
- Open the page and add a variation. Start from a copy of the current one, then change one thing that matters: the headline, the angle, the opening story, the offer presentation.
- Set the traffic weights — 50/50 for two variations is the usual start.
- Choose the metric that decides the test: click-through rate (visitors who clicked an offer link), lead rate or sale rate. Pick the one closest to money that you can measure; sale rate needs postbacks.
- Publish. Each active variation goes live at the same URL and visitors are divided between them according to the weights.
You can pause a test at any time. Pausing stops the split; the page keeps serving.
Reading the results
For each variation you’ll see unique visitors, outbound clicks and click-through rate, leads, sales and payout. Results only include activity on the published page — previews and the editor are excluded.
At the top, the results give you one of three plain verdicts:
| Verdict | What it means |
|---|---|
| Not enough data yet | Every variation needs at least 100 unique visitors, and the test needs at least 10 clicks, leads or sales in total (depending on the metric). We also estimate how many more visitors per variation a reliable read needs. |
| No clear winner yet | One variation is ahead, but the difference could still be chance. We show how far ahead it is and roughly how many visitors it would take to detect a 20% difference. |
| [Variation] is the likely winner | Shown only when that variation has at least a 95% probability of being best and the difference is statistically significant (p < 0.05 in a two-proportion test). |
Small differences on small samples are usually noise, so “ahead” is not the same as “winning”. The verdict is the most valuable part of the screen: it stops you shipping a lucky streak.
Practical rules
- Change one big thing. Testing a new angle teaches you more than testing a button color.
- Run full weeks. Traffic quality changes by day of week; stopping on a Tuesday can mislead you.
- Don’t peek and stop. Decide roughly how long you’ll run before you start, and don’t end the test the first time one variation looks better.
- Keep everything else equal. Same traffic source, same targeting, same offer link — or the test measures the wrong thing.
- Watch the metric that pays. A variation with more clicks but fewer sales is not the winner.
Choosing a winner
When you’re confident, choose the winning variation. All traffic goes to it and the test ends. The other variations, and the results, are kept, so you can look back at what you learned and start the next test from the winner.