Reading A/B test results
Results live in the Analytics app → A/B Tests tab. Open it from the platform with View in Analytics on a test card (the test comes preselected), or open Analytics directly and pick the test from the dropdown.
Use the period control and Refresh on that page to update the report. The A/B Tests tab does not use the shared page/referrer filter bar used by other Analytics tabs.
Summary cards
| Card | What it tells you |
|---|---|
| Visitors | Unique visitors assigned to variants. |
| Conversions | Unique visitors who completed the primary goal at least once (so conversion rate never exceeds 100%). |
| Conversion rate | Unique converters ÷ unique visitors. Shows N/A for Traffic split only tests. |
| Checkout clicks | Shown only when the primary goal is Checkout click. |
| Best variant | Declared winner when one exists; otherwise the variant with the highest conversion rate (including control). |
| Traffic health (SRM) | Whether the real traffic split matches your configured weights. |
Variant details table
Each variant row shows visitors, conversions, conversion rate, lift (relative change vs control), and a decision badge.
Lift = (variant CR − control CR) / control CR. Control always shows 0% lift.
Decision badges (statistical)
While the test is collecting enough traffic for a statistical read:
- Collecting data — not enough visitors yet (under the significance threshold). Keep waiting.
- Leading — this variant is likely better than control.
- Losing — this variant is likely worse than control.
- Inconclusive — enough data, but no clear difference.
After you declare a winner
When a winner was declared in the platform:
- The winning variant shows Winner.
- Other variants show Not selected.
- Best variant and the status area also show the declared winner (for example Winner: Control).
You can still open archived tests and change the date range to review historical numbers.
Traffic health (SRM)
SRM stands for sample ratio mismatch. If you configured a 50/50 split but traffic arrives 70/30, something is off — usually stale routing or a tracking problem. When SRM shows a warning, don't trust the lift numbers; redeploy the test and check that the entry URL serves variants correctly.
Charts
- Variant performance — visitors and goal completions by variant for the selected period.
- Traffic trend — daily unique visitors per variant (one line each).
- Conversion trend — daily conversions per variant (hidden for traffic-split-only tests).
When to make a decision
- Wait until decision badges leave Collecting data when you want a statistical call.
- Make sure SRM is healthy.
- Prefer results that cover at least one full week (weekday and weekend behavior differ).
- Then declare a winner or archive the test without a winner.
You can still declare a winner earlier if you accept the risk — Atomicat will record and display that choice even if statistical badges were still collecting data.
Frequently asked questions
How long should a test run?
Until the decision badge is leading or losing with healthy SRM — typically one to two weeks for pages with steady traffic. Ending tests on day one is the most common A/B testing mistake.
Why is Best variant Control when Variant B has negative lift?
Best variant uses the highest conversion rate, including control — or the declared winner if you already picked one. Negative lift on a challenger means control is converting better.
Why do visitors not match my ad platform clicks?
Ad platforms count clicks; Atomicat counts unique visitors who actually loaded the page. Bots, blocked scripts, and bounced connections cause differences.
Can I see results after the test is archived?
Yes. Pick the archived test in the Analytics dropdown and select a date range that covers the time it was running. Declared winners still show as Winner.