Goals & reading results

Goal types

A goal defines what counts as a conversion. Add one or more:

• URL visit — visitor reaches a URL (exact or contains). E.g. /thank-you.
• Element click — visitor clicks an element matching a CSS selector. E.g. .buy-button.
• Form submit — a form matching a selector is submitted.
• Custom event — a previously defined custom event fires.

Pick the goal closest to the money. A click on "Add to cart" is easy to measure and easy to win without selling anything more; a purchase custom event is harder to move and tells you the truth.

The results table

The Variation Comparison table compares each variation against control:

• Impressions — how many times the variation was served.
• Unique Visitors — distinct visitors who saw it.
• Conversions and Conv. Rate.
• Improvement — percentage change in conversion rate versus control.

Conversions Over Time charts the trend per variation, so you can see whether a lead was built early and then held, or is still moving.

Filters

Change the Date Range to look at a subset of the run, and apply a segment to read the same test for one audience — mobile only, paid traffic only. A variation that loses overall but wins decisively on mobile is a common and useful result.

Improvement is not the answer on its own

A +40% improvement on 30 visitors means nothing; a +3% improvement on 40,000 visitors can be worth shipping. The Improvement column tells you the size of the difference, not whether it is real. For that, read the statistics panel — it is the part that decides whether you have an answer.

Reaching a valid conclusion

• Let the test run long enough to gather sufficient data — don't call a winner on day one.
• Wait for the WINNER indicator rather than reacting to early swings. Early results move violently and almost always regress.
• Cover full business cycles (e.g. include weekends) to avoid day-of-week bias.
• Decide before launching what you'll do if the result is flat — "no difference" is a legitimate and common outcome.