Setting Experiment Goals from Analytics Events

By default an experiment is judged on your website’s configured conversion event. Goals let an individual experiment measure something else, or several things at once — useful when the change you are testing affects a step partway down the funnel rather than the final conversion.

You set them from the Tracking card on the experiment’s Settings tab. Drag Event Tag Tracking into the target area to open the goal picker.

This requires a connected analytics property

The card only appears when the website has a Google Analytics property connected, because the available goals are that property’s event tags. If you do not see it, connect the website to GA4 first.

Conversion tags and other tags

The picker splits the available events in two:

• Conversion Tags — events your analytics property marks as conversions. These are pre-selected the first time you open the picker, which is the sensible default for most experiments.
• Other Tags — every other event on the property, available when you want to measure something the property does not treat as a conversion.

Tick the events this experiment should measure and save. At least one goal is required — an experiment measuring nothing has no result to report.

Choosing goals well

Pick the metric the change plausibly affects. A test on a product page that only measures completed purchases has a long, noisy chain between the change and the metric. Measuring add-to-cart as well gives you a nearer signal that will reach significance sooner and tell you whether the change did what you designed it to do.

Decide before you start. Choosing a goal after seeing the results — picking whichever metric happens to look good — turns a null result into a false positive. If you want several metrics, name them all in advance.

Know which one decides. With multiple goals, agree beforehand which is the primary one. Otherwise "it won on one of the four" becomes the standard for shipping.

Reading multiple goals

On the Statistics page, each goal is available as an objective, and switching between them recomputes the whole page. Experiments frequently win on one goal and not another; that pattern is informative rather than contradictory, but it is not licence to report whichever is friendliest.

Notes

• Goals come from the connected analytics property. To add a new one, add the event there first.
• Changing an experiment’s goals mid-run changes what is being reported, not what was collected.
• Without goals, the experiment falls back to the website’s configured conversion event.