GA4Check the test first
- Did every variant get a fair share of spend and time, or did one run for two days?
- Did the variants face the same audience and the same placement?
- Is the result measured on the number agreed before launch, with the window stated?
Find the likely cause
- Hook: the first seconds or the first line did not earn the next look.
- Offer or price: the promise did not match what the buyer wanted.
- Landing page: the ad made a promise the page did not keep.
- Audience: the message was right but spoken to the wrong people.
Decide and record
- Keep it running if the test is not finished. Do not cut a test on day two because the numbers look bad so far.
- Change one element and test it again, if the cause is clear.
- Pause it when the test is fair, the loss is clear, and the cause is known or not worth finding. Pausing is a change, so it goes through the approval step and into the log.
Made-up scenario: a made-up apparel store pauses an ad after five days with a cost per purchase far above target. The review finds that the variant got a third of the spend and the ad sent people to the wrong page. The verdict is 'not a fair test, retest with the right page', and that goes into the log.
