A creative testing system that does not depend on luck
Kill on retention, not on cost per lead. Here is the weekly rhythm we use across every performance account.
Creative decides most of your performance on Meta and a growing share on Google. Yet most accounts test creative by feeling.
Judge early on retention
Cost per lead takes weeks to stabilise. Video retention stabilises in days. So at day eight we look at the 25 percent video retention rate. Anything below 10 percent gets turned off regardless of what its cost per lead says, because the sample is too small to trust and the attention data is not.
Separate the test budget
Keep a fixed testing budget in its own ad set. New creative goes there, proven creative goes into the scaling ad set. Mixing the two means your winners subsidise your experiments and you never see clean numbers for either.
Vary one thing at a time
Hook, format, angle. Change one per variant. Six creatives that differ in every possible way tell you nothing about why one won.
Retire on frequency, not on boredom
You will get tired of a creative long before your audience does. Watch frequency and the trend in cost per result. When frequency passes about 2.5 in a small geography and cost starts climbing for two weeks straight, it is time.
Feed the pipeline
The agencies that win here are not smarter, they simply produce more. Twelve decent creatives a month beats two brilliant ones, because the auction rewards volume of attempts.