Creative testing
Creative testing is the practice of running several deliberately different ad creatives against a known baseline in order to learn which creative elements move performance. It is distinct from A/B testing a landing page in one important way: the goal is usually to find a new winner quickly rather than to reach statistical certainty about a small effect.
Last updated
Why it matters
Creative is the highest-leverage variable left in most paid social accounts. Bidding, targeting and placement are increasingly automated, which leaves the creative and the offer as the parts a team still controls.
It is also perishable. A winner found this month will fatigue, so testing is a continuous supply problem rather than a one-off exercise.
How to apply it
- Establish the baseline: your current best-performing ad, still running.
- Ship a set of 8 to 20 variations into the same ad set as the baseline.
- Read hook rate at 48 to 72 hours and kill anything below half the set median.
- Read completion and cost per result at 5 to 7 days on the survivors.
- Promote the top two, retire the rest, and refill the set.
Resist the urge to split each variation into its own ad set. That divides your budget by the number of variations and delays every read, and it prevents the platform's optimiser from doing the allocation it is good at.
What people get wrong
- Chasing statistical significance on every variation. Most creative differences are large or irrelevant, and the borderline cases are rarely worth the spend to resolve.
- Testing without a baseline, which leaves you comparing losers to each other.
- Changing several elements at once, then attributing the result to one of them.
- Stopping the programme when a winner is found.
Frequently asked questions
How much budget does a creative test need?
Enough for each variation to accumulate a few thousand impressions before the first read. That is usually far less than teams assume, because the first cut is made on hook rate rather than on conversions.
Should each creative get its own ad set?
Usually not. One ad set containing the whole creative set lets the platform allocate, which is faster and cheaper than manually splitting the budget.
How long should a creative test run?
First cut at 48 to 72 hours on hook rate, decision on cost per result at 5 to 7 days or one conversion cycle, whichever is longer.