How to test Meta ad creative properly
Most accounts are not testing. They are running several versions of the same idea and calling the least bad one a winner.
Test three to five genuinely different concepts at once, not variations of one. Give each enough budget to reach roughly fifty conversions or two weeks, whichever comes first, hold audience and placement constant, and judge on cost per outcome rather than click-through rate.
Creative is the biggest lever on Meta and the one most teams treat casually. The platform will optimise delivery within what you give it. It cannot invent an idea you did not supply.
Test concepts, not variations
A different headline on the same image is not a test. It is a tweak, and it will produce a result inside the noise. Put three to five genuinely different propositions in market: a demonstration, a testimonial, a price-led cut, a problem-led cut, something odd.
Hold everything else constant. Same audience, same placements, same objective, same landing page. If two things change, you have learned nothing you can carry forward.
Give it enough to decide
Each concept needs roughly fifty conversions or two weeks before the number means anything. Below that you are reading variance. The common failure is killing a concept on day three because the first hundred impressions went badly, which is how accounts end up with one safe asset and a rising cost per acquisition.
Judge on the right metric
Click-through rate tells you the thumbnail worked. Cost per qualified outcome tells you the idea worked. Those come apart more often than people expect: the asset with the best click-through is frequently the one making a promise the page cannot keep.
What to do with the winner
Scale it, then immediately start building its replacement. On an active account the main asset degrades within four to six weeks: frequency climbs, click-through falls, cost per click holds. That is fatigue, and the fix is new work rather than new audiences. The fingerprints of each cause are in why your ads stopped working.
Tests that waste money
Button colour. Minor copy edits. Audience splits when the creative is the constraint. Anything run for three days. And testing during a sale period, when the offer is doing the work and the creative result will not repeat.
Questions we get asked
How many ad creatives should we test at once?
Three to five genuinely different concepts. More than that and none of them gets enough budget to produce a readable result.
How long should a creative test run?
Until each concept has roughly fifty conversions, or two weeks. Shorter than that and you are making decisions from variance.
Start with a Growth Audit
Every Colleviate engagement begins with an outside-in read of your brand and its competitive cluster, before we recommend a single line of spend.
Ask for a scoped auditWhat we do · How we produce it · Who we do it for · Talk to us