A system for creative testing in social ads

In the era of hand-cut audiences, targeting was the lever. Today automation handles that, and the only control left to a human is the creative. Yet in most firms creatives are made on instinct and judged on instinct.

A test starts with a question

The most common mistake is running two images against each other and afterwards knowing that image A was better. That insight is worthless because it does not transfer.

Instead formulate a question whose answer also applies to the next creative: does a face work better than the product alone; does a price in the image attract or deter; does before-and-after beat the result alone; does a problem pull harder than a promise.

The test matrix

A creative has four layers that should be tested separately, in this order.

Most firms test layer four exclusively and wonder why nothing moves. The gap between two concepts is routinely three times larger than the gap between two typefaces.

Set the hit rate realistically

Expect roughly one in five new creatives to beat the current favourite. That rate sounds low and is normal. It has a practical consequence: producing one new creative a month means improving every five months. Producing five means improving monthly.

The build that keeps things comparable

1. One campaign, one objective, budget at campaign level.
2. One ad set, the same audience for all variants.
3. Three or four ads differing on exactly one layer.
4. Same landing page, same copy, same call to action.
5. Minimum five days running, with no interference.

Point five is the hardest. Switching off the weakest variant after two days prevents exactly the data volume a decision would rest on.

Which number decides

Do not judge creatives on click-through rate. A sensational creative earns clicks and no customers.

The usable order: cost per conversion, the real measure where conversions are plentiful; cost per landing page view, a good substitute at low volume because it filters out mis-clicks; three-second hold rate, only for judging the hook; and click-through rate, only as a warning when it is extremely low.

Spotting and treating burnout

Every creative loses effect once the audience has seen it too often. It shows as rising frequency alongside falling click-through rate.

Once weekly frequency passes three and click-through has dropped by more than a third, the creative is exhausted. A variant with a new hook and the same core is often enough to recover the effect without starting over.

Keep an archive

Maintain a simple sheet with five columns: creative, question tested, result, period, conclusion. After a year it is the most valuable file in the marketing function, because it describes what your audience responds to rather than merely which image once did well.