The most important shift in Meta Ads over the past few years is not the algorithm changes, the iOS privacy updates, or the new campaign types. It is that creative has become the primary lever. The targeting does less differentiation than it used to. The algorithm is good at finding buyers. What separates a high-performing Meta account from a mediocre one is almost always the quality, variety, and freshness of creative, and the rigour of the process used to find what works.
Before ATT, you could build a precise audience, run a few ads, and generate reliable results because the targeting was doing most of the work. After ATT, the targeting signal degraded. Meta's Advantage+ audiences and broad targeting approaches now work better than narrow interest stacks for many advertisers, which means the algorithm is doing more of the audience selection automatically. What you can still control is the creative.
Creative is also how you talk to different segments of your potential audience without changing targeting. A prospecting ad that speaks to a pain point, a retargeting ad that addresses an objection, a social-proof ad that shows customer results, these are all doing different jobs for different audiences, and the way you control that is through what is in the ad, not who you target.
A creative testing framework is not "run a few ads and see what happens." It is a structured process for generating hypotheses, testing them with sufficient budget, reading results correctly, and feeding learnings back into the next round of creative.
The simplest version: define your creative concepts at the messaging level, not just visual treatments, but the core idea or angle each ad represents. Pain-point led. Outcome led. Testimonial. Comparison. Urgency. Each concept gets at least one ad execution, ideally with variation in format (static, video, carousel). Run each concept with enough budget to generate meaningful signal, usually a minimum of 50–100 relevant events before drawing conclusions.
Resist the urge to pause things too early. Creative that looks weak in the first few days often stabilises once the algorithm has found its audience. Give it time, then make decisions.
The highest-leverage variables to test are message and hook, the first three seconds of video or the headline of a static. This is where most of the variance in performance lies. Small changes to colour, font, or button style tend to produce small or inconsistent results.
Test concepts before executions. Running twelve variations of the same underlying message with different thumbnails is unlikely to reveal anything useful. Running three completely different messaging approaches with two executions each is much more informative. Concept testing scales; execution tweaking does not.
Test formats when you have a consistent message to test them against. If you want to know whether video outperforms static for your brand, run the same message in both formats and see what the data says. Do not change the message at the same time, you will not be able to isolate what caused the difference.
Three mistakes I see repeatedly. First: optimising towards CTR when your goal is conversions. High CTR is good, but ads that click well and convert poorly are not useful. Judge creative on cost per meaningful outcome, not on engagement metrics.
Second: declaring a winner too early. Meta's delivery system needs time to find the right audience for each creative. An ad that looks weak at day three can look strong at day ten once the algorithm has had enough impressions to optimise delivery. Use a minimum spend or event threshold before making calls.
Third: confusing creative performance with account performance. A creative might underperform in one ad set structure and outperform in another. Before killing a creative concept entirely, check whether it was given a fair test.
When a creative concept wins, the instinct is to scale budget. The smarter move is to iterate on the concept first: new executions with the same core message, different hooks, different formats. This extends the life of the concept and delays creative fatigue without the algorithm reset that comes with major campaign changes.
Creative fatigue on Meta is real but often misdiagnosed. Frequency alone does not cause fatigue, the same creative shown to the same person many times will lose effectiveness. But high frequency across a large audience is fine. Watch your frequency metrics at the ad set level relative to audience size, not just as an account-wide average.
If you want to build a proper creative testing process for your Meta account, I can help with that.