What to test first in Meta Ads, how to structure the test, how much budget and time to give it, which metrics to track and how to move winners to scaling.

In Meta Ads, the creative has become the most important lever you still have. Targeting is largely automated, Advantage+ decides placements and audience, and the algorithm picks who sees the ad based on what it sees in it. That means the ad itself does a big part of the targeting. If you don't test creatives constantly, you end up depending on two or three ads that wear out within weeks. Below is the simple method I use with the online stores I work with.
The most common mistake is testing small variations: another headline, another colour, another button text. The differences between them are usually small and hard to measure on a small store's budget. The test that changes results is the one between different concepts or angles:
The order I recommend: concept first, then format (video, static, carousel), then the first 3 seconds of the video (the hook), and only then headline and copy. Small variations only make sense once you have a concept that sells.
You have three options, depending on budget:
I wrote about the rest of the account, the scaling and retargeting campaigns, in Meta Ads campaign structure for online stores.
A practical rule: each creative should get a budget of at least 2 to 3 times your target cost per acquisition before you draw a conclusion. If your target CPA is 12 euros, that's 24 to 36 euros per creative. Let the test run for at least 7 days, to include weekends, and don't change anything in the first 3 days.
Hook rate and CTR are early signals: they help you stop a weak creative quickly, but they don't prove one sells. You make the final decision on CPA or ROAS, and for that you need at least 10 to 20 purchases per creative. With 2 or 3 sales, the differences are mostly chance. And you compare ROAS with the threshold calculated from your margin, as I explained in contribution margin vs ROAS.
When a creative passes your threshold, for example CPA below target with at least 15 purchases, move it into your main sales campaign, usually an Advantage+ sales campaign. Use the existing post ID, not a new ad, to keep the likes and comments. Don't stop the test right away: let the winner run in parallel for a few days to see whether it holds up at a higher budget. Some creatives that win on a small budget drop when they get more.
The ad name should tell you what you were testing without opening it. A simple format: date_concept_format_hook_version, for example 2610_back-pain_ugc_question_v1. That way you can quickly group results by concept or format, even months later. If you use the same names in your UTMs, you'll find them in GA4 too. I covered this in the UTM naming convention for online stores.
With the stores I work with, we agree at the start how many new creatives we can realistically produce each week, usually 3 to 5, and we start from different concepts, not variations. I keep a simple sheet with every test, concept, result and what we learned. After two or three months, that sheet is worth more than any single ad, because it tells you what your customers care about.
Creative testing in Meta Ads doesn't need big budgets, it needs order: test different concepts first, give them enough budget and time, judge on real conversions, not just clicks, and move winners to scaling by a clear rule. Repeated every week, this system frees you from depending on a single ad. If you'd like to build the testing routine together, see the Meta Ads page or write to me via contact.