Most brands find winners by accident.
An ad takes off. They scale it until it dies. Then they go back to guessing, usually by making "more creatives". It doesn't work, because nobody can say afterwards which change moved the number.
The WINMAP Method is how we break that loop. Every ad is a test of a hypothesis about the customer, written down before it launches. The goal isn't more ads. It's signal: knowing why something won, so you can do it again on purpose.
Volume without a system isn't testing.
Why "more creatives" isn't the fix
Meta now reads the ad itself to decide who sees it. The creative is effectively the targeting.
That changes what volume means. Ads that are near-copies of each other get grouped together and shown to the same people. Ten hook swaps on one video can behave like a single ad. What reaches new customers is diversity: different personas, different angles, different formats.
So the job isn't to make as many ads as possible. It's to make enough good, genuinely different ads that the winners show themselves, and to know why each one won.
What WINMAP stands for
WINMAP stands for Winning Insights, Messaging and Performance. The name gives the order of operations:
- Insights come first: why the customer buys, what stops them, the words they use.
- Messaging is built from those insights.
- Performance is how you find out whether you were right.
Most accounts run it backwards. They start with a format ("let's test UGC"), make the ads, and look at performance with no idea what they were testing.
The six stages
| Stage | What happens | What it produces |
|---|---|---|
| Research | Customer research: reviews, Reddit, comments, sales calls, post-purchase surveys | A customer psychology brief: triggers, objections, personas, awareness levels, customer language |
| Translate | Each insight becomes a written hypothesis, then a concept, then a brief. Format is chosen last | Briefs covering who it's for, awareness stage, concept, hook, message, format and the action |
| Produce | Approved briefs go into production: filmed, edited or designed | Finished ads, tracked through to launch |
| Test | New concepts run separately from proven winners, one question per test | A result read against the hypothesis written before launch |
| Iterate | Proven winners get variations: a new hook, a shorter cut, a new format, a new creator | More life from what works, and knowledge of which part did the work |
| Optimise and report | Regular reviews with the client and their media buyer, plus a monthly creative insights report | Learnings that become the next hypotheses |
Research is where most of the value is. We look for why people actually buy, what stops them, and the exact words they use, and we look at what's working in the category and why, without copying it.
Translate is where an insight becomes testable. Every concept starts as a sentence: "we think persona X will act on angle Y because Z." If you can't write that sentence, you're not ready to brief the ad.
Test keeps new ideas away from proven winners, so they actually get spend. A new concept dropped into a campaign full of winners rarely gets a fair read.
Optimise and report closes the loop. What we learn feeds straight back into research, so every round of testing starts smarter than the last.
The test loop inside each stage
The six stages are the full cycle. Every individual test runs a smaller loop inside it.
Both are right at different zoom levels. The cycle is how the account runs month to month. The loop is how each test earns its place in it.
What a test unit is
A test unit is one asset built to answer one question. Reskins don't count.
Assets come in three layers:
- Concepts are new hypotheses: a new persona, angle or offer.
- Variations probe one element of a concept, such as the hook, the CTA or the text overlay.
- Iterations refine a proven winner.
A concept gets several executions before it's judged, because one ad is a data point, not a verdict. Volume scales with spend, but only above a quality bar.
Test wide, then narrow
We change one variable at a time. But that rule only makes sense at the right level, so we test in two stages.
First, find which message lands.
Then learn what's doing the work, and extend its life.
- Wide first. Find which message lands by testing clearly different concepts against each other: different personas, different angles.
- Narrow second. Once a concept wins, change one variable at a time, starting with the hook, to learn which part is doing the work.
"Let's test UGC" is a format test pretending to be a concept test. Decide the angle first. Format comes last.
How we read a result
- Judge on the number the founder cares about. CPA and new customer acquisition cost first, meaning what you pay to win a first-time buyer. Hook rate and hold rate diagnose problems, but they don't decide the result. A great hook rate with a bad CPA is still a bad ad.
- Read at the ad set or campaign level, not the single ad. Meta credits the sale to the last ad clicked, so ads that warmed the customer up earlier look weak on their own. Switch those off and the "winners" often stop working too.
- Where Meta chooses to put spend is itself a signal.
- Judge every test against the hypothesis written before launch. If you didn't write down why you tested it, you can't learn from the result.
What the tracker records
Every test goes into a tracker. For each ad it records:
- The hypothesis, written before launch
- Concept, persona, angle and offer
- Awareness stage
- Format and hook
- Launch date
- Spend and CPA
- Win or loss
- Learnings: why it won or lost, broken down by hook, body, angle and framing
- Next step: scale, iterate, change format, or kill
After a few months, the tracker becomes a map of everything you've tested and where the gaps are. You brief into the gaps instead of going on gut feel.
Iterations are earned
Iterations only happen on proven winners. Never on losing creative, and only once the first round of tests has produced a winner, usually from month two.
The types we use most:
- A new hook on a winning body
- A shorter cut
- The same concept in a new format, such as video to static
- A different creator on the same angle
Most iterations won't beat the original. The ones that don't still tell you which part of the ad was doing the work, and iterating extends the life of a winner that's starting to fatigue. New hypotheses stay the majority of what we make. Iterations grow as winners appear.
Creative testing mistakes we see
- Testing everything and hoping something sticks.
- Copying a winner without knowing why it worked.
- Believing more creatives means better performance.
- Twenty near-identical variants of one ad, mistaking volume for diversity.
- Choosing the format first instead of the concept.
- No written hypothesis, so there's nothing to learn from the result.
- New tests in the same campaign as proven winners, so the new ads never get spend.
- Judging single ads on last-click ROAS and switching off the ads that were warming people up.
- Using hook rate as the main success metric.
- Taking every idea from ad libraries, so every brand in the category ends up looking the same.
- Creative teams that never look at performance data, making ads that look good but don't convert.
- One winner carrying all the spend with nothing else being tested behind it.
When to bring in a creative testing partner
It's usually time when:
- The product is proven, but creative can't keep up as spend scales.
- Your media buyer keeps asking for more creatives.
- The account has been flat for months.
- You know you need creative testing but don't know how to set it up.
- Peak season is coming.
Without a system, the account becomes a cycle of fixes. Winners arrive by accident and can't be repeated, and one ad carries everything until it fatigues.
It's too early if you haven't found product-market fit yet, or if there's no media buyer running the account. Creative testing works alongside media buying, not instead of it.
How we run WINMAP for clients
WINMAP is how every 225 Media account runs, from the first customer research to the monthly insights report. We do the research, write the hypotheses, produce the ads across video, statics and UGC, and run the tests with your media buyer. See our creative strategy and testing service.
Want to know what your account is actually testing?
Book a Creative Diagnostic. It's a live teardown of your ad creative on the call: which tests are telling you something, which are just volume, and what we'd test next.
Book a Free Creative DiagnosticA live teardown of your ad creative, on the call. No obligation.FAQs
What is a creative testing framework?
A creative testing framework is a repeatable process for deciding what ads to make, testing them so you know why one won, and feeding what you learn into the next round. Ours is the WINMAP Method.
What does WINMAP stand for?
WINMAP stands for Winning Insights, Messaging and Performance. Insights come first, messaging is built from them, and performance tells you whether you were right.
How many variables should you test at once?
One, within a winning concept. Test clearly different concepts against each other first to find which message lands, then change one variable at a time, starting with the hook.
Is hook rate a good way to pick winning ads?
No. Hook rate and hold rate diagnose problems, but the result should be judged on CPA and new customer acquisition cost, read at ad set or campaign level.
When should a brand use a creative testing agency?
When the product is proven but creative can't keep up with spend, your media buyer keeps asking for more ads, or the account has been flat for months. Before product-market fit, or without a media buyer, it's too early.