Most ad accounts we audit have plenty of tests in them. What they rarely have is a record of what those tests taught anyone. Ads were launched, some did better than others, the winners got more budget and the losers were paused. Three months later, nobody can say why the winners won.
That’s not testing. It’s just spending money in several places at once. A framework fixes it, and it doesn’t need to be complicated.
Separate the layers
An ad is really a stack of decisions. It helps to test them in order, from the biggest to the smallest.
The concept is the core idea or angle. Saving money, saving time, avoiding a risk, joining people like you. Different concepts can perform very differently, and this is where the biggest gains usually are.
The hook is the opening: the first line, the first two seconds, the image that stops the scroll. The same concept can live or die by its hook.
The format is how it’s delivered. Talking head, product demo, static image, carousel, testimonial style.
The details are everything else. Colours, button text, captions, music.
If you test details before you’ve found a strong concept, you’re polishing something that might not be worth polishing.
Write a hypothesis for every test
Before anything launches, write one sentence saying what you expect and why. “We think buyers worry more about installation hassle than price, so an ad leading on a one-day install will beat our usual price-led ad.”
It feels slightly formal. It’s also what turns a result into a lesson. When the ad wins, you’ve learned something about your buyers. When it loses, you’ve still learned something, and you won’t run the same idea again next quarter by accident.
How many variants?
Fewer than people think, but more different from each other than people make them.
For a concept test, three to five genuinely distinct concepts is usually plenty. Each one might have two or three hooks. Forty versions of a single idea with tiny changes will mostly tell you about the noise in the data.
Generative tools have made it easy to produce huge numbers of variants. That’s useful for hook testing and for keeping creative fresh. It doesn’t change the basic rule that a test is only as good as the difference between the things being tested.
Give it enough time and enough money
Calling a winner after one day is one of the most common mistakes we see. Early results swing wildly. An ad that’s ahead on Monday is often behind by Thursday.
As a rough guide, let each variant collect a meaningful number of conversions before deciding. For lead generation, that might mean a few dozen leads each, not five. If your budget can’t support that across many variants, test fewer variants.
Judge on the right number
For lead generation, the right number is rarely the click-through rate or even the cost per lead. A concept that produces cheap leads that never qualify is a losing concept, however good it looks in Ads Manager.
Wherever possible, judge creative on cost per qualified lead or, better still, on pipeline. That means feeding CRM outcomes back into your reporting, which we cover in our piece on lead metrics.
Keep a library
Every ad that runs should be tagged with its concept, hook, format and persona, and stored with its results. Over time this becomes a map of what your buyers respond to. New team members can read it in an afternoon. Briefs get sharper because they start from evidence instead of opinion.
A shared spreadsheet is enough to start. What matters is that someone keeps it up to date.
What to do with a winner
Don’t just raise its budget and wait. A winning concept is a direction, not a finished ad. Make new hooks for it. Try it in other formats. Try it with a different person on screen. The concept will usually outlast any single execution, and those new versions keep performance up as the original wears out.
That loop of test, learn, brief and test again is the heart of how our creative studio works with our media team. The ads change every week. The lessons build up.