Performance marketing · Creative
A creative-testing system for paid social that a small team can run
On Meta and TikTok the creative now does most of the targeting, which makes testing ads the main job of a paid social team. Most teams still test at random: new ads when the old ones tire, winners picked by gut. Here is a simpler system — one hypothesis per test, a set budget, a rule for calling the winner and a monthly calendar.

Why testing became the job
With broad targeting and automated placements, the ad itself decides who sees it. A video about back pain gets shown to people who respond to videos about back pain. In practice, the creative is the targeting.
So the team that learns fastest about its creative wins. Not the team with the most ads, but the team that knows why its ads work and can make the next one on purpose.
Start with a hypothesis, not an ad
Write one sentence before any production. “If we open with the price instead of the product, cost per purchase will drop, because our buyers compare prices.” Change, expected result, reason. If you can’t write the reason, you’re guessing, not testing.
Good sources for hypotheses: reviews and support tickets (the words customers use), comments under your ads, the sales team’s most common objections and competitor ads that have been running for months in the Meta Ad Library.
Test one variable at a time. Roughly in order of how much they move results:
- the concept: what the ad is about (a problem, a demo, a comparison, a testimonial, an offer);
- the hook: the first two or three seconds of a video, or the headline of a static ad;
- the format: creator-style video, studio video, static, carousel;
- the offer and the call to action;
- the details: colors, music, captions, length.
Big swings first. Settle which concept works before you test which caption font does.
A budget for every test
Set the budget from the metric you will judge by, not from what’s left at the end of the month.
An example: you judge by cost per purchase, yours is currently $40, and you want at least 25 purchases per variant before deciding. That’s about $1,000 per variant, or $3,000 for a three-way test. If that’s more than you can spend, judge by a metric higher in the funnel, such as cost per add to cart, and confirm the winner on purchases later.
Keep tests in their own campaign with a fixed share of the budget, say 10–20%, so they don’t compete for delivery with ads you already know work. One caution: several ads in one ad set do not get an even split, because the system quickly picks a favorite. When you need a clean comparison, use the A/B test option in Meta Ads Manager, or give each variant its own ad set with the same budget and audience.
When to call a winner
Write the rules in the test sheet before launch, so nobody moves the goalposts afterward.
- The metric. One primary metric, such as cost per purchase or cost per qualified lead, plus one or two diagnostics: hook rate (three-second views divided by impressions) and click-through rate.
- The minimum. Both money and time: the planned budget and at least seven days, so weekdays and a weekend are included.
- The margin. A winner beats the control by a gap you would act on, say 20% on the primary metric. Anything closer is a tie; keep the one that’s cheaper to produce.
- The confirmation. A winner moves to the main campaign and has to hold for another week at full budget.
The usual ways to spoil a test: calling it on day two, judging by click-through rate alone (clickbait wins clicks and loses sales), editing an ad mid-test, testing in the week of a sale.
A losing ad is still a result. Write down what it taught you.
A monthly creative calendar
A fixed rhythm keeps testing from turning into a scramble:
- Week 1: review last month’s tests, pick two or three hypotheses, brief production.
- Week 2: produce and launch; for example, two new concepts with three hooks each.
- Week 3: read the results by the rules; make iterations of the winner with new hooks and formats.
- Week 4: move winners into the main campaigns, retire tired ads, update the learning log, plan next month.
Split the output: most of it iterations on what already works, a smaller share new concepts, say 70/30. Iterations keep results steady; new concepts find the next winner.
The learning log is a shared sheet: date, hypothesis, variants, spend, result, decision, lesson. After six months it is the most useful document in the account, and it stays with you when people change.
What to do this week
- Pull the last 90 days of ads, sort by spend, and write one line on why each of the top five might be working.
- Write three hypotheses in the one-sentence format.
- Set up a testing campaign with a fixed budget and the decision rules written down.
- Start the learning log.
- Put next month’s four calendar slots on the team’s schedule.
Boson Group runs paid social and makes the creative for it, and Boson Factory, our content factory, delivers a steady stream of short creator videos — the raw material a testing system needs. Write to team@boson.group.




