Ask a DTC brand to walk you through their creative testing process and you usually get a shrug dressed up as a workflow. New ads go live whenever the backlog clears. Results get read when someone remembers to look. Two weeks vanish, the account is still leaning on the same two winners, and nobody can say what the last round of tests actually taught them.
The fix is rarely more ideas and it is almost never more budget. It is a calendar. What follows is the weekly creative testing process for Meta ads we run at Spark, written out as a checklist with a time box on every step, so you can run it yourself in about four hours of decision time a week. The important word is "run". The sprint only works if it happens every week, including the weeks when the numbers are ugly.
Why run creative testing on a weekly cycle?
Three mechanics make the week the right unit, and none of them are about preference.
First, creative is the biggest input you control. An August 2023 meta-analysis by NCSolutions and Nielsen, built on nearly 450 sales effect studies, put creative at 49% of advertising's contribution to incremental sales. Targeting, the thing most teams tinker with instead, came in at 11%. When half the outcome rides on the ad itself, the ad deserves a standing weekly slot in your diary, not leftover attention.
Second, Meta's delivery system already runs on a weekly clock. An ad set usually exits the learning phase after about 50 results in the week following its last significant edit, per Meta's Business Help Centre, and editing an ad set mid-learning resets that clock. A seven day sprint maps onto the exact window Meta uses to judge whether your ads can deliver stably.
Third, the volume bar has moved. In his January 2026 Motion talk on testing after Meta's Andromeda update, Caleb Kruse argues that launching 10 to 15 ads a month and hoping for one or two winners is obsolete, and that serious accounts now need 50 to 80 a month with genuine concept diversity. We would add a caveat from our own client work: cadence matters more than raw count. A brand that ships and properly reads five tests every single week will beat a brand that dumps forty ads into the account each quarter and reads none of them.
Most accounts do not have a creative problem. They have a cadence problem. The testing calendar is a bigger lever than the testing budget.
The sprint at a glance
| Day | Time box | The job | What exists afterwards |
|---|---|---|---|
| Monday | 45 min | Read last week's results, verdict every ad | A verdict log, one written learning per ad |
| Tuesday | 60 min | Market and competitor research | Three written hypotheses |
| Wednesday + Thursday | 60 to 90 min | Brief production, one variable per test | Briefs in the pipeline, ads named to a convention |
| Friday | 30 min | Launch tests beside your winners | New tests live before the weekend |
| Monthly | 30 min | Hit rate review | A decision: keep going, fix research, or raise volume |
Monday: read last week before you make anything new (45 minutes)
The sprint starts with reading, not making. Monday's job is to turn last week's spend into explicit beliefs.
- Pull every ad that spent in the last 7 days. Sort by spend, not by CPA. An ad that burned budget without converting is a verdict in itself.
- For the top five by spend, read the funnel from the top: hook rate, then hold rate, then CTR, then cost per result. The order matters because each metric points at the part of the ad that failed. Our guide to reading a Meta creative report covers the thresholds we use.
- Give every ad one of four verdicts: scale, iterate, kill, or too early. Friday's launches will almost always be "too early" on their first Monday; that is the correct call, not a cop out. We published the CPA rule we use for kill decisions if you want a hard trigger.
- For each verdict, write one sentence beginning "This tells us". That sentence is the actual asset. Winners fatigue and die; learnings stack.
- Keep the log somewhere the whole team can see. A spreadsheet is fine.
45 minutes is enough because Monday is not a debate. You are applying rules you already set, kill thresholds and scale criteria, to fresh numbers. If Monday regularly takes two hours, the problem is missing rules, not a missing analyst.
Tuesday: one research hour, three hypotheses
Tuesday exists so your next ads are informed bets rather than guesses. Split the hour deliberately.
- 25 minutes on the market. Read fresh reviews of your product and your category, the comments under your own ads, and whichever Reddit threads your buyers actually post in. You are collecting objections, use cases and exact phrasing, verbatim.
- 25 minutes on competitors in the Meta Ad Library. Anything a rival has kept live for six weeks or more is probably paying for itself. Note the hook, the angle and the format rather than the surface execution. Our Ad Library research walkthrough shows how to read it quickly.
- 10 minutes writing exactly three hypotheses, all in one fixed shape: because we saw X, we will change Y, and we expect Z to move.
A real example of the shape: "Because eleven reviews this month mention using the product at work, we will brief a desk setting UGC ad, and we expect CTR above the account median." One variable per hypothesis. "Try a new UGC ad" is not a hypothesis. It is a hope with a budget attached.
Wednesday and Thursday: brief it and load the conveyor
Midweek turns hypotheses into production. Each one becomes a one page brief: the hook, the angle, the format, and the single variable this test exists to answer.
Name everything to a convention while you are in there. An ad's name should carry its concept, hypothesis and version, because in six weeks someone will be staring at Ads Manager trying to remember what "Final_v3_NEW" was supposed to prove.
Be honest about lead times. A full creator shoot at Spark takes about 21 days from brief to live ads. The tests you launch this Friday were briefed two or three weeks ago; statics and recuts of existing footage can turn around inside the week, but net new video cannot. So the midweek block is loading the conveyor for future Fridays, not making this Friday's ads. Across the 5,000+ ads we have made, the accounts that compound are the ones where that conveyor never runs dry, and the accounts that stall are the ones that only brief new work after the current winner has already died.
Friday: launch beside your winners, then leave it alone
Friday is a 30 minute launch slot with three rules attached.
- New tests go in a dedicated testing campaign with ad set budgets (ABO), one concept per ad set, equal budgets. Top Growth Marketing's creative testing framework is a sensible reference here: roughly 50 conversions per variant before you trust a comparison, and 2 to 3 times your target CPA in spend per variant before you call it.
- Never pause a winner to fund a test. Winners pay for the testing programme. Keep the bulk of spend on proven ads and carve the testing budget out separately.
- Then stop touching things. Budget, creative and audience changes can all count as significant edits that reset learning. The test gets its week whether you are nervous or not.
What this costs depends entirely on your CPA. Our breakdown of how much to spend per creative test runs the maths by spend tier.
The monthly layer: judge hit rate, not hot streaks
Once a month, count two numbers: tests that got a fair read, and tests that either beat your control CPA or produced a learning you acted on. The second divided by the first is your hit rate, and it is the only score the sprint is judged on.
Expect it to be low. A genuine winner is a 1 in 5 to 1 in 10 event in most accounts, which is exactly why a single bad week means nothing. One bad week is variance. A bad month on honest volume means the research step is broken, so you fix Tuesday, not Friday. And if the hit rate is fine but growth is not, you have a volume problem: our weekly testing quota by ad spend gives the tier by tier numbers.
Run the same five moves every week: read, research, brief, launch, and once a month score the system on hit rate. Weekly discipline plus monthly judgement is the whole trick. Accounts fail at this by changing the system every time a single week disappoints.
Where the sprint falls apart
Five failure modes we keep seeing in audits, in rough order of damage.
- Skipping the Monday read after a rough week. That is the week the read pays most, and skipping it is how accounts end up flying blind for a month.
- Only iterating the current winner. All variations, no new concepts. Then the winner fatigues and there is no successor in the pipeline.
- Killing ads on day two CPA. "Too early" is a legitimate verdict. Use it.
- Editing live tests midweek. Every significant edit resets learning, which quietly burns the test you were trying to read.
- Letting hypotheses rot into "we need more ads". Volume without a question attached produces spend, not learnings.
More testing playbooks live in our resources hub if you want to go deeper on any single step.
FAQ: the weekly creative sprint
What is a weekly creative sprint in Meta ads?
It is a fixed weekly cycle for creative testing. Monday you read last week's results and give every ad a verdict. Tuesday you do one hour of research and write three hypotheses. Midweek you brief production, one variable per test. Friday you launch the new tests beside your winners, and once a month you review your hit rate. Decision time is about four hours a week, production aside.
How many new ad creatives should you test each week?
Scale it to spend. Post Andromeda guidance runs high: Caleb Kruse's January 2026 Motion talk argues for 50 to 80 ads a month on accounts with real budget. We care more about the floor than the ceiling: three to five concept tests a week, each read properly, beats a quarterly dump of forty ads nobody reviews. Every test needs enough budget to reach roughly 50 results in a week.
How long should a Meta creative test run before you judge it?
About a week. Meta says an ad set usually exits the learning phase after around 50 results in the week after its last significant edit, and editing sooner resets the clock. Judge a test after it has had that window, or mark it too early and wait. Day two CPA is noise, not a verdict.
Can one person run the weekly creative sprint?
Yes. The time boxes add up to under four hours of decisions a week. Production is the real constraint, so keep a rolling pipeline: briefs written this week become ads two to three weeks out, and the Friday launch slot never waits on a shoot.
If you would rather not staff it
The sprint is free to copy and most teams can run the decision layer themselves. The part that breaks first is production volume: three fresh concepts a week, every week, is a real content engine. That engine is the thing we operate for brands, research through production through the weekly read, and if you want it run for you, tell us about your account. Either way, steal the checklist. It will make your Mondays more useful than your last quarter of ad hoc testing.