The answer up front: if you spend $5,000 a month or more on Meta, test a minimum of 3 new creatives every week. That is the floor we hold Spark client accounts to. It rises with spend: 5 to 8 a week between $25k and $50k a month, 12 or more once you pass $100k. The full table is below, then the maths that produced it.
Most advice on how many creatives to test per week on Meta ads stops at "test more". Useless. The right number is a function of your monthly spend, your target CPA, and a fatigue clock that does not care how good last month's winner was.
The quota: new creatives per week, by monthly Meta spend
| Monthly Meta spend | New creatives per week | Per month | Expected winners per month |
|---|---|---|---|
| Under $5k | 1 to 2 | 4 to 8 | 0 to 1 |
| $5k to $10k | 3 (our floor) | 12 to 13 | 1 to 2 |
| $10k to $25k | 3 to 5 | 13 to 20 | 1 to 3 |
| $25k to $50k | 5 to 8 | 20 to 32 | 2 to 5 |
| $50k to $100k | 8 to 12 | 32 to 48 | 3 to 7 |
| $100k+ | 12 to 20 | 48 to 80 | 5 to 12 |
A winner here means a creative that beats your target CPA and earns a place in your scaling campaign. The winners column assumes a structured testing process running at a 10% to 20% hit rate. At the industry baseline of roughly 5%, halve those numbers. The gap between the two is most of this article.
Where these numbers come from
Two places. The first is our own playbook. Every Spark client account runs a standing rule: a minimum of 3 new creatives into the testing ad set every week, with no exceptions for weeks when the current ads happen to be performing. On accounts that moved from ad-hoc testing to that fixed weekly quota backed by proper research, the share of tests that win rose from roughly 5% to between 15% and 20%. Same brands, same budgets. The difference was volume and structure.
The second is the largest public dataset on the question. Motion's Creative Benchmarks 2026 analysed 578,750 creatives across 6,015 ad accounts and $1.3B in Meta spend between September 2025 and January 2026. Their weekly launch volumes by spend tier:
| Motion spend tier | Avg weekly launches | Top 25% of accounts | Winner hit rate |
|---|---|---|---|
| Micro (under $10k/mo) | 2.8 | 4.8 | 3.7% |
| Small ($10k to $50k/mo) | 4.1 | 8.1 | 6.2% |
| Medium ($50k to $200k/mo) | 6.7 | 16.0 | 7.3% |
| Large ($200k to $1M/mo) | 11.2 | 31.1 | 8.1% |
| Enterprise ($1M+/mo) | 18.9 | 54.6 | 8.2% |
Two things jump out. Top-quartile accounts launch roughly double the average at every tier. And the winner rate climbs with volume, from 3.7% under $10k a month to 8.2% at $1M+. Accounts that test more do not just find more winners in absolute terms; a higher share of everything they make wins, because every test sharpens the next brief.
Operator guidance runs more conservative. Foxwell Digital puts $25k to $50k accounts at roughly 1 new creative a week and $50k to $100k accounts at 2 to 4. Treat those as survival minimums: enough to hold CPA flat while winners fatigue, not enough to compound. Our quota sits between Foxwell's floor and Motion's top quartile on purpose, because the top quartile is what compounding accounts actually do.
Why the quota rises with spend
Winners are rare, and that is by design
Motion defines a winner as a creative that spends at least 10x the account median and $500+, which in practice means the auction chose to scale it. Roughly 5% of creatives qualify. Motion's own conclusion, in a piece bluntly titled "winners are rare", is that low hit rates are a statistical property of Meta's auction rather than a talent problem. The auction concentrates budget behind the few ads that earn it. So the honest planning number is 1 to 3 winners per 10 tests, and the way to get more winners is more tests.
Spend burns creative faster
According to Triple Whale's DTC panel, the average Meta ad decays meaningfully within 4 to 6 weeks of launch. Spend accelerates that. Push $3,000 a day through three ads and frequency climbs within days; the same three ads on $150 a day might hold for two months. Bigger budgets need a faster replacement pipeline for the same reason bigger fires need more wood.
Andromeda pays for variety
Meta rebuilt its ads retrieval stage on Andromeda, the engine it co-designed with NVIDIA, and reported +6% recall and +8% ad quality on selected segments. The practical consequence: targeting is now largely inferred from the creative itself. The system clusters near-identical ads and treats them as one idea, then matches distinct concepts to distinct pockets of buyers. Feed it four genuinely different angles and it has four doors into your market. Feed it twelve crops of one video and it has one. We unpacked the strategy shift in our Andromeda playbook.
Your account does not have a creative quality problem at 4 ads a month. It has an arithmetic problem.
What counts as one creative?
A countable creative is a distinct entry in your testing ad set with its own hypothesis. A new hook on a proven body counts, because the first 3 seconds decide most of an ad's fate. A new concept obviously counts. A resize, a crop, or a colour-graded duplicate does not; Andromeda clusters those together and your "volume" quietly collapses back to one.
The mix matters as much as the count. A workable split is two iterations of proven concepts for every one brand-new concept. Iterations protect your hit rate. New concepts stop the account from inbreeding, and they are the only route to the next big winner rather than a slightly cheaper copy of the last one. One filmed concept, briefed properly, cuts into 3 or 4 countable creatives through hook and opener swaps. That is what makes the quota affordable.
The maths at $25k a month
Take the tier most brands we speak to sit in: $25,000 a month in spend, target CPA of $40, quota of 6 new creatives a week, so 24 a month.
- Read budget: $120 per creative, or 3x target CPA, the same floor we set in our per-creative test budget guide. Monthly testing media: 24 x $120 = $2,880, about 12% of spend. The other $22k stays on proven winners.
- Expected winners: at a structured 15% hit rate, 24 tests return 3 to 4 winners a month (24 x 0.15 = 3.6). At the 5% industry baseline, the same 24 tests return 1, and some months 0.
- Media cost per winner found: $2,880 ÷ 3.6 ≈ $800. At a 5% hit rate the same winner costs the full $2,880 in test media. Structure, not extra spend, is the difference.
- Fatigue cover: a scaling campaign resting on 3 hero ads that decay in 4 to 6 weeks needs roughly 3 replacements a month. Finding 3 to 4 winners covers it. Finding 1 does not.
Now run the same account at the "sensible" 1 test a week. 4 tests a month at a 15% hit rate is 0.6 winners a month: one new winner every 7 weeks, against a fatigue clock that kills heroes in 4 to 6. CPA drifts upward and nothing is technically wrong. The ads are fine. The arithmetic is not. Kill discipline matters just as much at this volume: losers must exit at 1.5x target CPA with zero conversions, or they eat next week's testing pot.
Key takeaway
The quota is set by the fatigue clock, not by ambition. A winner decays in 4 to 6 weeks and only 1 to 3 tests in 10 win, so your weekly volume has to produce at least one replacement winner per hero ad per month. That is the whole table in one sentence.
How to hit the quota without quality collapsing
The standard objection: "we cannot make 24 good ads a month." Correct, if every ad is a bespoke shoot. So don't build it that way.
Batch at the concept level. 6 to 8 filmed concepts a month, each briefed with 3 scripted hooks and a modular structure (hook, body and CTA as separate blocks), cut into 20 to 32 countable creatives. One creator day covers 2 to 3 concepts. The brief does the heavy lifting, and the full loop from research through iteration is in our creative system guide.
Two failure modes to avoid. Don't fake the quota with crops and resizes; Andromeda collapses them and you learn nothing. And never fund volume by shrinking read budgets below the $100 to $150 floor. Twenty ads at $60 each is noise. Twelve at $120 is a dataset.
The honest constraint is production. Most brand-side teams we talk to can produce 4 to 8 finished creatives a month in-house. The quota at $25k+ starts at 20. That gap, not media buying, is usually what stalls scaling, and closing it is the specific thing our monthly creative engine is built to do. Pricing is per creative, so volume scales with your spend tier rather than with an agency retainer. The rest of the testing system, from research and briefs to benchmarks and kill rules, is mapped across the resources hub.
FAQ: creative testing volume on Meta
How many new creatives per month do you need on Facebook ads?
Work from spend: 12 to 13 a month at $5k to $10k, 13 to 20 at $10k to $25k, 20 to 32 at $25k to $50k, and 48 or more past $100k. Below $5k a month, 4 to 8 well-read tests beat a bigger number spread too thin, because each creative still needs $100 to $150 of spend for a clean read.
Is it better to test fewer creatives with more budget on each?
No. A creative needs roughly 2x to 3x your target CPA in spend (floor of $100 to $150) to give a readable signal. Spending much past that on a test buys very little extra information. Once a creative has a clean read, the next dollar teaches you more inside a new test. If budget is tight, cut the quota before you cut the read budget.
Does creative volume still matter with Advantage+ and Andromeda?
It matters more. Meta's Andromeda retrieval engine infers targeting from the creative itself, so the ads you feed it are the targeting. Accounts feeding it distinct concepts give the system more buyer segments to match against. Ten near-identical crops of one video do not count; the engine clusters them as one idea.
What percentage of tested creatives become winners?
Motion's 2026 benchmarks put the average at roughly 5%, ranging from 3.7% for accounts under $10k a month to 8.2% at $1M+. Across our client accounts, a structured weekly quota backed by proper research runs at 15% to 20%. Plan on 1 to 3 winners per 10 tests and budget for the occasional month with none.
If your spend tier says 20+ creatives a month and your current ceiling is 6, you now know exactly what is throttling the account. Tell us your spend and target CPA and we will map the quota, the batch plan and the cost of hitting it.