How Many Ad Creatives Do You Need Per Week on Meta? (2026 Benchmarks)

If you run paid social on Meta, Motion's 2026 Creative Benchmarks give you a well-grounded answer: the average account launches about 3 to 19 new creatives a week, depending on monthly spend, and the top 25% of accounts launch roughly double to triple that. The dataset covers $1.29B of Meta ad spend, 6,015 accounts and 578,750 creatives from September 1, 2025 to January 1, 2026. Below is that number by spend tier, how much of it should be versions of winners versus new ideas, what Meta's Andromeda system changes, and what 50, 100 or 300 videos a month cost to produce.

How many creatives per week, by monthly Meta spend

These are Motion's averages for new creatives launched per week (from its April 2026 research page), next to the top 25% of accounts in each tier (from Motion's March 2026 talk). The monthly column is the weekly figure times about 4.33 (52 weeks over 12 months).

Keep in mind this is Meta-only data. Motion counts unique creatives launched on Facebook and Instagram and doesn't split them by format, so assume statics and carousels sit inside these numbers, not on top of them.

If you've seen 50 creatives a week as a target, in Motion's data that only matches the top 25% of accounts spending $1M+ a month (54.6 a week). At $30K a month, the average is about 4 a week and the top quartile about 8.

Why volume matters: winners are rare

Motion defines a winner as a creative that spends at least 10 times the account's median creative spend and at least $500. By that definition about 5% of creatives win, and those winners take 55% of all spend. Hit rates run from 4.0% at the smallest tier to 8.8% at enterprise.

Motion's own framing is that hit rate is only useful alongside volume. Its example: a 20% hit rate on 5 launches gives you 1 winner, while 10% on 50 gives you 5. Its rule of thumb is that 20 launches produce about 1 to 1.6 winners and 50 launches about 2.5 to 4.

So work backward from the winners you need. If you want two fresh winners a month and your hit rate is around 5%, you need about 40 launches that month.

Not everyone agrees on the high end. Andrew Foxwell of Foxwell Digital (June 2026) argues that brands spending $100K+ a month need about 8 to 10 new concepts a month, that 40 to 50 new assets a month is plenty for most, and that around $5K a month, 4 to 5 assets is enough. His point is concept quality over endless versions. Part of the gap is counting: Foxwell talks about concepts, and Motion counts every creative launched. But even on assets, Foxwell's 40 to 50 a month sits near Motion's averages for $200K to $1M, not the top quartile.

Iterations vs new concepts: what share of your output

There is no measured benchmark here. Motion's methodology does not separate iterations from net-new concepts, so every split you hear is a practitioner opinion:

• Half to most iterations. Aaron Reardon at Conduit (July 2026) suggests roughly half your output should be iterations on proven ads at $50K to $200K a month, and at $200K to $1M gives an example of about 15 new concepts to 35 iterations a month (about 70% iterations).
• Mostly new. According to a Sepia Lab summary, Barry Hott recommends about 90/10 in favor of net-new.

A practical way to pick your own split: if you don't have a winner yet, lean heavily toward new concepts, because iterating on nothing teaches you nothing. Once you have winners, iterate, but make each iteration a real change. A different person, setting, opening or angle is a test. A new text overlay on the same footage mostly isn't.

The opening deserves special attention. A Meta and Analytic Partners study of 500+ ads, reported by Social Media Today in 2022, found ROI was 2.3 times higher when the product appeared in the first two seconds. The data is old, but it suggests that what happens in the first two seconds, such as when the product shows up, is worth testing.

What Andromeda changes about creative volume

Andromeda is Meta's ads retrieval system. According to Engineering at Meta (December 2024), it narrows tens of millions of ad candidates down to a few thousand, using an index built to scale to large creative volumes, including generated creative. In March 2025 Meta wrote that Andromeda supports advertisers who "upload more and more creatives to support a diversification strategy." Creative diversification is also one of the five pillars in Meta's Performance 5 framework, though that page gives no number of ads per ad set.

Meta's Business Help Center has an article titled "Creative diversity," which third-party write-ups describe as a score of how similar the creatives in an ad set are. The specific figures you may see quoted come from third parties, not Meta's documentation: a 60% similarity score (Billo's working number, according to Novoads) and "Entity ID" clustering (described by Atria).

The practical reading: ten near-identical versions may behave like far fewer distinct ads. Volume still matters, but only volume that is actually different. We go deeper in Meta Andromeda and creative diversity. How you structure the test matters too, since a campaign budget can starve new creatives before they get a fair read, which is covered in CBO starves your best creative.

What 50, 100 or 300 videos a month cost to produce

For a reference point, according to PPC.io's 2026 UGC pricing guide, a single UGC video averages about $212, with most rates between $150 and $300. Creators bring things generation can't, especially for brand-new concepts, but at that rate 100 videos a month is around $21,000. We break down creator pricing in UGC creator rates and the wrong number.

Riffkit bills by the second in credits, and failed renders are not charged. Here is what each public plan works out to per second on two engines, MiniMax H3 at 768P (40 credits a second) and Seedance 2.0 at 720p (100 credits a second):

On a monthly plan that puts a 15-second video at about $2.70 to $3 on H3 and about $6.75 to $7.50 on Seedance 2.0 at 720p. Seedance 2.0 Fast (80 credits a second at 720p) sits between them, and Seedance 2.5 (150 at 720p) above. The engines differ in more than price, which we cover in Seedance 2.5 vs MiniMax H3.

Now the higher volumes, priced at Studio's per-second rate:

To be clear about what this table is: per-second math for budgeting, not a price list. It assumes one render per video you keep. Failed renders aren't charged, but renders you decide not to use are, so budget extra seconds for retakes. The largest public plan, Studio, covers 36 H3 or 14 Seedance 2.0 fifteen-second videos a month, so all three of these volumes are above it. Contact [email protected] for more volume.

Where Riffkit fits, and where it doesn't

Riffkit starts from a short video that already won: a template from its library, a TikTok link, or your own upload, up to 45 seconds. It analyzes why that video worked (the hook, the pacing, the beats). Riffkit never re-uploads the source clip: your video is a new render, though flat graphics from the source (cards, stickers) can carry over into it. Everything you make is yours to post, run as ads, and monetize, with no extra license or revenue share from us; what comes from someone else isn't ours to give, such as a source video's music or, in a Swap, the people and voices it keeps. Adapt keeps the formula and tells a new story for your product. Swap keeps the shots, camera and action and changes the person, product or details. You can run one template with several characters in a batch, reframe a finished video to 9:16, 1:1 or 4:5 (re-rendered, not cropped; see one winning video for every placement), and generate in 10 languages. It runs as a hosted web app and as a skill for AI coding agents like Claude Code, Codex and Cursor.

What it doesn't do matters just as much. Riffkit doesn't connect to your ad account, launch or measure ads, or read performance data. It doesn't make static images or carousels. Edits to just the first three seconds of an existing file belong in a video editor, and platform AI labels stay as the platform applies them. For agencies: there are no client workspaces, white-label or client review links.

And one honest Andromeda caveat. Versions made from one source share its opening and rhythm, so Meta may still group them. Genuinely different ads need different people, scenes and angles, and often several different winning sources.

Which number should you plan for?

• If you spend under $10K a month on Meta, plan for 3 to 5 new creatives a week and favor new concepts until something wins. Creator on H3 covers about 13 fifteen-second videos a month, Studio about 36.
• If you spend $10K to $50K a month, aim for 4 to 8 a week, roughly 18 to 35 a month. Studio on H3 covers that range.
• If you already have a winner that's fading, make Swap versions that change the person and setting rather than the text, and pull in a second winning source so your next batch doesn't share one rhythm.
• If you spend $50K+ a month or run creative for several brands, the benchmarks put you at 7 to 16 a week at $50K to $200K, about 29 to 69 a month, and more above that. Studio on H3 covers the low end; past 36 videos a month, contact [email protected] for more volume, and plan to keep each client's work separate on your side.
• If you're planning for Black Friday, Meta's holiday guide says to start ads by mid-October and launch the Black Friday set at least two weeks before the peak, about November 13 this year. Our Black Friday and Christmas 2026 timeline maps that week by week.
• If you're not sure where your hit rate sits, count winners by Motion's definition over the last 90 days and divide by launches before you pick a weekly target.

You can start with free credits, no card needed, and see every per-second rate on the pricing page.