There is no formula for a winning ad. Here's what 578,750 of them told researchers anyway.

The pattern-hunting instinct is universal. You bring a media buyer two ads to review and the first question is always “but what actually works. Do we need UGC. Do we need a face. Do we need a hook. Do we need vertical.” What they’re asking for is a formula.

There isn’t one. Andrew Foxwell, summarising Motion’s Creative Benchmarks 2026, says so plainly: “there is not and will never be a perfect formula” for winning creatives.^[1] Which is honest, and a bit unsatisfying. So the rest of this piece is what the aggregate data does show, which is real, and why that data is prior signal, not a rule.

The size of the winner pool

Motion’s Creative Benchmarks 2026 analysed 578,750 unique creatives, 6,015 advertiser accounts, and $1.29 billion in Meta ad spend, from September 2025 through January 2026.^[2] Their definition of a winner is deliberately blunt: at least 10× the account’s median ad spend, at least $500 in total delivery. About 5% of creatives clear that bar.

The hit rate is not uniform across advertiser sizes. Micro accounts (under $10K/month) sit around 3.8%. Enterprise accounts ($1M+/month) hit roughly 8.2%.^[1] Two things drive that gap, and neither is creative genius. First, Motion’s winner definition requires at least $500 in total ad delivery, and micro accounts often kill or under-fund tests before crossing that floor even on their better creatives. The definition itself makes it harder for small accounts to book “winners.” Second, enterprise accounts run more mature process: more completed Learning Phases, cleaner budget allocation, faster kills. On top of both, they test 18.8 creatives per week on average versus a handful at micro, which produces more absolute winners even at a similar underlying rate. Volume drives the count of winners; it does not necessarily lift the percentage.

What Motion found about hook types

Motion also publishes the five hook formats that appear most often among $1M+/month enterprise winners: confession, bold claim, relatability, contrast, and curiosity.^[3] These are ranked directionally in Motion’s public materials; the per-hook hit rate is not published on their public pages. What is published is the shape of the winner pool at the biggest spend tier: five specific hook stances that recur across accounts.

Sepia, a creative-analytics platform, published a summary in June 2026 citing Motion’s dataset with more specific format hit rates: text-only ads 11.6%, product-plus-text-overlay 8.75%, UGC 7.56%.^[4] These percentages appear in Sepia’s report citing Motion 2026; I could not independently verify the exact numbers on Motion’s own public pages, so read them as “reported by Sepia, citing Motion.” The direction is not surprising, but the causal read matters. Text-only ads are cheap to produce, which changes both how many get tested and how long each gets funded before being killed. Whether that shifts what gets selected into the winners pool or reflects a real format effect, Motion’s public data does not tell us.

What Meta itself has measured

Meta’s own product page for Reels ads reports a global meta-analysis of 15 split tests across ecommerce, retail, and CPG SMBs: vertical 9:16 Reels ads with audio and creative in safe zones showed 34.5% lower cost per result than image ads on Reels placements, at 99.9% statistical confidence.^[5] This is a format lift within Reels placement, not a blanket claim that video beats static on every Meta surface.

The stat that surprises media buyers most is older. In partnership with Nielsen, Facebook pooled 173 of Nielsen’s Brand Effect studies and split viewers by video watch length. Viewers who watched less than 3 seconds of a Facebook video ad still showed 47% ad recall lift, 32% brand awareness lift, and 44% purchase intent lift.^[6] The sub-3-second window does most of the brand work by itself. Which is why Meta’s guidance keeps pushing “hook in the first 3 seconds.” Most of your viewers see nothing else.

“Viewers who watched less than 3 seconds of a Facebook video ad still showed 47% ad recall lift, 32% brand awareness lift, and 44% purchase intent lift.”

Facebook + Nielsen, 173 Brand Effect studies · Marketing Dive coverage, 2015

What academia says about stopping the scroll

The peer-reviewed anchor for “what stops a scroll” is Pieters, Wedel, and Batra’s 2010 Journal of Marketing paper, “The Stopping Power of Advertising.”^[7] They tracked eye movement across 249 print ads and separated two kinds of complexity. Feature complexity, meaning dense perceptual features on the page, hurt brand attention and ad attitude. Design complexity, meaning elaborate creative design, helped attention to the pictorial and to the ad as a whole. Translation for Meta creative: cluttered visual noise loses attention, considered creative design earns it. Print is different from feed video, but the eye is the eye, and the mechanism has held up across two decades of subsequent work.

What the pattern is not

Now the honest half. Everything above is aggregate prior signal, not a formula.

• Motion’s hit rates say text-only wins more often on average. That does not mean text-only wins for your D2C jewelry account.
• Meta’s Reels format lift is measured against image ads on Reels placement. Not against every creative choice you would make.
• The Nielsen brand-lift result is measured across a set of already-produced 2015 Facebook ads. It says the first 3 seconds do brand work; it does not say your first 3 seconds will do brand work.
• The five Motion hook stances are patterns that recur at $1M+/month. Under a tenth of Meta advertisers spend at that tier.

There is a shape to the data. Winning ads on Meta cluster around a small number of formats, hook shapes, and design disciplines. But 5% is the hit rate at scale, which means 95% of any advertiser’s test attempts fall short. Including 95% of the ones that check every recommended box.

WordStream’s 2025 benchmarks make this concrete by vertical. US Traffic CTR ranges from 4.13% at the top (Shopping/Collectibles/Gifts) to 0.80% at the bottom (Automotive Repair).^[8] Lead-objective CPL runs from $3.16 (Restaurants) to $76.71 (Dentists). If your vertical sits in the bottom half of either range, the aggregate patterns above do not carry the same weight. You are competing in a very different auction.

Why the formula fails anyway

Two structural reasons.

Aggregate patterns hide your account. The 5% hit rate is measured across 6,015 accounts spanning every category and audience Motion sees. Your account has one audience, one product, one seasonal cycle, one Learning Phase history. The distribution of winners for you is not the distribution for the aggregate. Every practitioner will tell you the same thing eventually: your account has its own patterns, and you have to find them.

The Learning Phase eats formulas alive. Because Meta requires roughly 50 optimization events per ad set per week to exit Learning Phase, even a genuinely great creative can read as a losing test if you kill it early, split too many ad sets, or edit the creative and reset the learning window. The pattern data above is what wins after Learning Phase stabilises. It is not what wins in the first three days of a test.

How Splitroom fits

A well-seeded batch of six or eight concepts, run through Splitroom against a defined audience, produces a directional read on which two or three are worth the Learning Phase spend to test live on Meta. The formula is still not there. The filter is. What a pre-Learning-Phase filter changes is the ratio of winners per dollar spent, not the underlying hit rate. Losers identified before they touch the auction are budget you keep. Learning Phases you do not reset are budget you do not spend twice. That is the mechanism, and it is separate from the aggregate hit-rate question. Whether your own account rate ends up at 4% or 8% will depend on your audience, your product, and your seasonal cycle. What the filter does is compound whichever rate you land on.

The one-liner

The pattern-hunters are half right. There are patterns. There is no formula. The 5% hit rate is not a design brief; it is a distribution. Winning consistently on Meta is a matter of respecting both halves of that sentence: seed the pattern space you have reason to believe in, and let the volume decide which seeds carry your audience.