Meta ad performance is not evenly distributed, and the size of the gap is the useful part. In a 2026 dataset covering 578,750 creatives, roughly 5% of ads spent at least 10 times their account median, while about half received little or no spend at all.
That shape changes how you should read a low hit rate. If most ads are expected to stay ordinary, a run of unremarkable creatives is not automatically evidence of a bad strategist. It may be the normal shape of a search process in which a few outliers carry the account.
What was actually measured
Motion’s Creative Benchmarks 2026 analysed anonymised Meta advertising data with this scope:
| Dataset field | Published value |
|---|---|
| Advertiser accounts | 6,015 |
| Unique creatives | 578,750 |
| Realised Meta spend | $1.29 billion |
| Platforms | Facebook and Instagram |
| Launch window | September 1, 2025 - January 1, 2026 |
| Minimum creatives per included account | 10 |
The window matters as much as the sample size. It covered pre-holiday testing, Black Friday and Cyber Monday, and the post-holiday reset, so promotional hooks were being measured in the quarter that flatters them most.
The study defined a winner as a creative meeting both conditions: spend at least 10 times the account median creative spend, and at least $500 in absolute spend. A mid-range creative ran at least 28 days without clearing that bar. A loser stopped, or never reached active spend, before 28 days. These are delivery classifications, not judgements of craft or profit. The full definitions are in the Motion Creative Benchmarks 2026 methodology.
Spend concentration is the headline, not the hooks
Motion reports that roughly 5% of ads spent at least 10x their account median, that about half of all ads received no spend or very little, and that around 6% were responsible for the majority of spend inside their accounts.
Spend is still an imperfect proxy for success. Meta can allocate heavily toward an ad while the advertiser remains below its margin-adjusted target, and the study says explicitly that it cannot infer ROAS, revenue impact, or causality from spend concentration. What the distribution does establish is the arithmetic of the search: if roughly one in twenty creatives earns real delivery, the number of distinct ideas you can put into the auction is a first-order constraint.
Hook hit rates in the published dataset
Hit rate here is the percentage of creatives in a category that met the winner definition.
| Hook or headline type | Reported hit rate |
|---|---|
| Offer only | 9.29% |
| Confession | 8.74% |
| Curiosity | 7.77% |
| Bold claim | 7.19% |
| Storytelling | 6.23% |
| Question | 5.47% |
| How-to | 5.47% |
| Explainer | 5.24% |
This table does not mean an offer-only hook is 77% better than an explainer for your business. Ads were not randomly assigned to hook categories, business models, offers, or audiences, and seasonality favoured offer-led language across the collection window. Read it as a hypothesis generator.
The test it suggests is to hold the offer and audience stable while varying the opening frame across five genuinely different angles:
- 1.Direct offer
- 2.Specific problem
- 3.Product demonstration
- 4.Customer objection
- 5.Contrarian or curiosity-led claim
That produces meaningful variation. Changing punctuation, background colour, or a single adjective usually does not, and it consumes the same delivery budget.
Asset hit rates complicate "high production always wins"
| Asset type | Reported hit rate |
|---|---|
| Text only | 11.60% |
| Product image with text | 8.75% |
| UGC | 7.56% |
| High production | 6.97% |
| GIF | 6.82% |
| Lifestyle image with text | 6.10% |
| Animation | 4.57% |
The finding is not that text-only ads are universally superior. Fast, text-forward formats are cheap to vary, so the teams using them tend to explore more propositions per month. High-production assets can build trust and explain complex products, but every revision that requires a shoot lengthens the feedback loop.
The advantage being measured is speed-to-learning: more distinct concepts leads to more tested propositions, which leads to more chances of finding an outlier. That chain holds up better than "cheap creative beats expensive creative", and it is the reason production cost per asset is the wrong number to optimise on its own.
What competitor research can and cannot reveal
Meta’s Ad Library shows observable details: the advertiser Page, the creative, the copy, the platforms, and the delivery dates. Eligible UK and EU ads carry additional estimated impression and targeting fields. Ordinary observation does not reveal a competitor’s purchases, CPA, profit, or ROAS, and no third-party tool can read those numbers either.
So the honest reading of a competitor’s library is narrower than most teams assume:
- A long-running ad is evidence of duration, not proof of profitability.
- Many active variants show testing activity, not creative quality.
- A repeated claim shows message consistency, not verified demand.
- A format that appears often is a pattern worth testing, not a winner to clone.
That still leaves plenty worth extracting, which is what tracking and monitoring competitor Meta ads is for: which propositions a category keeps returning to, which formats a competitor has committed production budget to, and how long a given angle survives. See the Meta Ad Library API documentation for what the platform itself exposes.
Turn the benchmark into your own taxonomy
Hit-rate tables only become useful once every creative in your account is attached to a hypothesis you can group by:
| Dimension | Example values |
|---|---|
| Audience problem | Cost, speed, uncertainty, complexity |
| Promise | Save time, reduce waste, improve control |
| Proof | Demo, number, testimonial, mechanism |
| Hook | Offer, question, confession, explainer |
| Format | Static, UGC, demo, founder, animation |
| Funnel stage | Awareness, consideration, conversion |
Then measure yourself against your own account rather than against a benchmark table:
- What share of concepts receive enough delivery to be evaluated at all?
- Which problems and promises repeatedly earn spend?
- Are your new ads genuinely distinct concepts, or correlated variants of one idea?
- Which stable mid-range ads should stay live while the search continues?
- Do platform winners also clear your contribution-margin target?
The last question is the one the dataset cannot answer for you, and it is the one that decides whether a winner is worth keeping.
Related Reading
For the testing layer that turns these hypotheses into scheduled work, see the Meta creative testing guide and how many ad creatives a Meta account needs. To organise observable competitor patterns into a taxonomy like the one above, AdRiseLab competitor intelligence does the collection step. And why most new creatives get little delivery covers the mechanism behind the 5% concentration figure.