Four creative benchmarks get quoted constantly across paid social, and none of them survives being followed back to its source. The first is a trade article from 2016. Another was measured and then deliberately never published. A third is real research about a completely different kind of advertising. The last does not exist at all.
Every check below takes minutes and you can repeat all four yourself. These are the creative benchmarks the industry treats as settled.
This is for you if you have ever been handed a target and wondered where the number came from.
Creative benchmarks one: the 85% sound off claim is not research
This is the most repeated statistic in paid social and it traces back to nothing resembling a study.
It traces to a Digiday article from 17 May 2016, which reported what four publishers said about their own analytics. No methodology was given. Sample size is absent. Facebook data does not appear anywhere in it.
Read the original and it gets worse. PopSugar, quoted in the same piece, gave a range of 50 to 80 percent rather than 85. So even inside the source, the figure is not agreed.
Then consider the age. This described a product that predates Reels entirely, on a platform that has rebuilt its video experience repeatedly since. A behavioural claim from 2016 about a video format that no longer exists is not evidence about anything you are running now.
There is a real number available. Verizon Media and Publicis surveyed 5,616 US consumers and found 80% more likely to watch an entire video when captions are available, with 69% watching sound-off in public and 25% in private. Disclosed sample, stated method, and it supports the same practical advice: caption your videos.
Creative benchmarks two: no hook rate figure exists at all
Not a wrong benchmark. No benchmark at all.
Meta has never published a target for hook rate, hold rate or thumbstop rate. Search for one and you will find plenty of pages offering figures, most commonly an eighteen to twenty eight percent range for hook rate.
Follow any of those citations. They terminate within two or three steps in another article with no source, usually on a domain publishing benchmark content across dozens of unrelated topics.
The research behind this cluster catalogued that pattern across sixteen sub-topic searches in creative. On hook rate, hold rate, thumbstop rate and creative testing volume, the entire visible first page was vendor content citing other vendor content.
So the honest position on these creative benchmarks is that they were invented to fill a gap. Reporting the absence is more useful than repeating a number, and almost nobody ranking for those terms will do it.
Creative benchmarks three: the threshold Meta chose not to publish
This one is the most instructive, because the measurement genuinely happened.
Facebook analysed frequency across 2,439 Brand Lift campaigns between November 2019 and March 2020. The analysis confirmed a plateau in effectiveness after a handful of impressions.
Then it published no threshold number. Deliberately.
Meanwhile the industry publishes several. Frequency above 2.5 means fatigue. Refresh every two weeks. Cap at three. None of those trace to Meta or to any dataset.
So the party with 2,439 campaigns of data declined to name a number, and parties with no data named many. That contrast is the whole argument, and it is checkable in one click.
Meta does define one thing, in Ads Manager itself: creative fatigue as cost per result reaching at least twice its previous level. That is the only defined threshold in existence and it is about cost, not frequency.
Creative benchmarks four: the 47% figure measures something else
This figure is real, well-sourced, and measures something other than what it gets used for.
It comes from Nielsen Catalina Solutions, across roughly 500 campaigns, and a later NCSolutions study put it at 49% across nearly 450 campaigns. Both numbers are correct and they describe different studies, which is one source of confusion.
The larger problem is the transfer. That research covers consumer packaged goods brand advertising, measured on short-term sales lift through purchase panels. It is not direct response, not Meta, not Google, and not lead generation.
Quoting it to justify creative investment on a lead generation account is a category error that almost every paid social article commits. The conclusion may still be right. The evidence offered does not support it.
A smaller detail worth knowing: many citations link to a Nielsen page that does not contain the number in its text, only inside a chart. So even people citing it correctly frequently have not read it.
What these four have in common
A pattern, and once you see it you will find it everywhere.
Each sits on a query with real commercial demand. Each has no available authoritative answer. In every case, content filled the gap with something specific-sounding, other content cited it, and within a few years the figure had the texture of established fact.
The mechanism needs nobody to lie. A writer needs a number, finds one, cites the page it was on, and the chain lengthens by one link. Repeat for three years.
That is why the checkable test matters more than the four corrections. Numbers get manufactured wherever demand exceeds evidence, and creative is simply the worst-affected area because so little is published.
How to check a benchmark in three minutes
The method is the same every time and it is not technical.
Click the citation. Then click that page’s citation. Keep going.
Real research terminates in a study with a stated sample size, a date range and a method. You will reach it within two or three steps and it will be obvious when you have.
Invented figures terminate in an article with no source, or in a loop where two pages cite each other, or in a vendor page with a number and nothing else. That is also obvious.
Do this once, on any creative benchmark you currently believe. The experience changes how you read every number afterwards more effectively than being told to be sceptical.
What to use once the creative benchmarks are gone
Comparison against yourself, which is what you actually needed.
Your real question is which of these adverts to keep running, and that is answered by comparing your own creatives on matched placement and period. No external reference point is required for it.
For absolute performance, cost per result is the measure that pays anybody, and it needs no benchmark either. Either the account is producing outcomes at a price the business can afford or it is not.
The published figures were never answering your question. They were answering whether you are normal, which is a less useful thing to know and, in this area, not knowable from what exists.
What to change this week
Three steps.
Pick one benchmark currently in your reporting and follow its citation chain to the end. Whatever you find will be instructive.
Then remove any target that terminates in an unsourced page, and replace it with a comparison against your own previous performance.
Then check whether anything in your reporting quotes the 47% figure or the 85% figure, because both are near-universal and both need the caveat attached.
Hook rate has no published benchmark and thumbstop rate is three different numbers cover two of these in detail, ad creative metrics covers what the denominators do, fund the ad set and read what the algorithm chose is the position underneath, and most cost per lead benchmarks are recycled is the same investigation in a different topic. What I do across paid, SEO and the website treats sourcing as part of the work. To have your reporting audited for invented figures, book a teardown.