Ad frequency thresholds are among the most confidently repeated numbers in paid social, and not one of them has a source. Meta studied this directly, across thousands of campaigns, and made a deliberate decision not to publish a threshold. Everybody else publishes several.
That contrast is the whole subject, and it takes one click to verify.
This is for you if somebody has told you your frequency is too high and you want to know against what standard.
What ad frequency measures
Impressions divided by reach. The average number of times each person saw your advert.
A frequency of three means the typical person in the audience was shown it three times over whatever period you are looking at. That last part matters and gets ignored constantly, because frequency accumulates.
A campaign at frequency two after one week is at a different place from one at frequency two after two months. Same number, completely different situation, and most reporting shows the figure without the period attached.
So the first honest thing to say about frequency is that a bare number carries less information than it appears to.
The study that declined to name a threshold
Facebook analysed frequency across 2,439 Brand Lift campaigns between November 2019 and March 2020.
The analysis confirmed what you would expect: effectiveness plateaus after a handful of impressions. Additional exposures keep costing money and stop adding much.
Then it published no threshold. No cap, no danger zone, no number.
That was a choice, made by the organisation with the most data on the question. The plausible reason is that a single figure would be wrong for most advertisers, because the right level depends on audience size, purchase cycle, creative variety and objective.
Meanwhile the industry supplies plenty. Above 2.5 means fatigue. Cap at three. Refresh at four. None of those trace to Meta or to any study with a disclosed method.
Why a universal ad frequency threshold cannot exist
Think about what such a number would have to survive.
A local trade serving three postcodes has a small audience by necessity. Frequency climbs quickly and that is the structure of the business rather than a mistake. A national ecommerce brand has millions of people available and a frequency of three would mean something entirely different.
Purchase cycle matters too. Somebody buying a sofa needs more exposures across a longer period than somebody buying a takeaway.
Creative variety changes it again. Frequency four across eight different adverts is a very different experience from frequency four on one advert repeated.
Any single threshold would be wrong for most of those cases on the day it was published. That is probably why the organisation with the data left it alone.
The arithmetic trap inside ad frequency
There is a mechanical effect here that produces false diagnoses constantly.
As frequency rises, impressions grow while reach flattens. So every metric divided by impressions falls automatically, whether or not anything about the creative changed.
Click-through rate declines. Any view rate declines. The report looks like decay.
Some of it usually is decay. Some of it is pure arithmetic from a growing denominator, and nothing in the interface separates the two.
The check is to look at cost per result alongside. If frequency rose and cost per result held steady, the falling rates are the denominator rather than the creative. If cost per result rose too, something real is happening.
That two-minute check prevents a lot of unnecessary creative replacement.
A real profile, without a verdict attached
Since no benchmark exists, here is one campaign described rather than graded.
A campaign on a US flag and banner ecommerce account ran at frequency 3.30, with a click-through rate of 2.03%, a return on ad spend of 1.12, and $87.75 per purchase.
That is above every threshold circulating online. Whether it is too high depends on what the account needs, what the alternatives are and what the margin supports, none of which a frequency number can tell you.
I am describing it rather than presenting it as a benchmark, because one campaign is not a benchmark and I would rather be useful than authoritative. It is one real profile at a frequency the internet would call excessive, still producing purchases.
What to do instead of watching the number
Two things, and neither involves a threshold.
Watch cost per result against its own history. That is the measure attached to money, and it is the one Meta uses in its only defined fatigue label, which triggers when cost per result reaches twice its previous level.
Then watch frequency as context for that cost, not as a target in itself. Rising cost with rising frequency suggests exhaustion. Rising cost with flat frequency suggests something else entirely, and replacing creative will not fix it.
Used that way, frequency is a useful diagnostic. Used as a number to keep below, it drives decisions nobody can justify.
When frequency capping makes sense
Rarely on conversion campaigns, more often on awareness.
On a conversion objective, a cap constrains delivery in exchange for a benefit nobody has quantified. The system is already balancing exposure against response, and overriding that with an unsourced number is unlikely to improve it.
On awareness activity, where reach is genuinely the goal, capping has a clearer logic. You are explicitly buying breadth rather than depth, so limiting depth serves the objective.
The better lever in both cases is creative count. Adding distinct adverts spreads exposure without narrowing your audience, so people see more variety at the same frequency. That addresses the actual problem rather than the number describing it.
Why ad frequency reads differently on small audiences
One situation deserves separating out, because the standard advice actively misleads there.
A business serving a small geography has a small audience by necessity. Frequency climbs quickly and there is nothing to be done about it, because the alternative is advertising to people who cannot buy.
Applying a general threshold to that account produces a bad recommendation: widen the targeting to bring the number down. That trades qualified reach for unqualified reach in order to improve a metric nobody can source.
I would rather run a tight audience at frequency four, producing results, than a wide one at frequency two producing enquiries from outside the service area. The frequency number looks worse and the business does better.
So before acting on a frequency reading, ask whether your audience could reasonably be larger. If it could not, the number is describing your market rather than a problem with your advertising.
What to change this week
Three steps.
Put frequency next to cost per result in your reporting, with the period stated, so the number carries the context it needs.
Then check whether any threshold in your process has a source. If it is 2.5, three, or a refresh interval in weeks, it does not.
Then, when frequency climbs, try adding creatives before widening the audience, because it fixes exposure without costing you targeting precision.
Fund the ad set and read what the algorithm chose is the position underneath all of this, creative fatigue has one real definition covers the only defined threshold that exists, creative benchmarks are mostly invented covers the wider pattern, ad creative metrics covers why denominators mislead, and scaling Facebook ads covers what happens to frequency as budget grows. The equivalent invented thresholds turn up on search too, which Google Ads management has to deal with constantly. To have an account assessed on evidence rather than folklore, book a teardown.