A frequency cap looks like the simplest setting in a media plan: show this advertisement no more than three times a week. The number is easy to enter and far harder to enforce.
What a cap actually counts
A cap counts impressions against an identifier, not against a human being. That identifier is usually a cookie, an entry in a device graph, or a logged-in account, depending on the platform.
Each of those is a proxy standing in for a person. Someone with a phone, a laptop and a television app can appear as three separate identifiers, each carrying its own untouched allowance.
The cap is therefore honoured exactly as it was specified and still fails at the job it was bought to do, which is limiting how often one person sees the same message.
Identifier resets quietly restart the counter
Browser restrictions have shortened the working life of many identifiers, and an identifier that disappears takes its impression history with it. The counter starts again from zero against what looks like a new person.
This produces a pattern advertisers recognise without being able to explain it. Reported frequency stays neatly inside the cap while complaints about repetition rise steadily through the campaign.
The reporting is not lying. It is measuring the only thing available to it, which is the number of impressions delivered per surviving identifier.
Caps do not travel between platforms
Each platform enforces its own cap against its own identifier space. A campaign running across three environments has three independent counters that never compare notes.
A person who uses all three can legitimately receive the full allowance from each. The total exposure is the sum, which is often several times what the planner intended.
Cross-platform capping requires a shared identity layer that every seller agrees to honour, and the commercial incentives to build one are weaker than the incentives to sell more impressions.
Capping too tightly has its own cost
Tight caps suppress delivery. Once the cap is reached for the reachable audience, the campaign either stops spending or reaches further into audiences with weaker intent.
That second outcome is the more common one, and it is usually invisible in the reporting. Spend stays level while the quality of who is being reached falls.
The trade is between wasting money on repetition and wasting money on strangers, and the right balance depends on how much persuasion the product actually needs.
What planners do about it
The practical response is to treat reported frequency as a floor rather than a measurement. Actual exposure is assumed to be higher, often substantially so, and plans are built with that assumption.
Some teams sample directly instead, asking a panel how often they recall seeing a campaign. The answer is imprecise but it measures people rather than identifiers.
Neither approach fixes the underlying problem. They simply stop the plan from depending on a number that describes the plumbing rather than the audience.