What a frequency ceiling actually costs you
Most accounts set one and never revisit it. The number that matters is not the ceiling but how long you sit against it.
Almost every media buyer has a frequency number they consider safe. Three, usually. Sometimes two and a half for cold prospecting, sometimes four if the offer is seasonal and the window is short. The number gets set once, early, in a conversation nobody wrote down, and then it sits in the account like a load-bearing wall that nobody wants to touch.
The problem is not that the number is wrong. It is that a ceiling is a threshold, and a threshold only tells you about the moment you crossed it. It says nothing about what happened next, which is where the money actually goes.
A ceiling is an event. Fatigue is a duration.
Consider two ad sets that both crossed 3.0 last month. The first crossed on a Tuesday, ran at 3.1 for two days, and came back down when a new creative entered rotation. The second crossed on a Tuesday and was still at 3.4 eleven days later, because the new creative was stuck in review and nobody noticed.
A dashboard that alerts on the ceiling fires identically for both. A weekly review catches the first one, which did not need catching, and misses most of the second one, which did. If you are only ever looking at whether the number is over the line, the two situations are indistinguishable, and one of them is the expensive one.
The measure that separates them is time above the line. Not the peak, not the current reading: the integral. How many days, and how far over, and on how much spend.
Working the arithmetic
Take an ad set at $400 a day. Assume acquisition cost degrades roughly linearly with time spent above your ceiling, which is a simplification but a defensible one over a short window, and assume the degradation reaches about 40% by day ten. That is the shape most people describe when they talk about an ad set going stale.
- Two days over, averaging 8% worse acquisition cost: about $64 of the spend buys nothing it would not otherwise have bought.
- Eleven days over, averaging 22% worse: about $970 on the same daily budget.
- The second is fifteen times the first, on an ad set whose alert fired exactly once, on the same day, in both cases.
Those figures are worked from the assumptions in the paragraph above, not measured from an account. The point is the ratio, which holds across a wide range of assumptions, rather than the dollar amounts, which do not.
What to do about it
Keep the ceiling. It is a fine trigger and everyone already understands it. But stop treating the crossing as the finding, and start treating it as the beginning of a clock.
Three changes, in the order they are worth making:
- Record when an ad set crosses, rather than only that it did. A crossing with no timestamp cannot become a duration later.
- Re-read the same ad set on a schedule after it crosses, rather than waiting for the next review. The second reading is what turns an event into a trend.
- Rank by spend above the line, not by frequency. The ad set at 3.1 on $2,000 a day is a bigger problem than the one at 4.6 on $80, and a list sorted by frequency puts them in the wrong order.
None of this requires a new metric or a different platform. It requires reading the same account more than once a week, which is exactly the part of the job that does not scale with attention and does scale with automation.