Ad frequency: after how many exposures does the effect level off?

Seeing the same ad three times in one break is irritating. For the advertiser, it is mostly a budget question: at what point does one more exposure add nothing?

Campaign reports rarely answer that. They give an average frequency, and an average of 4 can hide half of the target exposed once and a small fraction exposed twenty times.

What average frequency hides

The distribution of exposures is almost always skewed. The heaviest consumers of the medium concentrate a large share of impressions. Capping is meant to limit this, but it usually applies per platform, per device or per identifier. One person reached across three environments can go far beyond the cap you thought you had set.

So the first thing to ask for is the distribution: what share of the target was exposed once, 2 to 3 times, 4 to 6 times, 7 times or more. That table already says more than the average.

Reading the effect by exposure band

A Brand Lift study compares an exposed group with a comparable control group, using the same questionnaire, and reads the difference with its 95% confidence interval. When each respondent's number of exposures is known, the exposed group can be split into bands to see how the difference evolves.

The pattern one hopes to see is a curve that rises and then flattens. The first exposures establish the message, the next ones consolidate it, then the difference stops growing. That inflection point is what matters for the media plan: beyond it, budget would be better spent reaching new people.

The trap: heavily exposed people are not like everyone else

This is where caution is needed. Whether someone is exposed to the campaign can be controlled by the protocol. How many times they are exposed cannot: it depends on each person's behaviour.

People exposed ten times are the heaviest users of the medium. They are often older on TV, younger on some social platforms, and do not necessarily relate to the brand in the same way. If their score is higher, part of the difference comes from who they are, not from what they saw.

Two precautions limit this bias. The first is to compare each band with controls who have the same media consumption profile, not with the control group as a whole. The second is to read the shape of the curve rather than absolute values: a plateau or a decline among the most exposed is a robust signal, even if the exact level of each band is debatable.

Then there is sample size. Splitting a sample into four bands divides precision accordingly. With 150 respondents in a band, the confidence interval is wide, and two neighbouring bands can rarely be told apart. Three solid bands are better than six unreadable ones.

What can and cannot be concluded

One can reasonably conclude that a plateau exists and locate its zone: around 3 to 5 exposures, or rather beyond 8. One can spot a decline among the most exposed, a sign that capping should be tightened.

One cannot derive a universal optimal frequency. The result holds for this creative, this format, this target and this campaign duration. A complex creative needs more repetition than a simple message, and a six-second format does not behave like a thirty-second spot.

In practice

1. Ask for the distribution of exposures, not just the average.

2. Check at what level capping applies: platform, device or person.

3. If an impact study is planned, decide on frequency bands before the campaign, with enough respondents in each.

4. Read the shape of the curve, and be wary of differences between neighbouring bands.

5. Reinvest in reach what you take out of repetition, then measure again.

The right frequency is not a number to apply. It is a hypothesis to test, campaign after campaign.