Atlas Solutions

Finding Optimal Ad Frequency in a People-Based World

by James Dailey, Head of Atlas Marketing Sciences

A typical person sees hundreds of ads each day. Show your ad only once, and they may not remember it. Show an ad hundreds of times, and people will likely get frustrated and annoyed. Managing ad frequency improves yield and enables marketers to redeploy impressions and broaden the reach of their marketing messages. Finding the ideal ad frequency depends on having accurate information. Unfortunately, cookie-based measurement complicates this issue instead of clarifying it. If marketers want to find the optimal ad frequency to drive their conversions efficiently, they need to rely on people-based measurement instead.

Because cookies are unable to keep up with the cross-everything habits of real people, they often provide a distorted view of reach and frequency. When relying on traditional cookie-based measurement alone, a typical Atlas client sees their reach overstated by 58%, and their ad frequency understated by 141%.1 So how does this translate in the real world? It means that if cookie-based measurement tells you your ad reached 100 people, in reality it only reached 42. And if you originally set a frequency cap at 10 ads per cookie, it’s more likely a real person would’ve actually been exposed to 241.

We can see a further example of this by looking at a frequency histogram — a depiction of numerical data in graphical form. What follows is Atlas’ analysis of a DR client’s digital ad serving activity in May 20152:



Looking at the chart, we see that 71% of cookies had an ad frequency of one (orange) vs. 52% of people (blue). Many people were exposed to two, three, four or more ads, but some of an individual cookie only captured one ad impression during the same time period. For this reason, the cookie-based reporting is artificially “shifted left,” and does not accurately depict the ad frequency experience of real people.

But what are the practical implications of this observation for setting frequency caps? And how does this information distort our understanding of the relationship between ad frequency and conversion? Pulled from the same dataset, the next chart shows the conversion rates for people and cookies when exposed to different ad frequencies:



This chart tells a similar story as the first histogram, only now we see conversion rates as well. Note how the cookie-based frequency curve (orange) is much flatter and shows higher conversion rates at low frequencies. The people-based line (blue) is steeper and shows a more pronounced leveling-off pattern above 15 impressions.

With cookie-based reporting (orange), a portion of the population is artificially “pulled left,” and we’re given a lower frequency than what people truly experienced. As a result, the relationship between frequency and conversion is obscured and the curve appears flatter than it should be. Looking at cookies, you’d think the difference in conversion between showing 5 ads and 10 ads isn’t very much. The conversion data are still accurate, but the overall depiction is not — the people who made purchases at the “lower end” of the orange curve were actually exposed to higher ad frequencies than cookies could capture.

In contrast, the people-based curve (blue) provides a more complete picture and gives us a better understanding of real people and their behavior. This is because it captures ad impressions served across browsers and devices over time. We can now see that there’s a meaningful increase in conversion performance between people who saw 5 ads and those who saw 10. We know we can rely on this trend line to help us set frequency caps in the future.

When marketers focus on accurate people-based reporting, the relationship between frequency and conversion becomes clearer and reflects traditional expectations. Now we’re able to start exploring the question of ideal ad frequency and help steer our brand away from overexposure and annoyance. This is one more way people-based marketing can help marketers adapt to real audience behavior and find value-based insights for their campaigns.


1. Source: Atlas internal data, March 2015
2. Source: Atlas internal data, May 2015