What is Frequency?

A campaign delivers 840,000 impressions to 210,000 people, so average frequency is 840,000 ÷ 210,000 = 4.0. Breaking the audience apart, half of them saw it once or twice, a third saw it three to six times, and the remaining sixth saw it around ten times. The mean of 4.0 describes almost nobody, and the sixth at the top is where the complaints and the falling click-through rate are coming from.

Frequency divides impressions by reach. It answers how many times the average person in the reached audience was served the ad.

Frequency = impressions ÷ reach

The three quantities are one identity, and any two of them fix the third. That makes frequency the diagnostic of the set: it is where the relationship between what was bought and who it landed on becomes visible. A campaign can report excellent impression volume and a healthy price while doing nothing except showing a small group the same ad repeatedly, and frequency is the number that says so.

Exposure is distributed very unevenly, and the mean sits in a part of the distribution where few people actually are.

The mechanism is straightforward. Platforms serve ads against available inventory, and heavy users generate far more inventory than light ones. Somebody who opens an app throughout the day presents many times the opportunities of somebody who opens it once. Targeting does not change this; it selects who is eligible, not how often they appear.

The result is a long right tail. A campaign sitting at a comfortable average shows most of its audience one or two ads and a minority a great many. Every consequence people associate with over-exposure lives in that tail: the falling response rate, the rising cost, the hidden-ad and report-ad actions on social placements.

Where the platform offers a frequency distribution rather than an average, it is the report worth opening. It changes the question from "is four too high" to "what do I do about the slice of the audience seeing this a dozen times".

Nobody sets out to over-expose an audience. It happens because of how delivery interacts with a finite audience.

Early in a flight, most impressions go to people who have not been reached, so reach grows and frequency stays low. As the addressable audience fills up, a rising share of impressions necessarily go to people already reached. Reach flattens; frequency climbs; spend continues.

This is arithmetic rather than a setting, and it means every campaign on a fixed audience has a frequency trajectory built into it from the start. The only variables are how large the audience is and how fast the budget is spent against it.

Raising the budget on a saturated audience does not buy more people. It buys more repetition, faster.

Repetition is not a defect. For most messages it is a requirement: recognition, recall and the erosion of hesitation all take more than one exposure, and a campaign capped at one is frequently a campaign that never lands.

The cost appears when exposure passes the point at which the message has done its work. Beyond that point, additional impressions produce diminishing response and eventually negative response — irritation, hidden ads, reports, and on some platforms a quality penalty that raises the price of subsequent delivery.

Where that point sits varies with almost everything: how complex the message is, how much it asks for, how distinctive the creative is, how long the flight runs, and how warm the audience already was. Which is why an imported rule of thumb is a poor instrument and the account's own curve is a good one.

The observable signal is the click-through rate over the flight against the frequency curve. Where response turns down while frequency continues up, the campaign has passed its point. That turn is specific to the campaign and it is visible in the account's own data within a week or two of a flight starting.

A single frequency target applied across an account misprices both ends of it.

Prospecting runs against large audiences and should keep frequency low, because the job is to add people and repetition is a poor substitute for reach.

Retargeting runs against small audiences by construction — site visitors, cart abandoners, a customer list — and the frequency will be high whatever the settings say. A remarketing pool of eight thousand people receiving a modest daily budget will produce double-digit frequency within a fortnight. The relevant controls there are the membership window, the exclusion of converters, and the size of the daily budget, not a frequency setting.

Judging both against one number leads to capping the campaign that needed repetition and ignoring the one that was drowning in it.

Widen before capping. More people is the only intervention that lowers frequency without reducing delivery. A frequency cap on a saturated campaign trades repetition for unspent budget, which is sometimes the right trade and is always a trade.

Rotate creative. Fatigue attaches to the execution more than to the message. A new creative against the same audience resets the response curve, which is why a rotation schedule set to the account's own turning point is worth more than a cap.

Exclude people who converted. Continuing to show an ad to somebody who has already bought is the most avoidable form of over-exposure and one of the most common.

Shorten membership windows on retargeting. A ninety-day window with a steady budget guarantees very high frequency against the people who joined it first.

Pace the budget to the flight. Spending a month's budget in a week against a fixed audience produces a month's worth of frequency in a week, with none of the reach.

• Read the distribution, not the average. A mean of four can hide a tail seeing the ad thirty times, and the tail is where the fatigue and the complaints are.
• Widen the audience before capping the frequency. A cap on a small audience suppresses delivery; a larger audience lowers repetition without throwing away impressions.
• Rotate creative on the schedule the frequency curve suggests rather than on a calendar. The point to change is where the click-through rate turns, and that point moves with audience size.
• Expect retargeting frequency to run much higher than prospecting and judge the two separately. A single account-wide target misprices both.
• Track frequency over the flight rather than at the end. The end-of-campaign average is the least informative version of this number.