Forecasting Frequency Capped Impressions

Published by: J Zilberbrand , Mediavest Worldwide
Published on: October 14, 2010

The question of being able to forecast both total and available impressions with a frequency cap has been coming up a lot lately. I've heard from a number of sources that an accurate prediction of both total and available impressions is mathematically not possible. The reason being that it isn't possible to know if the person you're delivering an impression to will consume any number of additional impressions.

You can estimate the total unique user reach but the total number of impressions delivered will always be some number less than the possible reach multiplied by frequency cap.

In other words, with a 100 user reach and a frequency cap of 3, the total impressions served will be somewhere between 100 and 300.

One best practice is that these campaigns should run at the highest possible priority in the ad server in order to maximize reach. However, even in that scenario, a campaign won't deliver reach * frequency cap number of impressions.

Further complicating this is any competing frequency capped campaign. Only one campaign can reach a frequency of one user and that user is removed from the possible delivery pool after it is delivered.

All that said, I've heard a couple people recently mention certain ad serving vendors are saying that they are able to forecast frequency capped impressions. I'd love to know if that is indeed true and if so, how accurate this is. How are publishers seeing the results? Working? Not Working?

If someone has solved this riddle, would be great to hear about it and what results are like in the field.


Rocket Fuel