Why use reach and frequency instead of impressions in Marketing Mix Models?

With MMM, one expects to obtain insights on media contributions, saturations and help in future decision making. To do that, the model must estimate the target KPI (sales, number of leads, …) for different scenarios. Using only impressions to describe a campaign will reduce the comprehension of the strategy and lead to bad decisions. Here are some examples where an impressions-only model will fail:

• if a campaign/touchpoint (media channel) took place only in a few geos and we decided to generalize the campaign/touchpoint to the whole country, the impressions and impact should linearly increase. Classic MMMs will apply the saturation they learned on the geos and thus return low ROI. With a reach and frequency approach, we can easily generalize and increase the reach.
• if past campaigns on a touchpoint reached almost their maximal reach, increasing the budget will only increase the frequency and may not bring additional revenue. This form of saturation may not be captured by classic MMM if the reach saturation point has not been observed by the model.
• if past campaigns on a touchpoint reached only a small fraction of its maximal reach, increasing the budget will increase the reach and thus increase the budget linearly. Classic MMM are built to learn a saturation response which will prevent from grasping the full potential of the media channel.

In order to leverage reach and frequency, one can use the model developed by Google in July 2023, which is available in fifty-five MMM solutions or through Meridian. At fifty-five, we also rely on Agent Based Models (ABM) to deal with reach and frequency. In ABM, we simulate fictive consumers (agents) and their behavior on the market based on the marketing strategy they are exposed to. Working at the consumer level enables us to natively integrate reach, maximal reach and frequency because, in ABM, each individual consumer will be exposed to an ad a certain number of times. We can also integrate MarCom target features in order to get insights and make decisions at this level of granularity. Finally, we learn the relationship between budget and reach+frequency to forecast future scenarios as accurately as possible. Provided that available data are granular enough, with these models you will not only know if a touchpoint is saturated or not, but also why. 

‍

‍

*Reach: Among the targeted population, the percentage of people that have seen the campaign at least once.

*Frequency: Among reached people, average number of times a user has seen the ad.

*Impressions = Reach x Frequency x Targeted population size.