Prediction in marketing (an intro to the concept)

Posted by on March 10, 2019 · 2 mins read

Prediction is an area of marketing that is getting a lot of attention for its usability in identifying highly profitable channels and building a predictable business. Doing prediction well enables:

  • More efficient capital allocation. Marketers can quickly identify non profitable ways of acquisition and stop those quickly saving money and time during the capital allocation process.
  • More predictable revenue. Marketers and company owners will have a better look into the future revenue of their company.

In order to do prediction thoroughly, it’s important to collect as much data that makes prediction accurate. In marketing, predictable revenue comes from understanding your user base’s demographics and historical spending/usage patterns.

The more spending and usage patterns you can get per user-demographic dimension, the more predictable any new user LTV will be once acquired for that same user-demographic.

As an example, if you have a significant amount of data that shows you have high LTV with customers who are male, the probability that a male produces high value is good. But if you find that males who live in Hawaii produce the majority of that value, then marketers can have more confidence when trying to acquire male users from Hawaii.

Prediction is all about quantifying confidence levels based on the dimensions of demographic data you have.

At a high level the things drive monetization help you predict monetization:

  • Retention - make sure you understand how retention translates to revenue. You can also use the inverse of retention which is churn.
  • Mechanics that drive ARPDAU - how well active users monetize (ARPDAU) and the drivers of that. In an ad revenue based business this could be the types of ad impressions that you’re serving and the decay curve that’s associated with the eCPM of that impression. In an IAP based business this is how many users will pay and the distribution of the whales vs. minnows.