Cool Tool: Decomposition Tree for Ad Monetization

Published by Digital Limbo on

In collaboration with Rubén Barrena Crespo

Decomposition tree: Concept (Part I)

When you want to understand why your Ad Revenues have increased or decreased, it is easy to waste your time (and lose your mind!) diving inside the Data to find explanations.

Why has revenue decreased? Was it because of a change in eCPM or impressions? And why do impressions change? And so on.

This Decomposition Tree tries to summarize the main ad monetization KPIs and determine how revenue is impacted by the variation of other factors.


This model is based on daily average values, since Daily is the most common period to check your KPIs in digital market, but it can be adapted for your own needs.

Starting from Revenues, each box represents the Daily Average Value of the current period (in this example, the whole June ’18) and the variation compared to the previous Month (May ’18). Then each KPI is disaggregated in two new KPIs that explains the variation. As a Summary we have, from top to bottom:

  • Revenues: the most important KPI, it indicates how much money you have earned during the period. It is calculated by the product of number of Impressions and eCPM.


  • Impressions: total number of VideoAds that your Audience consumes every day. It is the result of the relation between VideoAds Watchers and Frequency (VA watchers * Frequency).
    • VideoAds Watchers: total number of users that consumes VideoAds inside your Application (also known as Unique Impressions). Is the result of Daily Active Users and Engagement. (DAU*Engagement)
      • Daily Active Users: the number of users that enter in your Application, no matter if they watch VideoAds or not.
      • Engagement: the % of your Active Users that became VideoAds Watchers.
    • Frequency: average of video watched by a unique user (impressions / VideoAds Watchers)
  • eCPM (effective Cost Per Mille Impressions): the money earned by the publisher for one thousand VideoAds watched in the application. It is affected by Price Correlation and Unknown effects.
    • Price Correlation: it is the direct impact in the eCPM caused by the increase or decrease of the Frequency. It is known that the price that we earned from the first VideoAd seen by a user is higher than its sixth or tenth. Collecting ad networks data, we defined the following curve that explains how the eCPM goes down in its value as the number of VideoAds watched by the same user increases:


In other words, for the first VideoAd watched, we will get the maximum eCPM (100% of the price), for its second VideoAd 75% of the eCPM, and then until reach around 40% of the eCPM (more on this model in this article about eCPM per impression number).

    • Unknown effects: all those variables that affects the eCPM but we can’t control and whose origins are difficult to determine. As the most important we could find:
      • Seasonality: in some moments of the year eCPM can increase, for instance during Christmas or summer when there is more competition across the advertisers (and bids are raised).
      • Apps launched: at least in Gaming sector, the launch of a hit title from a big company can change the price market for some weeks due to its intensive marketing campaigns.
      • VideoAds mix: the quality and price of the VideoAds mix shown in the App would affect the final Average eCPM (waterfall optimization, country fill rate..)
      • Suppliers mix: finally, the share of VideoAds shown by each Ad Network would affect the final eCPM.

In part 2 we will share the excel file that you can fill with your own data and run the analysis for your game or app. Stay tuned!


Rubén has been working in the Strategy Department at Social Point for 3 years and now is working as a Freelance Consultant. With a combination of product, market and customer understanding, he is able to provide meaningful insights to companies and set up the next steps in strategic topics.


Niels SEO · May 26, 2019 at 18:06

very nice post, i certainly love this website, keep on it

Leave a Reply to Niels SEO Cancel reply

Avatar placeholder

Your email address will not be published. Required fields are marked *