The Case Against eCPM as a KPI

When reviewing metrics to ascertain performance as a publisher, there are several options. One of the most common is to look at eCPM. However, eCPM is a flawed metric from a publisher performance perspective when comparing how one platform performs against another. In this article, we attempt to explain why, using real-world examples.

When a publisher should ignore eCPM as a KPI

Consider the following two scenarios: [table id=7 /] In these examples, we are A/B testing two monetisation solutions on a page visited one million times per month.

Solution A has 3 ad slots served on the page, which results in 3M ad script loads. Solution B has 2 ads per page, which results in 2M ad script loads.
Solution A does not refresh the ads. It also filters out what it deems invalid ad calls from its reporting count, giving us a 2.8M ad calls/pageviews/ad requests/impressions (the naming of these metrics differ wildly from platform to platform).
Solution B refreshes ads 3 times on average every time the page loads and—unlike solution A, it reports all ad calls, valid or invalid, giving us a 6M ad requests number.

In these examples, solution A generated $5,000 at the end of the month and its eCPM is $1.75. Solution B generated $8,000 at the end of the month and its eCPM is $1.33. If we were to refer to eCPMs at the KPI, which are based purely on platform count, solution A is clearly performing better. However solution B has generated 40% more revenue out of the same traffic—an ultimately more successful result. Knowing this, is eCPM really the metric we should be looking at? At the end of the day, solution B generated more revenue for each visit. It did so by using less screen real estate (2 ads vs. 3). There are really two things we can trust to be accurate: On one hand, we have the number of users interacting with the site. On the other, the revenue generated by advertising. Whatever sits in the middle means very little in such a comparison.

SO I SHOULD NEVER USE ECPM?

Is that to say eCPM should never be used as a metric? Absolutely not. eCPM would be relevant if, for example, you were comparing two ad units belonging to the same setup. Or performance of a campaign you were delivering on your inventory. However, if comparing performance between one advertising vendor or platform against another, there are too many variables which affect the output to make eCPM a meaningful metric for comparison.