Why eCPM sucks and what to measure instead?

The ultimate goal of online advertising from a publishers angle is a certain level of revenues, “certain” meaning at least high enough to cover the costs and secure a safe margin. While it may sound obvious, it is not always that easy to tell the impact of changes like adding new tech partners, delivery rules etc. on the final numbers. While online business has always been in love with measurement and hooked on many different metrics, picking the right one might be a little tricky.

The eCPM trap

eCPM is not a great proxy of total revenues. Full stop. In order to understand why, let’s analyse the following example. Publisher X manages around 30 000 000 of ad requests a month and imposes a strict floor price at the level of 5 EUR. With such a rule they manage to sell around 20% of their inventory, what results in 30 000 EUR monthly revenues. But then they decide to ease their floor pricing policy and remove them at all. Their eCPM falls down to 2.5 EUR, but their fill rate raises to 95%, i.e. vast majority of inventory is sold. So even though they trade it 50% cheaper, their total monthly revenue hits 71250 EUR, ie. more than twice as big as before. The summary is presented in the table below.


The eCPM is a misleading metric as it takes into consideration only the impressions that were sold. And every unsold impressions represents a lost opportunity cost which finally affects the real value of publishers inventory. The higher the eCPM, the lower the fill rate and any attempts toward yield optimization should be made with awareness of this relationship. It also means that it does not make any sense to optimize either against eCPM or fill rate as the holy grail lays in an optimal balance between the two.

What’s the remedy?

As it was mentioned before, a reasonable approach to measurement should include both sold and unsold impression. The sum of two represents the total number of ad requests available on publishers properties. When you calculate the eCPM based on ad requests (let’s call it ad request CPM), things start to look different. Let’s get back to the previous example. In the first scenario ad request CPM is just one 1 EUR, whereas in the second one it raises up to 2.38 EUR. So while the eCPM does not necessarily have to be correlated with total revenues, the ad request CPM does.

Alternatively, one can measure so called RPM (Revenue Per Mille), which is usually translated to the revenue from thousand page views (not ad impressions). Since the number of page views represents the whole inventory, not only its part that publisher managed to sell, this metrics can also be considered a good proxy of final ad revenues.

Is yield optimization really that simple?

Theoretically, yes. In practice, it always depends on pricing policy, strength of direct relationships with advertisers and purchasing power of agencies. Still, these two very simple and extremely useful metrics can help when it comes to estimating the outcome of either strategic decisions or everyday operational tasks. One thing is certain, though, and it is that eCPM should not be taken as an ultimate reference point by any means.

Still hungry for more?
E-mail us at [email protected]

No comments yet