Search
Clear search
Close search
Google apps
Main menu

Understanding eCPM in Ad Exchange

The Query Tool offers three eCPM metrics to help with your business needs. Each metric is derived from a different stage of the ad delivery workflow. 

Introduction to Impressions

To help you better understand the distinction between the three Ad Exchange eCPMs, this article explains the applications of each metric as well as what the different denominator definitions can tell you.

  • First, here is the definition of CPM = Revenue per one thousand impressions
  • Second, here is the process behind how an ad is requested from an ad server or network:
    1. A website or app puts in a request for a creative. 
    2. A creative is then matched and served back. 
    3. That ad must be rendered--this is the process of the creative being drawn, but not necessarily shown to the user.
    4. Finally, an ad is displayed to a user. 
Show me the visual representation
Ad request metrics excluding text ads

As shown in the illustration,throughout the process of ad serving, fewer impressions are seen than queries are made. There is a drop off at each step of the process. Why?

  • Not every request for an ad will find a campaign that matches.
  • Not all ads that are matched and served are ultimately rendered, and even fewer are seen by a user.
But any one of these points in the process is used to calculate a CPM. The further you go down the chain when you select a CPM denominator, the higher the CPM.

Definition of the Ad Exchange eCPMs 

Metric Definition Formula
Ad requests  Overall number of Ad requests made to Ad Exchange, but not necessarily filled by Ad Exchange. Ad request eCPM = Revenue / Ad requests
Matched requests  Number of requests where a buyer was matched with the Ad request and purchased the inventory. This metric displays the CPMs buyers are willing to pay.
If there are rendering issues, Matched Requests > Impressions.
eCPM (default) = Revenue / Matched requests
Ad impressions NEW
 

Ad impressions often delay counting an impression to until the ad is rendered or later. This metric shows the resulting CPM that buyers paid per Ad Exchange Ad impression.

Prior to early 2016, ad impressions counted each text ad in an ad slot as an Ad Impression. We're changing this behavior to only count the ad slot once, deprecating the prior definition of Ad eCPM (legacy), and adding Ad eCPM NEW.  Here is a summary of changes to eCPM.
The Ad impressions metric is the most relevant metric for evaluating floors (excluding text ads).
Ad eCPM NEW= Revenue / Ad impressions NEW

When to use which eCPM metric

Each of these metrics serves a different purpose.

  • Ad Request eCPM: Revenue /Request Queries
    Represents the revenue you receive per request you made for a creative; the calculation factors in fill rate. So ad networks that may have high CPM ads, but offer low fill or whose ads are unlikely to convert are disadvantaged.
     
  • Matched eCPM: Revenue/ Matched Queries
    Matched eCPM removes the impact of the fill rate. While fill rate is ignored, a network’s CPM is punished if many of their ads are high CPM, but less likely to pay--for example a video that only pays after a second of play time. This metric also helps to standardize for different networks counting impressions and paying at different points in the ad serving process.
     
  • Ad eCPM: Revenue/ Ad Impressions
    Ad eCPM removes the impact on the denominator of ads that are served, but do not result in impressions and may not generate revenue. As more networks count impressions and generate revenue at points after the creative is served, this measure of CPM is becoming more common. A version of this CPM metric is used by many other networks. This is one reason that it’s now the default eCPM metric in Ad Exchange.

By looking across eCPM metrics you can begin to understand the impact of a network’s fill rate, its creative latency, and the attractiveness of its ads on your revenue.

eCPM troubleshooting

Learn about why effective CPM could be lower than your Min CPM.
Learn about discrepancies in reporting impressions.

Was this article helpful?
How can we improve it?