+6 votes
I've heard multiple ways to calculate eCPM decay. One is through analyzing lifetime, the other is through analyzing lifetime session. Which is the best approach?
by (1.6k points)

1 Answer

+6 votes
Best answer

Some time ago I ran an analysis based on feedback from several video ad networks and it turned out that eCPM goes down a lot after the first 4 impressions.

Looking at the trend, if the eCPM of the first impression is 100%, you can see how much will it decrease when the same user watches several videos:

eCPM per impression number


From the advertiser perspective, the more videos the same user watches, the lower her/his conversion rate is (they will be less likely to click on an ad and actually install the advertised app).

My advise in general would be to focus on the number of users who watch videos (so called engagement) rather than on the number of videos watched by the same users ( known as frequency).

by (890 points)
selected by
Yes, but how do you measure this tradeoff? That seems mostly like a hunch that you have. Do you know in fact this method is more profitable?
There are many more factors to take into account:
- game genre
- IAP vs IAA share
- retention
This is a basic model, then everything needs to be tested and reiterated based on the goals of ad monetization. I hope it helped!
I did some research and it seems like there is a different ecpm decay curve for each x-day retained cohort. Do you find something similar with your study?
Can you share your findings? I'm curious to see more data on this, my study might be a bit outdated but if you think from an UA point of view I think it's still valid with some premises, as the CR of a campaign decreases the more videos are watched. Then of course it depends also on the geo/platform and the ad network mix in the stack
Not sure how you'd like me to share directly, but the intuition is that ecpm decay varies by the x-day retained day.