One of the oldest debates in the short history of in-app ads have been what advertisers should be blacklisted by publishers. Many companies have already started using SOOMLA to gain valuable data in support of such decisions as shown in this case study. However, we’ve noticed recently that many publishers face a problem, even when they have the data.
The problem – how do you weigh ad revenue vs. churn from ads
Even when companies have the full data of the eCPM rates paid by each advertiser along side the churn rates, it’s not always enough to reach a complete decision. What’s needed is a formula to weight the pros and cons. In other words, companies want to know what eCPM lift justifies a 1% lift in churn.
For example, let’s consider two advertisers:
- Billionare Casino with eCPM of $17.54 and ad resulted churn 5.2% (from users who clicked the ad, how many haven’t returned)
- WGT Golf with eCPM of $27.27 and ad resulted churn of 18.5%
Who do you think is better? Does the eCPM increase justify the additional churn?
The analysis – revenue lost vs. revenue made
To answer the question, it’s not enough to look at the basic parameters. The basic analysis that needs to be made is how much revenue was lost vs. how much revenue was made. To determine this, we have to first put a value on a lost user. A good place to start is the overall LTV of a user. If the ad is presented to the user in the first days of activity than the overall LTV of the user is pretty close to the value. For users who have been in the game for some time, the value of a lost user would be the future LTV from that point on. It’s important to note that the number could be higher due to users already having an emotional investment in the game but it can also be lower if the game doesn’t have a lot of depth. Right now, we will assume the value for all lost users is the overall LTV. Now that we figured out how much a user is worth we can multiply it by the number of users lost to determine the amount of potential revenue lost. This factors in the churn ratio but also the CTR as the churn ratio is calculated from the clicks. The revenue that was made is given directly by SOOMLA in the advertiser analysis screen.
Going back to our example – the value of a lot user was determined at $1.28:
- Billionare Casino – generated a total of $623 and while their churn was only 5.2%, the number of users churned was 1,509 so potential revenue loss was $1,931 and the net revenue was a loss of $1,308
- WGT Golf – generated a total of $1,573 and only churned 188 users which are worth $240. Net revenue made was $1,333
As you can see, comparison becomes much easier this way. One has a negative impact and the other has a positive one.
Comparing 2 advertisers with positive net revenue by using nCPM
The analysis above does help weed out advertisers with negative contribution, however publishers also wants to be able to compare between advertisers and give more priority to the ones with higher eCPM and low churn. In many cases, there is a need to compare the net revenue of each advertiser on a quantity of 1,000 impressions to determine who the impressions should be given to. This ratio can be called the nCPM / nRPM (net revenue per mile) as opposed to eCPM / eRPM (revenue per mile).
So back to our example:
- WGT Golf – generated a net revenue of $1,333 on 57.7K impressions which makes his nCPM $23.1
Improving the formula
One way to improve this analysis is to have a better understanding of the lost revenue. Some games don’t have the depth to keep users retained for a long time so the loss might be lower while for other games. Also, some of the games only expose users to ads once they predict the potential for IAP revenue is very low. If they are successful in such prediction, the revenue loss from churning such user would be much lower.
Better way to prioritize advertisers
nCPM is a better way to prioritize advertisers than eCPM. However, the tools available to publishers for optimizing are limited to blacklisting. In reality, the task of prioritizing advertisers for the publisher mostly falls on the shoulders of the ad-networks. The ad providers have an algorithm that tries to predicts the eCPM of each ad. In an ideal world, there will be a way for a publisher to add a “toll rate” for each advertiser rather than just blacklisting them. This will allow the ad-networks to prioritize based on nCPM instead of eCPM.