In today’s mobile app ecosystem, ad mediation became the standard for any app company who is doing more than $10K a month from in-app advertising. This makes perfect sense since different ad-networks have strengths and weaknesses when it comes to eCPM and fill rates. For example one ad network might have great campaigns in Russia but underperform in India and vice versa.
There are many problems with the way ad mediation is done today. Here are 3 of them:
Country level segments are not granular enough
The basic function of the ad mediation is to allow you to create a waterfall of ad-networks. Basically this means that you can select which ad-network gets the 1st chance, which one gets the 2nd chance if the first one doesn’t have an ad, which one gets the 3rd chance and so on. This waterfall can be configured differently for different countries. However, ad mediation platform do not allow configuring waterfalls for more advanced segments. For example, you might want to have a different waterfall for males vs. females – there is no way to do that with current mediation solutions.
All ad mediation platforms except for 2 are owned by ad networks
The mediation platform’s role is to be an unbiased 3rd party that works in the service of the publisher to maximize his advertising yield. The selection of what ad-network to show first should be guided only by the interest of the publisher. In reality however, all ad mediation platforms are owned by ad networks. This makes it hard for them to be objective about their decisions and in many cases you can see that the mediation is making sub-optimal setups from the publisher’s perspective.
Automatic optimization mode relays on yesterday’s data
Automatic optimization mode is the default choice for most mediation platforms. The idea is pretty cool – if there was a way to know how much an ad network is going to pay for certain impression, the auto-pilot mode of the mediation could easily decide which network should get the impression and give it to the highest bidder. In reality however, the data that the ad mediation platform has to make that decision is insufficient. It is only able to extract yesterday’s eCPM per country from each network. Since the eCPM is based on how many users ended up converting to installs for the advertisers, the eCPM rates fluctuates quite a lot. The mediation auto-pilot ends up with this cycle:
- Good network is on top and bad network is receiving less impressions
- Bad network eCPM fluctuates more due to increased randomness in small sample sizes
- Every few days the bad network gets “good day” and finds itself on top the day after
In most cases, you will see that the auto-pilot makes a bad decision like that every 4-5 days. Bad days resulting in a loss of about 25% of the revenue that day.
If your app is using a mediation platform with multiple ad networks you should probably start attributing your ad revenue. Check out SOOMLA Traceback – Ad LTV as a Service.