Mamba is a popular dating app with an audience of more than 40 million users from more than 50 different countries. The app is free and generates revenue from ad placement.
Experiment details
The app developers ran a two-week test to compare the classic waterfall model of mobile ad mediation with Mobile Mediation from Yandex with automatic threshold selection.
The key goal of the experiment was to compare how effective these models were for selecting the best monetizers and ultimately increase revenue from ad blocks in the mobile app. The experiment was run on native banners in the app. Half of the requests to demand platforms were based on the waterfall model, and the other half used automatic threshold selection. Monetization from the Yandex Advertising Network and AdMob was integrated into the app via Mobile Mediation. In order to also serve ads from other demand platforms, the IDs of blocks in those platforms were added in the Mobile Mediation interface. Then the latest version of Yandex Mobile Ads was integrated into the app and mediation was launched. The experiment lasted about two weeks.
Results
The results showed that in all cases, regardless of where the ad block was shown in the app, the revenue from impressions with Mobile Mediation was higher than the revenue from the classic waterfall approach. Advertising revenue for all ad blocks with Mobile Mediation increased by 30% on average, with an average fill rate of 75%. The ad block in the photo carousel showed the largest increase in revenue — 55%.
How Mobile Mediation works
Automatic threshold selection helps maximize revenue from in-app ads. It uses machine learning to predict which demand platform will win the auction and sends the request to them first. The algorithm is able to predict which platform will offer the highest bid for an impression to a particular user by analyzing the history of rendering statistics, the device type, the user's behavior and location, and hundreds of other factors. As a result, fewer requests are sent out to demand platforms and the ad can be served faster.
The ad host gets the maximum possible revenue, and the advertiser gets higher quality traffic and a relevant audience. It took about two weeks of training for the algorithm to reach peak efficiency in the Mamba app.