What we set out to test
Can machine learning help app businesses efficiently acquire high-value users at scale?
The background
IGG (I Got Games) is a China-based online game developer with over 300 million players worldwide. They are always looking to acquire and grow a valuable user base for their games. They had previously done so by running manually optimized YouTube install campaigns. But manual optimization meant they had to analyse a vast amount of data (such as search, YouTube and Google Play data) across multiple channels to find the right gamers, build the perfect campaigns for them and find the best time and place to serve them.
The release of a new game “Lords Mobile” was a great opportunity to test the potential of Universal App Campaign to promote their new game application; a Machine Learning-powered solution that creates highly customized and targeted ads across multiple platforms, all done automatically and with real-time optimization.
How we set the experiment up
Experiment control and test groups split based on geography. Within the same market, control and test geographies were identified — roughly equal in terms of size and behaviour.
Control group: YouTube install Campaign
- Manually optimized
- On YouTube network only
Test group: Universal App Campaign (UAC)
- Uses Machine Learning to optimize for business goals (eg new app installs or in-app actions) and find your ideal audience
- Across Search, GDN, YouTube network
Solutions we used
What we learned
Universal App Campaigns can find high value app users, by selecting the right events to optimize for and by following creative best practices.
Competitive target Cost Per Acquisition bidding can strengthen Machine Learning predictions.
This case study is part of the Experiment with Google Ads Program.
