Originally posted on Android Developer blog
Posted by Lily Sheringham, Developer Marketing at Google Play
Editor’s note: This is another post in our series featuring tips from developers finding success on Google Play. We recently spoke to games developer Kongregate, to find out how they use Store Listing Experiments successfully. - Ed.
With Store Listing Experiments in the Google Play Developer Console, you can conduct A/B tests on the content of your store listing pages. Test versions of the text and graphics to see which ones perform best, based on install data.
Kongregate increases installs by 45 percent with Store Listing Experiments
Founded in 2006 by brother and sister Jim and Emily Greer, Kongregate is a leading mobile games publisher specializing in free to play games. Kongregate used Store Listing Experiments to test new content for the Global Assault listing page on Google Play. By testing with different audience sizes, they found a new icon that drove 92 percent more installs, while variant screenshots achieved an impressive 14 percent improvement. By picking the icons, screenshots, and text descriptions that were the most sticky with users, Kongregate saw installs increase by 45 percent on the improved page.
Kongregate’s Mike Gordon, VP of Publishing; Peter Eykemans, Senior Producer; and Tammy Levy, Director of Product for Mobile Games, talk about how to successfully optimise mobile game listings with Store Listing Experiments.
Kongregate’s tips for success with Store Listing Experiments
Jeff Gurian, Sr. Director of Marketing at Kongregate also shares his do’s and don’ts on how to use experiments to convert more of your visitors, thereby increasing installs. Check them out below:
|Do start by testing your game’s icon. Icons can have the greatest impact (positive or negative) on installs — so test early!||Don’t test too many variables at once. It makes it harder to determine what drove results. The more variables you test, the more installs (and time) you’ll need to identify a winner.|
|Do have a question or objective in mind when designing an experiment. For example, does artwork visualizing gameplay drive more installs than artwork that doesn’t?||Don’t test artwork only. Also test screenshot ordering, videos, and text to find what combinations increase installs.|
|Do run experiments long enough to achieve statistical significance. How long it takes to get a result can vary due to changes in traffic sources, location of users, and other factors during testing.||Don’t target too small an audience with your experiment variants. The more users you expose to your variants, the more data you collect, the faster you get results!|
|Do pay attention to the banner, which tells you if your experiment is still “in progress.” When it has collected enough data, the banner will clearly tell you which variant won or if it was a tie.||Don’t interpret a test where the control attribute performs better than variants as a waste. You can still learn valuable lessons from what “didn’t work.” Iterate and try again!|