Unstable Android apps punished with poor Google Play placement

Google plans to punish crash-prone Android apps with worse rankings in the Google Play Store, in an attempt to force developers to write more stable code. The move will see Google wield some machine intelligence along with human assessors onto the vast quantity of titles available in the Google Play download store, following persistent complaints by Android device users that much of what's in there can be flaky and dissatisfying. Just as in search, Google plans to hit the developers responsible where it hurts the most: visibility.

By looking at things like user ratings and engagement, as well as performance data, Google's new AIs will rank Android apps according to how stable and reliable they are. That rating will then be used to help decide how high up in the search results that app shows. Efficient, stable titles will get good visibility and be near the top of the list; those which aren't will be downranked.

The new system of stability-related ranking is the latest step in Google's attempts to clean up its download store. A purge back in February cleared out apps with over-reaching permissions policies, for instance, and more recently Google said it would use machine learning to identify titles that were too demanding in terms of performance or access to device hardware.

It's unclear, though, whether this will address the big complaint many have about Android: that Google simply doesn't take the time to vet the titles allowed through into the store. Where Apple assesses every title submitted to the App Store before it's available for iPhone and iPad users to download, Google takes a more retroactive approach to filtering, only taking action once software has already been made available to download. Though that has spurred plenty of submissions to the store, it has also made it more likely that malware, apps that steal personal details, or software that tries to access premium services in the background have made it through to users' phones.

Using machine learning to automate the process as much as possible is typical Google behavior, mind. The company already does something similar with search results, taking into account factors like page load speed, mobile readability, and the presence of pop-up adverts and other elements that go against easy reading. AMP pages, which have been pared back in the name of speedier load times, also get SEO priority.

Initially, Google will test the new Google Play ranking algorithms on a small subset of users. However the goal is to expand it across all Android devices in due course.