Causal Impact Analysis Applied to App Releases in Google Play and Windows Phone Store

App developers would like to know the characteristics of app releases that achieve high impact. To address this, we mined the most consistently popular Google Play and Windows Phone apps, once per week, over a period of 12 months. In total we collected 3,187 releases, from which we identified 1,547 for which there was adequate prior and posterior time series data to facilitate causal impact assessment, analysing the properties that distinguish impactful and non-impactful releases. We find that 40% of target releases impacted performance in the Google store and 55% of target releases impacted performance in the Windows store. We find evidence that more mentions of features and fewer mentions of bug fixing can increase the chance for a release to be impactful, and to improve rating. UCL DEPARTMENT OF COMPUTER SCIENCE