A Data-Driven Model for Mobile Game New Version Update Evaluation

The paper proposes the Fish Tank Model (FTM), a data-driven framework for evaluating mobile game version updates, motivated by a gap in publishing-focused game analytics and the difficulty indie developers face in judging whether a new release is well received. The framework combines two metrics, New Version Rating (NVR) and New Version Change Rate (NVCR), which blend player reviews with changes in installs, paying users, revenue, and daily active users (DAU) to assess whether players favor a new version. The approach is demonstrated through a design science case study of a casual mobile game studio's version 2.7 update, released on iOS and Android.

The update's reception was mixed across storefronts. The iOS App Store rating rose from 4.4 to 4.7, while Google Play held steady at 4.7, so the authors supplemented ratings with in-game data. Net revenue increased by roughly 58% on iOS and 54% on Android, paying users rose 25% and 63% respectively, and average DAU grew 3% on iOS and 1% on Android. Install gains were also observed, though the authors caution that these may reflect concurrent marketing efforts rather than the update itself.

Correlation analysis reveals platform-specific dynamics. On Android, installs and DAU were strongly correlated (γ = 0.74), but new installs contributed little to revenue (γ = 0.09). On iOS, the install-DAU relationship was weak (r = 0.33), and the install-revenue relationship was slightly negative (r = -0.31), indicating that new installs contributed little to engagement or monetization. The authors conclude that the update performed well overall, that new players on Android need better monetization such as first-payment rewards, and that iOS requires stronger new-user acquisition. They recommend evaluating installs, DAU, and revenue separately, since no consistent relationship among them emerged for this update.