how do you estimate the revenue numbers for apps that don't publicly report earnings, and how accurate is it compared to sensor tower on the same datasets?
Report
Maker
@egemensafe95621Ā Downloads: On Android, Google Play itself publishes a "min installs" floor (e.g. "500M+") ā that's real public data, so I anchor to it. iOS doesn't expose an installs number, so I estimate from rating count using the industry rule of thumb that roughly 1 rating comes in per 100-200 downloads.
Revenue: I take estimated monthly downloads and multiply by a category-specific average revenue-per-download (games monetize very differently than, say, social apps), adjusted for price and whether the app has in-app purchases.
Accuracy vs. Sensor Tower: honestly, not as precise, and I don't claim otherwise. Their edge is a real device panel ā actual usage and spend data from millions of devices, built over a decade. That's a fundamentally different (and expensive) data source that public signals can't replicate. My numbers are directionally useful ā right order of magnitude, good for comparing apps to each other ā not exact, and I say that openly in the product rather than pretend otherwise.
I'm working on closing the gap: I just started daily historical tracking (rank/rating movement over time) to build my own calibration dataset, and the architecture is designed to plug in a real paid data provider later for users who need enterprise-level precision, while keeping the free heuristic tier for everyone else.
how do you estimate the revenue numbers for apps that don't publicly report earnings, and how accurate is it compared to sensor tower on the same datasets?
@egemensafe95621Ā
Downloads: On Android, Google Play itself publishes a "min installs" floor (e.g. "500M+") ā that's real public data, so I anchor to it. iOS doesn't expose an installs number, so I estimate from rating count using the industry rule of thumb that roughly 1 rating comes in per 100-200 downloads.
Revenue: I take estimated monthly downloads and multiply by a category-specific average revenue-per-download (games monetize very differently than, say, social apps), adjusted for price and whether the app has in-app purchases.
Accuracy vs. Sensor Tower: honestly, not as precise, and I don't claim otherwise. Their edge is a real device panel ā actual usage and spend data from millions of devices, built over a decade. That's a fundamentally different (and expensive) data source that public signals can't replicate. My numbers are directionally useful ā right order of magnitude, good for comparing apps to each other ā not exact, and I say that openly in the product rather than pretend otherwise.
I'm working on closing the gap: I just started daily historical tracking (rank/rating movement over time) to build my own calibration dataset, and the architecture is designed to plug in a real paid data provider later for users who need enterprise-level precision, while keeping the free heuristic tier for everyone else.