How Appfigures Estimates Downloads and Revenue for iOS and Android of Apps
Every day, Appfigures estimates data for 1,500,000 iOS and Android apps to provide app makers, marketers, investors, and analysts visibility into the mobile market.
When estimating, we have two critical focuses: accuracy and privacy:
Accuracy is essential because we know our members rely on our app intelligence to make real-world decisions. Whether it's spending on ads to beat a competitor or investing in a new game developer, the data you use will make (or break) your potential for success.
Privacy is important to us because we know that we can produce accurate estimates without using any data that can identify its owner or give a competitor any information they can't have otherwise.
We're obsessed with both, and that why so many app and game developers, marketers, analysts, investors, and journalists rely on our app intelligence every day.
Appfigures uses sophisticated models that turn app performance, which can be observed in the store, into download and revenue estimates by training our models with data from hundreds of thousands of apps.
Statistical modeling means making assumptions. Because we're estimating, we have to make assumptions, but the fewer assumptions we make, the more accurate the estimates are.
We minimize the number of assumptions in our models by training our models with data from hundreds of thousands of apps at a very granular level, which goes as deep as the day of the week, the app's category, and trends in the store.
Such granularity allows us to fine-tune our models, update them daily, and continue to refine them as trends evolve.
When estimating, we apply a double-opt-out process, which ensures every estimate we produce is actually an estimate.
This makes it impossible for any of the data we used for training to "leak" out as an estimate, which is why many popular app developers are included in our training pool, enabling us to have high accuracy.
Adapting to Trends
In addition to all of the above, our team constantly updates our models to fit trends that are going on in the store. For example, 2020 was full of megatrend shifts due to COVID, working from home, and lockdowns.
We've made more than 20 changes during 2020 to ensure our accuracy remains at our expected levels.