Written by


Live on Airnow Data From: Monday 12th April 2021

Model(s) Updated: Downloads

Platform(s) Affected: Google Play (Android)

Dates Considered: 2021-01-01 to 2021-01-31

The development team at Airnow Data is pleased to announce a substantial update - Google Play Model Update (11).

This release - detailed in full below - marks the beginning of an exciting period for Airnow Data, following a great period of growth in our team and capabilities. We have a series of exciting initiatives to share with you over the coming months, and look forward to getting you involved in our journey in due course.

In the meantime, find out more about our latest model update which improves further the accuracy within the platform, below:

What did the update do?

Google Play Model Update (11) calculates Airnow Data’s Download estimates where all Google Play apps are considered. The update replaced the previous algorithm used to predict these markers with a new and improved derivation; a novel approach allowing our platform to get better, smarter and more accurate data than ever before.

As at all App Analytics and SaaS companies, Partner Data is of paramount importance, as it is from this data and other data sources that we can infer the metrics that are displayed to customers. Whenever new partners are added, it is imperative that we undertake a ‘refit’, which allows all of the algorithms to retrain themselves to better consider all of the new data available to them.

What can you expect?

To calculate error, we use a formula called Relative Absolute Error, or RAE.

We consider RAE to be the relative amount of distance we are away from the real figure, or ‘actuals’. As a consequence of the latest update, we have reported a 20% + increase across the board in the accuracy of their data relative to the previous model GP10.

By country, the biggest improvements can be seen in the table below:

By category, some of our biggest improvements are:

If you have any questions about this most recent update, please get in touch.


The Airnow Data Team