Usage, in the open

Every device that connected.

These numbers come from an anonymous launch ping: a few booleans and a version string, no identifiers, nothing that ties a count to a person. Here is all of it.

— total installs

— purchased a credit pack, of the devices active this month

Daily active devices

A device counts once on the first launch of each UTC day, in the free lane or the paid one. Last 30 days.

paid free

— today, both lanes · peak — in the window

Monthly active devices

Counted once per device per calendar month, stacked free below paid. Up to the last 12 months.

paid free

Version split

Which build the active devices are running, this month.

OS split

Which iOS/macOS the active devices are running, this month.

What these numbers are, and are not.

There is no identifier at all, on any path. No device ID, no hardware hash, no salted hash, no account, no cookie. The ping payload carries exactly eight fields: five booleans, an app version, an OS version, and a license state. That's it.
Aggregate counters only. Each figure is a single number in a table that goes up by one. There is no row per device and no way to go from a count back to a device or a person.
Client-side deduplication. Your device decides whether a launch is the first today or this month and sends a boolean. The server just adds it up.
Credit-pack status is self-reported. The app reports whether it has ever purchased a credit pack — one bit, "free" or "paid". Receipts and transaction details never reach the stats path.
Counted per device, not per person. Two devices count twice. The conversion rate is "devices that have ever purchased a credit pack and were active this month" ÷ "all devices active this month".
Opt-out, in the app. Account → Privacy has a toggle that turns the launch ping off entirely. Privacy-minded users are therefore underrepresented in every number here.