Clarification: iTunes Connect Analytics Data

It's all about purchases vs. installs, apparently (think buying an iOS app on a computer first, then choosing to run it on a given device post-purchase).

About 'App Units', which are defined as...

"The number of first-time app purchases made on the App Store using iOS 8 or later. App updates, downloads from the same Apple ID onto other devices, and redownloads to the same device are not counted. Family Sharing downloads are included for free apps, but not for paid apps."

Since purchases are not something users can opt into or out of in terms of sharing with Apple, units don't fall into voluntary metrics.

In contrast to 'purchases', however, 'installs' are part of the opt-in metrics.

"Installations - The total number of times your app has been installed on an iOS device with iOS 8 or later. Redownloads on the same device, downloads to multiple devices sharing the same Apple ID, and Family Sharing installations are included. App updates aren’t counted. Totals are based on app users who agree to share their data with you."

Anything on the main dashboard marked with Opt in is covered by this. On the metrics page it's basically everything under Usage.

As I understand it the distinction comes down to where the data comes from - if it's straight from the App Store, there is no privacy implication, as they know this on the server side, so Apple gives you the actual numbers. Though for me these numbers always differ slightly from Sales & Trends, but average out the same over time.

For the usage stats Apple is actually relying on the device 'phoning home' with this information. This is considered to be a privacy issue - maybe the user doesn't want this sort of behaviour happening on their device. This requires OS support (ie this is only available since iOS 8 as that's the first version of the OS had this code to phone home), and it seems to be done in batches - which is why there's a day or two's delay before some of this data is available. On iOS 8, the dialogs getting the user to opt in to this was really badly worded and the proportion of users who opted in was really bad (for me anyhow). On iOS 9, this is about 10% better, but yes, you're still having to treat these like statistical samples in order to estimate the approximate numbers for the usage stuff.