download count and per device statistics mismatch?

flocsy🤠 said:

This is also the reason these stats are almost of no value (unless you only support system5+ devices)

If I'm right and device download stats include updates, then the stats have even less value.

So I have a data field app which only has 687 total downloads, and it has had a large number of updates (more than 10). (It's a "clone" of another app of mine, so the user can install it more than once) (Obviously this phenomenon is much more likely to be noticed for apps with a small number of total downloads. For any app with 10,000 or more downloads, we'll never see it, bc 1 / 0.01% = 1 / 0.0001 = 10000)

Sure enough, the lowest percentage is 0.01%. If that represents a minimum of 1 "device install", then the lowest possible number of device installs is ~10000. Even if we suppose that the value was actually 0.019% but rounded down to 0.01%, that's still ~5263 "device installs".

I'm sure we could see the same thing for app with a small number of downloads and a decent amount of updates (10+).

I really think the most likely explanation is that "device installs" includes updates (and we know that the total download number does not.)

Wow! Thank you for a great and thorough explanation!

I did not know that older devices (FR935, F5 etc) didn't provide stats. And I didn't know that the per device stats was so "inclusive" but I guess thats ok. Then I shouldn't bother too much about the per device stat other than perhaps see what kind of distribution over devices my app has. But I have to admit though that the per device stat combined with the number of versions and your computaiton of a "loyalty factor" was quite interesting. :)

Np. To be fair, “device install stats include updates” is pure speculation on my part, but I can’t think of any other explanation (other than a bug and/or bad data).

If you make a new app, you can try to track the numbers while the number of updates and total downloads are low, and see if there’s ever any stats that contradict my guess (like similar percentages which suggest many more device installs than total downloads, but can’t be explained by updates.) Ofc you can never prove my theory this way, but you can disprove it.

Or you could file a bug report and see if anyone on the Connect IQ team will clarify the situation definitively.

If anyone has some time on their hands and two CIQ 3.3+ / System 5+ devices, they could try to prove or disprove my theory by releasing a beta app (so it’s private to their account) and updating it several times.

On one device, they would install the app once, and never update it.

On the second device, they would install the app, and update it each time an update is released.

If the device percentages change at some point (like when the 2nd device receives its first update), then my theory is correct. If the percentages never change, then I’m wrong.

Ofc this might be complicated by the fact that we don’t know how often the device stats are updated. If it’s instantly, then this test could be done in a few minutes. If it’s daily, it might take a couple of days or longer.

Probably just easier to ask garmin haha

EDIT: nvm it looks like stats aren’t collected for beta apps, not even total downloads

Indeed no stats for beta. 

And I think the stats are updated about once a day.