Is First Run On Device

[winsmith]

Based on the sheer number of people asking me about this every time I post this graph, it might be worth automatically adding some kind of simple "isFirstRunOnDevice" flag on all logs into the SDK

• Gate this code so it only runs on platforms which du supply a UserDefault implementation
• Check if com.telemetrydeck.[APP_ID].hasInitializedBefore is set (replace [APP ID] with the app's ID. If it is not set, set it to true.
• On Signal send, add a payload value with key "isFirstRunOnDevice" and value either "true" or "false"

[Jeehut]

There is a more reliable API to check if this was the first app install using the AppTransaction API which as an originalPurchaseDate field that returns the date the app was first downloaded on the App Store. This way even installs after switching a device would not get re-detected as "first installs".

Having that said, we still need a way to persist that we have reported this already. But it probably should be combined with the original purchase date and if that was more than – say 3 days ago – then it's likely this is not the first install.

And rather than using UserDefaults, at least on iOS we could use the keychain. Because entries there are not deleted even if the app is uninstalled and reinstalled on the same device.

[kkostov]

If the app is reinstalled after some time/different device, doesn't this count as a "first run" as well?

[Jeehut]

@kkostov That depends on what the stats really will represent and how they will be used. You are right that "first run on device" is more clear when it's not synced across devices, but I'm thinking about this as part of the Pirate Metrics approach overall and there, when thinking about acquisition/activation, it's more relevant what "users" are doing rather than what "devices" are doing. I don't think it's "user acquisition" if a user just buys a new phone.