Add product analytics for the extension

[chrisshaw]

This would help us

1. figure out what we need to do to comply with privacy requirements
2. what custom setup we desire
3. get used to the tool

Flow

As the product manager

• I launch the app by enabling the extension
• Using the PostHog toolbar, I create an action for when someone adds an annotation (before save)
• I then test the functionality by adding and not saving an annotation
• When I visit the PostHog admin, I see a $pageview event from the app launch and an Action related to adding an annotation

Constraints

• The source code for the extension should remain "pure" for the community. That means we should not be collecting usage analytics automatically unless it is our product (the extension). Ideally, the community doesn't even see the analytics code, but if they do, then they need to be able to disable/enable it.
• The analytics tool should not create an practical performance hits
  • We don't have any actual performance thresholds set, so this is just what feels right.

Notes

In conversations

Reference

• PostHog docs
  • Integration docs
  • JS snippet docs
  • JS library docs
• Slack thread covering the data privacy requirements

[robertknight]

Hi @chrisshaw - Can you confirm that this request is just about the browser extension and not for the embedded client (ie. when publishers embed Hypothesis on their website) or the client in the LMS app?

[chrisshaw]

@robertknight Open to suggestions; the intent is reporting on the client. With that said:

• definitely not LMS yet
• my thought with focusing on the client in the extension was that it would make it easier to avoid listening from clients who didn't want to send us reporting.

[robertknight]

A general comment I have regarding any kind of data collection that might require user approval, is that we could only use it in scenarios where we have a way to persist these preferences. This means either the extension or logged-in users (for whom we can store account-level preferences). The embedded client does not have reliable access to persistent storage when the user is not logged in. This is due to the way that browsers (especially privacy-focused ones) treat iframes, as part of anti-tracking measures.

As an example of such a setting that we already have, there is an account-level setting that tracks whether the user has seen and dismissed a tutorial panel in the client.