Replace Google Analytics with self hosted solutions

[ZhangYiJiang]

Google Analytics is great, but we don't really use a lot of its features. We mostly use it to track user browser (to update our support matrix), and to track events (to see if students are using certain featuees). In addition, using a third-party to collect user data is somewhat sketchy privacy wise, especially since Google explicitly uses GA data to support its own advertising by tracking users across multiple sites. In addition with more and more students using adblockers, it is possible we aren't collecting good data due to GA being blocked (I know my adblocker does block GA).

We can try self hosting our analytics instead to improve privacy and address the concerns above. Here are the candidates I've found - feel free to edit this issue to add more

Fathom

Pros

• Lightweight - written in Go, so should be fast and reasonably simple to set up
• Uses MySQL - no additional set up needed since we already have a MySQL installation

Cons

• No explicit SPA support - tracking code has no documentation, but from reading the source there's no explicit way to send pageviews I think
• New - not 1.0 yet. May be unstable

Mamoto (formerly Piwik)

Pros

• Old, well known - Mamoto is really old. That makes it pretty stable
• Uses MySQL - no additional set up needed since we already have a MySQL installation
• SPA support - not particularly easy to use, but shouldn't be too hard either

Cons

• No Docker - it's a PHP app, so updating probably won't be too problematic, but still worth taking note
• No npm package - analytics JS has to be served separately. Not a huge deal, and probably preferable from a performance point of view anyway

[ZhangYiJiang]

Fathom doesn't work because we need browser / device data, and they don't have that. Testing Mamoto now.