Selfhosted

I'm considering Matomo, but its tracking script is huge and can't easily be bundled with my site's code (whereas systems like Plausible have their tracking scripts available on npm). Looking for something a big lighter on the JS.

Each request to the site may use a different port though (if the keepalive times out).

I'd grab the data from the http logs if it was me and use a tool like goaccess.

Unfortunately this has the same issue with multiple people sharing an IP appearing to be one user.

Another major use case I thought of are people stuck on CGNAT. It's easily possible for tens of thousands of customers to share just dozens of public IP addresses. Unfortunately some providers are rolling out CGNAT instead of IPv6 even though it causes so many problems.

I don't think this would fully help? It'd still show that one user performed all the events.

I just need a way of explicitly specifying a user identifier instead of it being assumed based on IP and user agent. I don't think Plausible has a way to do that. I was chatting with the developer of Swetrix and they said it might be possible to add this as a feature.

Umami is one I'm considering. I don't think it allows a session ID to be provided though.

It also uses a row-based database (like MySQL or PostgreSQL) rather than a column-based database (like Clickhouse), which for analytics data is way less efficient and significantly slower once you have a lot of data. Pretty much all queries you'd want to run load a lot of data by column rather than row (eg charts of most common browsers, countries, etc over the past 6 months) so they benefit greatly from a columnar database.