Site usage analytics

[thibaudcolas]

We have been discussing setting up analytics on the website for a while (forum: Add Plausible Tracking to DjangoProject.com?). At this point I think there is a pretty clear case for it and it’s more of a matter of deciding what tool(s) we could use and setting them up.

So I think opening this as a GitHub issue will help make this more of a task that we want to happen, not just an idea.

Tasks

• Discuss analytics use cases and potential problems
• List possible approaches
• ⌛️ Try out 2-3 promising approaches/ tools
• Check with ops team on self-hosting
• Pick top 1 and top 2 options
• Set up the most relevant options on the site
• Access for website WG, fundraising WG, Board, and any other relevant stakeholders

Tools considered so far

For the dedicated analytics / tracking tools, likely options:

• Plausible CE (used by the PSF for their docs)
• Plausible SaaS
• Cabin (has a very generous free tier)
• Umami SaaS
• Umami self-hosted
• Google Analytics (industry standard but requires cookies)

Requirements

We want something that does not require cookies, is compatible with privacy laws (GDPR, CCPA), ideally has minimal to no impact on performance for site users, ideally allows us to get good reports with little efforts. I think there are three approaches that are viable:

1. Use the existing Django server logs (behind Fastly caching) as an incorrect but likely interesting data source. We were hoping to do this for Docs search: tweak results ranking so release notes have lower priority #1628 but this never happened.
2. Reach out to Fastly to see if we could get access to their analytics product, and use that.
3. Set up a dedicated JS-based analytics beacon

I have no experience with Fastly analytics / logs tools, but from my experience with Cloudflare’s equivalent this is pretty viable to understand popular pages and high-level geographic audience details. That’s about it. Big benefit is it doesn’t require any additional, potentially invasive tracking.

Those come with the clear drawback of loading more code for users, and tracking more data than we would often want. But I think there are ways to mitigate that - picking the least-intrusive option, or only having tracking turned on occasionally.

Use cases

Just reiterating the use cases we have - here is what we want to know:

• Rough geographic distribution of our audience across countries
• Page views / Session counts site-wide
  • Top landing pages
• Hits per search queries
• Bounce rate of specific key pages (fundraising, docs, etc)

To help illustrate the benefits of this kind of data, I wrote a blog post about the Python docs analytics: What we can learn from Python docs analytics. For Django, we would use this data to:

• Know what information is so popular it might warrant more attention / restructure
• Decide which translation efforts to encourage
• Revamp the docs based on common searches
• Understand where there is friction in our donation flow

Relevant existing issues that this data would support:

• Docs search: tweak results ranking so release notes have lower priority #1628
• Website redesign #2287
• Delete Committees page #2257
• Non-docs content enhancements from user research #1505
• Documentation search results relevance improvements #1097
• Improving the organization of information #1495
• Optimize the donations page #1500

[ehmatthes]

I am fully in support of sorting this out, with a clear description of what the analytics capture and what they can't capture. One concrete thing I'd love to know: when people start the tutorial, how far do they get?

[MRigal]

What about Umami ?

[thibaudcolas]

@MRigal since you’re proposing it, could you tell us more about why it’s suitable based on the requirements I outlined above?

[MRigal]

Hi @thibaudcolas

I just thought to drop that in because we (at A2EI) had similar requirements and the person in charge of the investigation chose Umami. I was actually wondering/following based on what would you make a decision for one or the other software and this is why I've suggested it to the list, I have no relation whatsoever with Umami.

From your use cases:

Rough geographic distribution of our audience across countries
Page views / Session counts site-wide
Top landing pages
Hits per search queries
Bounce rate of specific key pages (fundraising, docs, etc)

I think only the last but one would be missing out of the box. Although the "hits per search queries" might be overcome by adding some specific https://umami.is/docs/track-events. But maybe someone finds/knows a way to do that in a better way out of the box.

What made us chose Umami:

• self hosted
• open source
• privacy by design
• simple to use
• not much bloat, only the features that we need
• without cookies
• data can be easily accessed via API or database
• automation via API (e.g. setting up new sites)

[wsvincent]

What can we do to make progress here team? For example, just today PyCharm marketing was asking me if we had analytics on how the Django Developers Survey post did. This is standard stuff: marketing departments collect analytics for a campaign to analyze what worked and didn't. The fact that we are totally flying blind is not great. To say nothing of efforts to attract premium sponsors.

Anyway, would love to help on this, it just feels a little stalled if someone in the know could weigh in on next steps?

[thibaudcolas]

I’d like to push this but have very little time for it until probs Feb 2026. But progress update:

• Proposal for a Django marketing working group, folks who would definitely need those kinds of analytics to do their work. There’s an explicit point on there about the group defining and tracking "KPIs aligned with objectives of the DSF". There’s been some work of our social media and fundraising groups on this so far, that is promising. Nothing too complex, just retrieving our follower counts and engagement rates in a way that is sponsors-friendly.
• I’ve been trialing Umami and Cabin on a subset of the of the site’s pages (not that particular blog post unfortunately).

Trial is going well so far. I chose a varied subset of pages that are not mission-critical so we would have interesting data. Here’s the links to the public reports:

• Public Cabin reporting
• Public Umami reporting

---

I chose those two because of their ease of setup. I can’t imagine ourselves having the bandwidth to self-host Plausible, or wanting to use Google Analytics, so to me an obvious next step would be for someone to do a three-way comparison between Umami SaaS, Cabin, Plausible SaaS. Can start with a simple matrix of requirements from this issue, and add more later.

[jefftriplett]

I don't have strong feelings on which service to use. I am personally using Umami on a bunch of websites (self-hosted on Hetzner), and I'm happy with the results.

With stats, it's harder for us to make some decisions, and it would help with sponsorship if we had numbers to back things up. I think it's long overdue, but I'm not sure what's stopping us from moving forward.

The forum post has been open for well over a year now, and I'm not seeing any pushback or reasons to now move forward with a solution. https://forum.djangoproject.com/t/add-plausible-tracking-to-djangoproject-com/35581

What are the infrastructure concerns or blockers?

[wsvincent]

As long as we use something that anonymizes users, it's not "that" hard to switch services. And if we wanted something hosted, I bet we could find a company who would donate it to us.

[pauloxnet]

I can confirm that Umami works well also for my personal blog with the free account. If we want to try it self-host it I think we don't have any blockers.

[thibaudcolas]

I’ve set up an internal but publicly-visible doc so we can structure the next steps a bit better than via issue comments: Web analytics for djangoproject.com. I’ve contacted the ops team to ask for their input on self-hosting specifically, as it would be owned by them in the long run.

The next step I think is to find someone / multiple people keen to refine our requirements for us, and complete the comparison matrix in the doc accordingly. This isn’t something where we have to make the best possible choice "or else", but I think we all definitely have enough collective experience with this that we might as well take the time to decide what’s our exact needs. For example,

• Some of those tools might have integrations that would be relevant for us. (Google Search Console?)
• I’d say it’s a must have that the reports produced are publicly visible
• Maybe some of us would like to see weekly / regular reporting emails

[thibaudcolas]

Update from our ops team that they’re happy to own the setup and maintenance of a self-hosted option. I’m not big on self-hosting myself but can see how it clearly aligns with our values :) So with that in mind, I think we can proceed with a narrower comparison of "Plausible CE vs. Umami".

I’ll try to write down a more specific list of use cases we foresee, that go beyond the basics. We will need to do this anyway once we have the data available, so might as well do it now when it can be useful for selecting an option. If anyone has good expertise with analytics for non-profits sites / docs sites, please do make suggestions