Web analytics are ridiculously easy to implement (the server already collects all of the necessary information on which pages are hit), and there was a lot of financial incentive to develop them (since they can directly impact how much business a company gets from their web site), so they took off and have received a lot of development effort to make them easy to use and ubiquitous.
Desktop software (both open source and commercial) has a few obstacles compared to this:
- Harder to implement - Information isn't automatically collected, so you'll likely have to add some tracking calls yourself.
- More privacy concerns - It's something that users have to opt in to.
Additionally, analytics aren't always as useful as they first appear:
- If feature A is used more than feature B, is it because B isn't as useful, or is it because B is too hard to use, or is it because not enough people know that B is there?
- I suspect that desktop analytics would be harder to interpret. Analytics require interpretation (statistical analysis, visualization, etc.). For the web, companies like Google have commoditized analytics: there was enough financial incentive to justify the effort to make the data so easy to understand that anyone can easily benefit from it. This hasn't happened for desktop software; even if you could get the raw data, interpreting it would be more work.
Open source software in particular has the reputation (stereotype) of not giving a lot of attention to usability, so it's not surprising that it hasn't used a (relatively) advanced usability technique.
With that said, some desktop open source software does to analytics. For example:
- Firefox has telemetry (usage tracking), health reports, and crash reporting.
- ToME (an open source computer game) tracks player statistics and uses it to help balance gameplay.
- Debian and Ubuntu have popcon to track which packages are used.
In each case, the projects are large enough to be able to afford the difficulty of implementation (Firefox), or already have an online component (ToME and popcon) and so can more easily overcome the implementation and privacy concerns.