Product usage analytics on the fleet - how do you all see what's actually getting used?

Been running openBalena for a bit over a year now, around 800 devices in the field. We currently use Grafana so I can graph CPU/mem/temp/disk across the fleet and get alerted. For monitoring, node-exporter on the devices, Prometheus scraping the public URLs, and Grafana. So I can graph CPU/mem/temp/disk across the fleet and get alerts.

I dont have a good way of seeing how to product is actually being used after it is shipped.

To answer questions like:

how many of our devices have ever used [feature X]

whether someone who sets one up is still using it 60 days later

Whether [feature Y] usage went up or down after the recent update?

Just to answer some questions on the product side of things. I looked at just pointing a normal analytics tool at it but they all assume a browser and a session.

I am curious how the rest of you all handle this.

do you track product/feature usage at all, or is it device health and that’s it?

if you do, build your own? What is the stack, and how has that been going?

I would appreciate it if someone could point me in the right direction or provide any insight into their experience with this. Thanks!