Make the case to restate ubiquity

[chadwhitacre]

Self-hosted is a component of Sentry's ubiquity metric (the "90k orgs" stat we put on all of our job descriptions, for example). With https://github.com/getsentry/sentry/pull/62441 we now have quite a different picture🔒 for self-hosted.

We are already using this new understanding in OSPO (e.g., https://github.com/getsentry/team-ospo/issues/232). This ticket is to track working with Sentry's Biz Ops team to update our overall company understanding for reporting. The task is to:

• [ ] redo the work I did to produce the above charts
  • [ ] export this explore🔒 from 217🔒
  • [ ] modify the explore to reflect the old state and export that
  • [ ] for both old and new, aggregate versions locally with a script (or figure out how to do it in Looker)
• [ ] document it better (I scrapped the aggregation script I used :sob:)
• [ ] incorporate a ratio of active to inactive installs from the cohort where we have that data to adjust where we don't.

[hubertdeng123]

So are we essentially trying to find the ratio of inactive to active installs from 24.1.0+ to try and estimate the amount of active installs we have in the versions 20.0 to 23.12.0, where we don't have that data?

[chadwhitacre]

Yes, with the addition that we also have active vs. inactive in versions prior to 20.*, yes? What would be interesting is to see the ratio of active vs. inactive over time for all versions where we can get both numbers.

[hubertdeng123]

Yep, we do have the info for active vs inactive in the older versions too. Good call

[BYK]

Keep in mind that there are companies ahem that run self-hosted behind isolated networks.

[chadwhitacre]

Hehe. ;-)

Yes, we've always known that reported instances is a lower bound, that's still very much true. It's quite eye-opening to me to realize that our self-hosted community is much larger and alive than we previously could see easily. Worth investing in! :-)

[chadwhitacre]

Lost momentum on this, gonna close it out for hygiene.