Stats on Discourse replies

I did some analysis of the topics and replies in the categories here.

Announcements: 4 topics, 25% had replies
Announcements/Architecture: 4 topics, 0% had replies
Announcements/Deprecation: 10 topics, 30% had replies
Announcements/Releases: 3 topics, 67% had replies
Announcements/Security: 1 topics, 0% had replies
Community: 21 topics, 86% had replies
Community/Discourse: 10 topics, 80% had replies
Community/Events: 15 topics, 93% had replies
Community/Introductions: 24 topics, 75% had replies
Community/Jobs: 7 topics, 43% had replies
DevOps: 119 topics, 76% had replies
DevOps/Ops Help: 171 topics, 83% had replies
Development: 110 topics, 76% had replies
Development/Architecture: 12 topics, 92% had replies
Development/Collaborative Proposals: 9 topics, 78% had replies
Development/Design: 9 topics, 67% had replies
Development/i18n: 3 topics, 67% had replies
Educators: 43 topics, 93% had replies
Educators/Authoring: 14 topics, 79% had replies
Educators/Instructional Design: 7 topics, 71% had replies
Working Groups: 2 topics, 50% had replies
Working Groups/Build-Test-Release: 7 topics, 86% had replies
Working Groups/Marketing: 6 topics, 83% had replies

I don’t know what this means exactly, but there it is. I did this with the Discourse API. I’ll be cleaning up the code and putting it on GitHub somewhere.

@nedbat Thank you for compiling these stats!

For reference, this came from a discussion in the Open edX Contributors Meetup, where we were wondering what is the proportion of requests for help that goes unanswered.

Historically, the devops help requests were the ones which had the most unanswered questions. The stat “DevOps/Ops Help: 171 topics, 83% had replies” seem to indicate that this is a lot better than this used to be - though I agree that it’s hard to conclude from this stat alone, since we don’t know what the answers were – sometimes the original author replies to their own message with precisions, some of us ask the author to provide more precisions, etc. If we wanted to become more precise with this, we would probably need to start using the StackOverflow-like “Solution” checkbox in the replies, and identify which posts are actually questions. But maybe it’s a good enough proxy for now?