developer-community-health

Pick at most 8 metrics across four pillars: responsiveness, contributor funnel, activity, and sentiment. Give each a formula, a baseline from at least three prior periods, a threshold, and a named owner with an action, then review monthly.

Start with the code-centric starter set: time to first human response, change request closure ratio, contributor absence factor, and release frequency, plus a twice-yearly recommendability survey item. Read them as trends against your own baseline, never against other projects.

No. The CHAOSS reference model publishes no numeric cutoffs, and commonly quoted figures like 24-hour response or 20% monthly active ratio are unsourced folklore. Use them only as starting hypotheses and replace them with your own trailing median after three periods.

Use short surveys instead of automated scoring, which misreads blunt bug reports and dry humor as negativity. A two-item quarterly pulse with a recommendability question works, corroborated by behavioral signals like responsiveness and return rates.

Only when three conditions hold together: multiple venues, a named owner for the pipeline, and a question cheaper sources have failed to answer twice. Otherwise forge APIs and venue analytics panels cover most needs at near-zero effort.

When diagnosis shows sustained near-zero non-staff activity and unanswered questions over many months, the honest answer may be folding the space back into a smaller surface. Keeping a dead room open costs trust every time a newcomer posts into silence.