04 METRICS ✣
Community Metrics.
Layer 2 of the metrics stack: how is the developer community around your product doing? These metrics are leading indicators for Layer 1 business outcomes. A weakening community in Q1 produces weakening retention in Q2 and weakening reve…
Layer 2 of the metrics stack: how is the developer community around your product doing? These metrics are leading indicators for Layer 1 business outcomes. A weakening community in Q1 produces weakening retention in Q2 and weakening revenue in Q3.
Active members
The foundational metric for any community.
Definitions
- DAU / WAU / MAU. Daily, weekly, monthly active users.
- What “active” means. Posted? Reacted? Logged in? Each is a different threshold. Pick one and apply it consistently.
- Per-channel and aggregate. Track per Discord server, Discourse forum, GitHub repo, etc., and a deduplicated aggregate.
What’s healthy
- DAU/MAU ratio of 20–40% indicates a sticky community.
- Sustained MAU growth quarter over quarter for a growing product.
- Sustained MAU at scale for a mature product (decline is a serious signal).
Contributor return rate at 90 days
What percentage of community members who contributed something in a 90-day period return to contribute again in the next 90-day window?
- Healthy. 25–40% return rate for sustained programs.
- Unhealthy. Under 15% suggests members try, get a bad first-contribution experience, and never come back.
This is one of the most diagnostic community-health metrics because it measures whether the community actually retains contributors rather than churning them.
What counts as a contribution?
Define carefully. Common definitions:
- A merged code PR.
- A reviewed PR.
- An answer that the asker accepted.
- A community-channel message at a defined threshold of quality.
- A submitted talk.
- A published content piece tagged to the company.
Many teams use a quality-weighted contribution score: different contribution types are worth different points. A talk delivered = 50; a merged PR = 30; a substantive forum answer = 10; a Discord message = 1.
First-response time
How long does it take for a community-posted question to receive a substantive response?
- Goal. Under 4 working hours during business hours for first response; under 24 hours always.
- Important. Track median, p90, p99. The p99 (worst case) often matters more — it’s the user whose first impression is “no one answered me.”
Member NPS
Net Promoter Score asked specifically of community members: “How likely are you to recommend this community / product to a peer?”
- Scale. 0–10.
- NPS = % promoters (9–10) – % detractors (0–6).
- Benchmarks. For developer products, scores above +40 are healthy; +60 is excellent; +20 or lower is a yellow flag.
NPS is most useful when:
- Run quarterly.
- Followed up with qualitative free-text question.
- Trended over time (the trajectory matters more than the absolute number).
- Stratified by member type (advocates vs. casual users).
Sentiment
What’s the emotional valence of conversations about your product?
- Manual. Review a sample of community posts and AI-assist or analyst-classify into positive / neutral / negative.
- Automated. AI-driven sentiment analysis over message corpora (Common Room and similar tools provide this).
Most useful trended over time and cross-referenced with product/event launches. A sentiment dip following a release tells you something specific.
Orbit metrics
If you operate the Orbit Model (see ../03-frameworks/orbit-model.md):
- Orbit-level distribution. How many Explorers, Participants, Contributors, Advocates?
- Transition rates. Explorer → Participant, Participant → Contributor, Contributor → Advocate, over each period.
- Reverse-transition rates. Outflow from Advocate / Contributor / Participant orbits.
Healthy communities show inflow at every transition; struggling communities show outflow at one transition that bottlenecks the rest.
Member-generated content volume
How much content (posts, videos, talks, PRs, sample apps) is your community producing?
- Volume per period.
- Reach of that content (where measurable).
- Cross-platform. Some members are loud on Discord but produce no public content; others are quiet but write influential blog posts.
This is one of the strongest signals of Advocate-orbit health.
Cross-platform identity resolution
To compute most of these metrics rigorously, you need to resolve the same human being’s identity across platforms. The same person may be:
- A GitHub username.
- A Discord ID.
- An email address.
- A Twitter / Bluesky handle.
- A LinkedIn name.
Without identity resolution, you can only measure per-platform activity. With it, you can measure the human.
Tools that help: Common Room, LFX Community Data Platform, custom warehouse joins. See ../08-tools/community-crm-platforms.md.
Geographic and demographic distribution
For global programs, monitor:
- Geographic spread. Country / region distribution of active members.
- Language distribution. What languages do community posts happen in?
- Time-zone distribution. Is your community always-on or concentrated?
Useful for identifying regions where investment would expand the community vs. regions where existing investment is paying off.
Survey-based metrics
Annual or biannual surveys of the community surface things metrics cannot:
- What are members’ goals?
- What blocks them?
- What do they wish your product did?
- Why did they choose your product over alternatives?
- What would make them recommend it more?
The SlashData State of the Developer Nation survey methodology is a good reference for designing one.
Anti-patterns
- Tracking only growth, not retention. A community can grow rapidly while bleeding members at the back end.
- Treating Discord member count as community health. Many of those members will never post.
- Conflating engagement with sentiment. Loud is not the same as positive.
- Reporting community metrics without business connection. “MAU is up 20%” produces “and so?” from executives unless connected to downstream outcomes.