How we measure DevRel
Smoower scores each organization across six pillars derived from public signals — GitHub activity, docs quality, package downloads, AI-readiness, community presence, and content cadence. Everything here is computed from data anyone can verify.
The six pillars
Repo polish across the org + package adoption on the registries that matter.
Docs quality + how AI-ready READMEs are for assistants and new devs.
External contributors, discussions, sentiment, Discord, programs.
llms.txt, hosted MCP, external mentions, dev-focused content.
Active repos in the last 90 days + blog/video cadence.
Samples, templates, forks, downstream usage.
How the overall score works
Each pillar is scored 0–100 from its own sub-metrics, then combined into a single weighted average. Code & Distribution and Reach & AI Readiness carry the most weight; Education and Community sit just behind them; Momentum and the DX Multiplier round things out. Orgs that are strong on both engineering polish and onboarding get a small quality boost on top.
When a signal is missing — no docs URL, no package registry, no Discord — its weight drops out instead of penalizing the org with a zero. We only score what we can see.
Data sources
- • GitHub REST + GraphQL (repos, PRs, contributors, discussions)
- • Firecrawl (docs structure, blog/YouTube content)
- • npm, PyPI, NuGet, RubyGems, Crates, Packagist, Go
- • Reddit + Hacker News search
- • Stack Overflow tag stats
- • Live MCP probes (mcp.<domain>, /mcp, /sse, …)
Refreshed daily. Manually triggered re-analysis is available on each org page.
Limitations
- • Public-data only. Private repos, internal dashboards, and paid-only content are invisible to us.
- • Signal not certainty. A high score reflects strong public artifacts, not whether a program is well-run internally.
- • Not a vendor ranking. We measure DevRel surface area, not product quality.
- • LLM-graded signals (docs grade, sentiment) carry model bias and are sampled rather than exhaustive.