Education & Onboarding
18% of overallCan a new developer get to 'hello world' without asking anyone?
What it measures
We crawl the docs site (and a sample of READMEs) and grade it on discoverability and structure: is there a quickstart, an API reference, tutorials, in-page search, copy-pasteable code samples, and a visible version selector? The repo-level AI-readiness signal rates how well READMEs work as context for LLM assistants.
Why it matters
Docs are where the funnel collapses or compounds. A good docs site deflects support load, lets AI assistants answer accurately about your product, and converts evaluators into integrators on a Sunday afternoon.
Sub-metrics
| Metric | What we read | Weight | Source |
|---|---|---|---|
| Docs grade | Quickstart, API reference, tutorials, search, code samples, version selector | 70% (when a docs URL is provided) | Firecrawl + LLM grader |
| Repo AI-readiness | Structured READMEs, copy-pasteable snippets, install/usage clarity | 30% | GitHub READMEs |
How it's weighted
Docs grade is the dominant signal when a docs site is provided; repo-level AI-readiness fills the rest. If no docs URL is on file, the score is driven entirely by how well the READMEs read as context.
Best practices
- Host docs on a dedicated subdomain (docs.yourdomain.com) — easier to crawl, easier to share.
- Make the quickstart the landing page of /docs. Install, configure, first call — three steps, all copyable.
- Provide a visible table of contents and in-page search. AI crawlers and humans both rely on it.
- Ship runnable examples per language/SDK, not just code fences.
- Add a version selector so users land on docs that match their installed SDK.
- Publish an OpenAPI spec or SDK reference page that mirrors the actual API.