Analytics for documentation sites

The idea

Docs teams measure success with tools built for marketing sites. Plausible tells them the getting-started page got 4,000 views; it says nothing about the 300 people who searched webhook retry, got zero results, and filed a support ticket instead. The crowd has repeatedly judged simple analytics a commodity — three privacy-first analytics tools at $9-20/mo all drew verdicts that generic pageview counting is replaceable.

That is exactly why the wedge is vertical, not cheaper: docs-native events that a generic tool cannot ship. Zero-result searches, pogo-sticking between articles, exits to the support page, code-block copies as a success signal — and a monthly AI-written report that turns all of it into a ranked list of the five articles to write next. Pageviews are the commodity; write these five articles is the product.

Who pays

The buyer is a DevRel lead, docs engineer, or technical writer at a developer-tools company, paying from a tooling budget that already covers $500/mo services without a procurement cycle. $29-49/mo lands as trivial. Distribute through Write the Docs community, DevRel Slack groups, Docusaurus and Mintlify plugin directories, and a free docs health check tool as top-of-funnel.

MVP scope

• One-line snippet with auto-instrumented docs events (search, scroll, copy, pogo-sticking)
• ClickHouse ingest with cookieless daily-hash sessions
• Dashboard: failed searches, dead-end pages, article scroll detail
• Monthly Claude-generated Gap Report emailed as five article briefs
• Free tier to 10k events with self-serve Stripe upgrade
• Skip for v1: Docusaurus plugin packaging, A/B testing, feedback widgets, multi-site rollups

Build DocSignal, docs-native analytics installed with one script tag. Stack: Next.js (App Router) for the dashboard and ingest API, ClickHouse for event storage, Claude API for the monthly gap report. No cookies, no PII — hash IP plus user agent daily for session grouping, Plausible-style.

The tracking snippet auto-instruments docs-specific events beyond pageviews: search queries and result counts (hooking Algolia DocSearch and common search inputs, capturing zero-result queries), scroll depth per article, copy events on code blocks, exit-after-search, rapid back-navigation (pogo-sticking between articles, a confusion signal), and outbound clicks to support/contact pages, which mean the docs failed.

Dashboard pages: overview (top articles, search health score), failed searches ranked by frequency, dead ends (high-traffic pages with low scroll depth and high exit), article detail with scroll heatmap, and the Gap Report — a monthly Claude-generated brief that clusters failed searches and dead-end topics into five concrete article briefs with suggested titles and outlines, emailed to the team.

Non-obvious details: dedupe search queries with trailing-keystroke collapse (only record a query after 1.5s idle or result click) or the failed-search list drowns in prefixes; detect the docs framework (Docusaurus, Mintlify, GitBook) from meta tags to auto-configure search selectors.

Pricing: free to 10k events/mo, $29/mo to 100k, $49/mo to 1M. Stripe billing.