Observe
Observation answers a narrow question: where does documentation help, and where does it block people? Collection is consent-first, and the numbers you see are the events the server accepted.
By default
- No analytics and no replay until a purpose is consented to. Nothing is captured before consent, including on the first page load.
- Code, chat, terminal and password fields are blocked from replay; sensitive values are masked before storage.
- The publishable site key grants ingestion only — never reads and never money.
- Billing counts server-accepted, deduplicated events; dropped and rejected events are excluded and reported.
What you get
- Readership and task-completion views with an explicit data window, sample rate and drop count.
- Search terms that returned nothing, feeding the Grow proposal engine.
- Replay for debugging with a short signed URL, a stated reason and an audit record.
- Retention you set, with a sweep that purges sessions past the window.
Actions
Actions
- View synthetic dashboard
Opens the workspace observe surface. It requires a running control plane and says so honestly when none answers; no sample statistics are shown on this public page.
- Enable on site
Ingestion needs a site key minted in a workspace; this build runs no hosted analytics deployment, and nothing is captured before the published site consents.