Insights & Feedback

It starts with a reader looking for something: portal search, scoped to the version they are actually reading.

Every search leaves a trace: what was searched, what returned nothing, and what returned results nobody clicked.

How readers rated each page in this version: responses, views, the helpful rate, and the comments behind the number.

Broken down page by page, so a score belongs to a specific document rather than to the portal in general.

What readers actually wrote, each tied to the page it was left on, so the feedback points at something you can edit.

All of it belongs to one portal, alongside the Insights surfaces behind it: engagement, analytics, experiments, AI cost and the audit log.

Who did what, and when. The log is signed and chain-verified, so the record holds up when someone audits it.

Reader behaviour over time, including the searches that returned nothing — the clearest signal of a document you have not written yet.

First-party traffic, split by who is reading: people, versus the AI assistants and crawlers now reading your documentation too.