Analytics

An artifact that nobody can measure is an artifact you cannot improve. Two signals are collected today, both by the edge worker, and both derived from the request rather than asserted by the page.

Views

A small beacon reports four things only the page knows: how long it was visible, whether anyone touched it, which path, and the referrer. Everything else — device class, operating system, browser — is derived by the worker from headers it observed.

That split is a security property rather than a style. The beacon runs inside a page the platform did not write, so every field it could assert is a field it could forge: a client that can name its own country can name a hundred of them.

The visitor's address is the sharpest edge. It arrives on every request, it is the one genuinely identifying value in the exchange, and it is never stored — it goes into a hash and out of scope in the same expression.

Errors

When an artifact throws in somebody's browser, the fault is relayed to the platform and the owner is told. Before that, an artifact that broke on a phone reported nothing to anybody, which is indistinguishable from an artifact nobody opened.

What is not built yet

There is no per-artifact analytics report. The pipeline is live and the gallery shows a view count and who is reading right now, but the page that breaks those numbers down — over time, by device, by path — has not been built. It is named here rather than left out, because a missing report and an empty report look identical from outside.

That is the only gap here. Collecting what a visitor types is built and is a different feature from this one — see Collecting data. Analytics is what the platform observes about a request; collected data is what the page asks for and the visitor chooses to give. They are counted, stored, read and erased separately, and neither is a substitute for the other: analytics cannot tell you who is coming to a party, and a response cannot tell you whether anybody opened the page without answering.