How we measure this site

We measure our own front door the way we tell you to measure yours: on our own metal, in the EU, with no cookies, no device storage, and no US processor — and we publish the list of what we keep. This page is that list. It describes facts and configuration, nothing more.

The tool, and where it runs

Site measurement is Umami, an open-source analytics tool we run OURSELVES — one container on our own PostgreSQL database, on EU infrastructure. There is no Google Analytics, no hosted analytics service, and no external operator anywhere in the path. The container sits on an internal-only network with no route to the internet; its update-check and telemetry are switched off, and it makes no outbound connection. Its data lives in our database, subject to the same residency as everything else here.

Cookieless, and no IP kept

The tracker sets no cookie and reads nothing from your device. To tell one visit from another within a day it uses a hash of your IP address and user-agent, salted with a key that ROTATES DAILY — so yesterday's hash cannot be matched to today's, and the raw IP address is never stored. It also honours your browser's Do-Not-Track signal.

The query allow-list

Only five query parameters are ever sent, and they are an ALLOW-list, never a block-list: utm_source, utm_medium, utm_campaign, utm_content, utm_term. Everything else in a URL's query string — including a search term (?q=), a request id, and any parameter anyone adds in future — is stripped in your browser BEFORE the event is sent. A new parameter is invisible to measurement by default; it has to be added to this list on purpose.

Paths are patterns, never ids — and why /verify matters most

The tracker reports the page's ROUTE PATTERN, never the real path. /verify/[requestId] is recorded exactly like that — with the literal text [requestId] — never with the actual request id in the URL. This one matters most: a real id in that URL would reveal which specific attestation someone is auditing, and when. That is precisely the class of metadata this product exists NOT to accumulate, so accumulating it in a marketing tool would be indefensible. The same pattern rule applies to every dynamic route (/models/[...slug], /office/*, /tenant/*, and any other).

What is never measured

The admin console is not measured at all — the tracker is not even loaded there; operator behaviour is not a marketing metric. White-label tenant sites are not measured either: a tenant's visitors are that tenant's business, decided by the same single first-party-host rule the rest of the platform uses. And nothing here touches your API requests or inference — prompts and completions are never seen by this tool.

The first-party script path

The tracker script and the endpoint it sends to are served from our own domain under a neutral path (/z/s.js), not from a third-party host and not under the name every ad-blocker recognises. We do this so the measurement is accurate — an ad-blocker would otherwise erase real visits and flatter the numbers — and we say plainly that we do it and why. Nothing about it evades your Do-Not-Track choice, which the tracker still honours.

Retention

Measurement rows are kept 365 days and then pruned on a schedule — the promise is enforced by code that actually runs, not merely stated. Sign-up attribution (the referrer host and campaign parameters read once when you create an account) is removed automatically when the organisation is deleted.

What this does and does not tell you

This describes what the site records and where. Whether your own use of NeuralRing needs a specific consent control is a question for your counsel; we set out the facts so they can make that call. And every claim above is checkable: the allow-list and the script path on this page are printed from the same values the running code uses.