What a silent stall costs. Without this address a stale queue is invisible — your report still opens and the numbers on it are simply older than you think, which is the one failure that ruins a month-on-month comparison. CiteAngle publishes the values raw, sealed to the same discipline as our figures, so you can check us yourself.
The scale behind the queue is worth knowing: a full audit runs 50 buyer questions across 15 engines, 7 times each — every planned reading on 17 surfaces, checked by machine and sealed on completion. That is why a stalled queue is worth watching: enough work sits behind those four values that a silent delay costs you a comparison.
4
signals the service publishes about itself — answering, jobs waiting, longest wait, event ledger
All four sit in one short piece of JSON at /healthz
1800
seconds — thirty minutes — is the oldest-wait threshold that flips the page into a warning state
Field name: oldest_queued_age_s
0
summary badges. We publish the raw values and the thresholds, not a verdict we wrote about ourselves
A warning state answers with a different status code
There is a second reason the raw values sit at a public address. Without a number you can read yourself, you lose a month before noticing that a reading came in late, and the missing month is the one your comparison needed. The only other evidence on offer is a sentence written by the company selling you the measurement, which is not evidence at all.
Why a measurement service publishes this
A stalled measurement queue does not look like an outage. Your report still opens, the dashboard still draws, and the numbers on it are simply older than you think they are — which is the one failure that matters when you are about to compare this month against last. That is the risk these four values are here to remove. The longest-wait figure tells you whether the reading behind your report is current, and it tells you in seconds rather than in a paragraph we wrote about ourselves.
Publishing the raw values instead of a summary badge is a deliberate choice, and it is the same one we make everywhere else. Reports carry the conditions each figure was measured under. Public figures are traceable through the claims registry. Cells the grid planned but could not read are printed as a count and kept in the denominator. A company selling numbers should be checkable on its own numbers first, and this address is where that starts.
Four signals
The service reports itself in four values. Open /healthz in a browser and they are all there.
| Signal | What it means | Threshold |
|---|---|---|
| Answering | Whether the address responds at all. A response means the server and its database are both up | A response is healthy; any breached threshold below moves it to a warning state |
| Jobs waiting | Measurement jobs that are running or queued behind others | The count alone does not decide anything — the wait below is the deciding axis |
| Longest wait | How long the oldest waiting job has been waiting, in seconds | Past 30 minutes it moves to a warning state |
| Event ledger | Whether what happens on the site is landing in the ledger on time | A backlog moves it to a warning state |
In a warning state the same address answers with a different status code, so an external monitor only has to watch that code.
Reading the values
What comes back is one short piece of JSON. The field names are the plain ones.
01
ok
The four signals rolled into one verdict. True is healthy; false means one threshold above has been passed.
02
queued
How many measurement jobs are waiting right now.
03
oldest_queued_age_s
The oldest wait, in seconds. 1800 is thirty minutes.
04
event_repair_*
Whether the ledger has a backlog, and if so how large it is and how long it has stood.
What these signals cover
They cover our own service: the public pages, order and measurement processing, and ledger writes. The outside search and AI answer surfaces we measure are run by those companies, and how each behaved on the day is recorded inside the report itself.
A contracted audit carries a delivery date in writing, and progress shows in the console. When a run takes longer than usual, we reach out before you have to ask.
Maintenance and change notices
Planned maintenance is announced by email before it starts, with the window and the affected scope. Contracted clients get the same notice through their engagement contact.
Changes to page structure and data shape land in the docs first, and the last change date of any public page is readable in /sitemap.xml.
Tell us when something looks off
If a screen behaves differently from what you expected, email us with three things: when it happened, which address you were on, and what you were doing. That makes it reproducible fast. Security reports have their own route — see the security section of the trust center.
If you need these operating terms in a contract document, say so. Name the items and the format, and we will come back on that.
Contact usNext — Trust center · Docs · Methodology