Agentic Modules
The Analyst flags readings far from normal, and slow drift, each with its baseline. It has no schedule. The Auditor calls it in its sweep.
How it works
Two tests, outliers and drift runs, against each series’ own baseline. Every anomaly also goes out as an event, so other Modules can read it.
Anomalies
Each reading is compared to the average and spread of its own series, so the target moves with your process instead of sitting fixed. You set how far out is too far for each check, 3.0 sigma by default. Every finding shows the baseline and how far the reading sat from it.
Drift
Eight readings in a row on one side of centre count as a run. That catches the slow movement no single reading would trip. A series without enough history is skipped until it has enough to trust.
Form drift
Every numeric field on a submitted quality form gets watched, including the ones nobody thought to check. Drift is measured across a run of submissions. The finding names the field, the run and the baseline it drifted from.
Auditor handoff
Findings on quality metrics feed the Auditor’s root cause run. The Auditor proposes a hold, and it is carried out or sent for approval depending on the authority you’ve set. Anomaly events also go out for other Modules to read.
Diagnostics
The same threshold, run length and baseline apply whether it runs on the Auditor’s schedule or when you ask. Pick the series and the sigma, then run it. Findings come back with the exact points that tripped them.
Test types
It runs two tests today: outliers and drift runs. It has no schedule and no inbox of its own, so it only runs when the Auditor or a diagnostic calls it. A field-drift finding raises a notification through the Auditor and stops there, with no root cause draft and no hold.
Every reading is scored against its own series. Cross the sigma threshold and it becomes a finding with the baseline and deviation attached. The reading itself is never changed.
Eight points in a row on one side of the baseline count as a run. A series without enough history is skipped until it has enough to judge.
No. The platform runs it when the Auditor’s quality sweep calls it, or when someone asks for a diagnostic. It has no schedule of its own.
From the series itself, and every finding reports the baseline it was measured against. A series with too little data is skipped.
They share the two tests. The SPC Engine charts your measurements. This runs inside the Auditor’s sweep, where it feeds root cause and holds.
Quality-metric findings go into the Auditor’s root cause run and, depending on severity and your authority settings, a hold. Field-drift findings raise an event and a notification.
30 minutes on your quality data, and how outliers and drift runs get flagged against their own baseline.
We use cookies to understand how the site is used and to connect form submissions to earlier visits. No analytics or marketing cookies are set until you choose. Privacy Policy