Monitoring

The Monitoring page answers one question: is production running normally right now, and if not, what needs attention?

It is a live view of today. It refreshes on its own every 20 seconds. For historical trends over weeks and months, use the Reporting page instead.

Monitoring covers production only, and only projects that have been published. A project still in design has nothing running to monitor, and the page says so rather than showing zeros.

Scope

The Projects control, on the row of page controls above the funnel, sets what you are looking at. It sits with the other page controls rather than in the title bar, because it changes what this page reports on and nothing else. It is a checklist rather than a single choice:

• Tick any combination of projects. The whole page aggregates the ones you tick, and adds a per-project breakdown so the outlier is one glance away.
• All projects ticks everything you are allowed to see. It stays current: a project published later joins the view on its own, without you revisiting the picker.
• Only, on any row, jumps straight to that one project. Use it instead of unticking everything else.
• Only the projects you have permission to monitor are listed, and only published ones. You cannot untick the last project, since there would be nothing left to show.

Your selection is remembered the next time you open the page, and it applies to this page only. It never changes your current project anywhere else in the app.

You can also ask the assistant to open the page for you with a scope already set, for example "show me monitoring for Invoices and Receipts".

Reviewers working across several projects get the same checklist. What the wider team-wide permission adds is the in-flight reviews section, not the ability to combine projects.

Controls

• Count by switches the funnel between counting transactions and counting documents. A transaction can contain many documents, so the two answer different questions. The choice is remembered per scope.
• Compare to sets the baseline every funnel figure is measured against: yesterday, the same weekday last week, a 7-day average, or a 30-day average. The green or red figure under each number is today's difference from that baseline.

Reading the page

The sections are ordered so that the earlier one usually explains the later one. Work down, and report the first thing that is wrong rather than the loudest.

License

One traffic light for the licence, because an expired or exhausted licence stops intake and would otherwise show up here only as a page of zeros. The card shows the pages left in the current period (or in total, for a fixed-volume licence), the day they run out (the period end when pages reset, or the expiry date, whichever comes first), the expiry date, and one line saying why the light is not green.

• Healthy (green): nothing to do.
• Attention (yellow): the licence expires within 21 days, 80 percent or more of the period's pages are used, or the licence has not been refreshed from the license service recently. Processing continues.
• Action needed (red): the licence is expired, exhausted, missing, over its volume, or could not be refreshed within its offline grace period. New transactions are, or are about to be, refused.

The thresholds are the same ones that trigger the licence notification emails, so the card, the email and the user menu never disagree. Anyone who can open Monitoring sees the card; the Manage license link appears only for holders of the Manage license permission and opens the License page. The card hides itself when licensing is not enabled on the deployment.

Import connectors

Live health of the project's published import connectors. Check this first. A feed that has stopped explains every zero further down the page, and a low document count on its own cannot tell you the difference between "quiet day" and "nothing has arrived since Tuesday".

This section starts collapsed, showing how many connectors are running out of how many exist, plus a count needing attention. That headline is usually all you need; click it to open the full list, and it stays as you left it next time.

Each connector shows its control state, its health, and when it last imported. Two signals are worth knowing:

• Paused means somebody turned it off deliberately. It is not a fault.
• Stale means the connector is polling happily with no errors, but nothing has arrived in a long time. A dead feed reports no errors at all, so this is the only signal that catches it.

With the right permission you can Pause, Resume, or Retry a failed item from here, without publishing anything.

This section hides itself when the project has no published connectors.

Today's funnel

Where today's work is, in four numbers:

Received	Arrived today, by upload, connector, or API.
Auto-processed	Completed without needing a human.
Routed to review	Sent to a review step.
Completed	Finished today, whether automatically or after review.

A funnel that is entirely zero usually means intake is broken, not that throughput is low. Check the connectors section above before drawing conclusions.

By project

Shown when more than one project is selected. The same funnel numbers per project, plus how many items each one has in review right now, so you can see which project is dragging the aggregate.

Projects that did nothing today are hidden, so a long list stays readable. A link in the section header shows how many are hidden and brings them back. A project counts as active if anything moved today or if it has items waiting in review, so one with a backlog and no new arrivals is never hidden. The section itself can be collapsed.

Review queue health

How the review queue is holding up:

In review now	Waiting for a person.
Oldest waiting	How long the longest-waiting item has been there.
At risk	Will breach its service level soon.
Breached SLA	Already past it.
Assigned / Unassigned	Whether anyone owns the work.

Anything in Breached SLA is already late. A rising Oldest waiting combined with a high Unassigned count means nobody is working the queue, which is a staffing answer rather than a system one.

Team pace

Throughput over the last rolling hour: items per hour, and the average and median time an item spends in review. Below that, how many items each reviewer decided in that hour.

A median much lower than the average means most reviews are quick and a few are getting stuck.

Forecast

Queue depth divided by current throughput, projected forward to when the queue would clear at today's pace.

A projection of Unknown while the queue is not empty is a meaningful signal in itself: nothing cleared review in the last hour, so review has stalled. That is different from an empty queue, which also shows no projection.

In-flight reviews

Who is reviewing what at this moment, how long each item has been waiting, and how long it has been locked by its reviewer. Observe opens the transaction read-only so a supervisor can look without taking it over.

When more than one project is selected, this covers all of them in one list, ordered by how long each has been waiting, with a Project column. Observe opens the transaction in its own project, whichever project you were last working in.

This section needs the wider monitoring permission.

Memory learning

Whether the project is still learning from the work your team reviews, and if not, why not. Full background in Memory.

The row shows:

• Learning or Paused, plus who paused it and when.
• When a sample was last captured.
• A breakdown of the last seven days: of the transactions that completed, how many were captured and, for the rest, the reason each was skipped.

That breakdown is the point of the section. Memory can sit empty for weeks with nothing anywhere reporting an error, and the reasons are not equally serious:

Reason	What it means
Captured	Working.
Paused by operator	Deliberately stopped. Not a fault.
Accumulation turned off / Document examples disabled	Switched off in the published configuration. Change it in Memory settings and publish.
Reviewer not permitted	The person who approved the transaction lacks the "System can learn from user" permission, so their work was discarded. This is a permissions fix in Team → Roles, and it is by far the most common cause of empty memory.
No human reviewer	The transaction completed without anyone approving it, typically an unattended or API submission. Expected for those flows, not a fault.
Already captured	A re-run of something already learned. Normal.

The first three reasons that need a human are highlighted, with the fix named underneath. Pause and Resume are available to anyone with the "Pause and resume memory learning" permission and need no publish. Note that Resume returns the project to its published configuration rather than turning learning on; if the published settings have it switched off, resuming will not start it.

When more than one project is selected, this section shows one row per project, with a count of how many need attention.

Like the connectors section, it starts collapsed. The collapsed headline shows how many of the selected projects are actively learning and how many samples they captured over the reporting window. Click it to open, and it stays as you left it next time.

Permissions

Capability	Permission	Built-in roles
Open Monitoring and see the funnel, queue health, pace and forecast, for one project or several at once	View own monitoring, or the team-wide one below	Everyone, including Reviewer and Submitter
See in-flight reviews — who is holding which review right now	View all monitoring	Admin, Designer, Supervisor
See the import connectors section	View connectors	Admin, Designer, Supervisor
Pause, resume, or retry a connector	Control connectors	Admin, Designer, Supervisor
See the memory learning section	View memory learning status	Admin, Designer, Supervisor
Pause and resume memory learning	Pause and resume memory learning	Admin, Designer, Supervisor

Sections you do not have permission for are not shown at all, so the page shape differs by role. Whatever you select, only the projects you are allowed to see are included, in every section.

Ask the assistant

The in-app assistant reads the same data and can triage for you:

• "Is everything running normally?"
• "Why is nothing being imported?"
• "Which project is behind on review?"
• "Is this project still learning into memory?"

It can also act: pausing a misbehaving connector, assigning a stuck transaction to a reviewer, raising an item's priority, or pausing memory learning. It will ask you to confirm before anything that touches production.