Your KnowDoc dashboard, explained
Five status cards, a week of volume, and the two donuts that tell you what went wrong.
The greeting and the period filter
The page opens on a greeting and a period filter set to This Week. Everything below it — the counts, the bars, the donuts — is scoped to that window, so the first thing to check when a number looks wrong is the filter. Widen it to compare a slow week against a busy one; the layout doesn't change, only the scale.
Five status cards
The row of counters is the pipeline in miniature: how much arrived, how much made it all the way through, and how much stalled. Read them as a funnel rather than five independent numbers — a spike in one is always paid for somewhere else.
- Total documents — Everything that landed in the window, whatever the outcome.
- Processed — Read end to end with a usable extraction attached.
- Needs review — Extracted, but waiting on a person to confirm before it bills.
- Failed — The read did not complete — bad file, unreadable scan, no carrier match.
- Rejected — Deliberately turned away: not a billable document at all.
Last 7 Days Volume, and who is sending it
The volume chart shows the shape of your week — most freight billing arrives in bursts tied to carrier billing runs, not evenly across seven days. Beside it, the Senders table ranks the addresses actually feeding the pipeline. That table is the fastest way to find a carrier who suddenly stopped invoicing, or a new one who has started and hasn't got an extraction profile yet.
Recent Extractions
The live stream underneath shows each document as it finishes: the file, its confidence score, its status, and when it landed. Confidence is the number worth watching. A run of 100% reads on a familiar carrier means the profile is doing its job; a drift downwards on one sender usually means that carrier has changed their invoice template.
The two donuts: Failed and Rejected
Failures and rejections are broken out separately because they mean different things and have different fixes. A failure is KnowDoc saying "I couldn't read this"; a rejection is KnowDoc saying "this wasn't something to read." Chasing the first improves extraction accuracy. Chasing the second usually improves what people email in.
- Failed — Fix at the source — usually a missing carrier profile. See the AI Agent guide.
- Rejected — Fix upstream — usually signatures, logos, and other inline images.
The rejection table tells on itself
In the filmed week, 73 of 77 rejections were inline images — email signature logos, scanned letterheads, and the like, pulled in as attachments and correctly discarded. That is not a problem to solve so much as a number to recognise: once you know what the rejection pile is made of, you stop investigating it every Monday.
How to use the page in two minutes
Read it in one direction: cards for the shape of the week, volume and senders for who drove it, Recent Extractions for whether quality is holding, then the two donuts for what to fix. Anything the dashboard surfaces as a repeated failure reason has a home elsewhere in KnowDoc — the Carriers page for extraction profiles, the Flagged and Review queues for individual documents.
Ready to build your own?
Start free and ship your first app in minutes — or have us set it up with you.
More guides in Getting Started
More resources