Reading an Extraction Receipt

What it is

Every invoice, certificate, estimate, scorecard, or other file that flows into Verinode gets read by the extraction pipeline: it classifies the document, matches it to the right vendor, client, carrier, or job, and pulls out the fields that document type is expected to carry (amount, dates, reference number, and more). The extraction receipt is the audit trail for that read. It is attached to every parsed document and answers three questions for that one file: what did Verinode capture, what did it miss or get wrong, and how complete is the read overall. That last number, the completeness percentage, is what drives the colored badge you see next to the document in your list.

Verinode does not decide what a document means on your behalf. It reads what is on the page, shows you exactly what it captured, and hands you a fast way to correct or retry anything it got wrong. You stay in control of every value that ends up feeding your benchmarks, your margin figures, and your signals.

Where to find it

Open Vault from the My Data section of the sidebar (iq.verinode.ai/data). The Vault home opens with a hero panel labeled "What Verinode Is Tracking," an Explore row of tiles (Uploads, Tracking, Playlist, Reading, Saved, Notifications), and a Most Recent row of your latest captures. Click the Uploads tile, or click any document tile in Most Recent, to open the document list where every extraction receipt lives.

Every document in that list, invoices included, carries its own receipt. There is no separate "Invoices" list: invoices are one document type among many (certificates, estimates, statements, assignments, payment notices, carrier scorecards, and others) that all pass through the same extraction pipeline and get the same receipt treatment.

The document list: what each column means

The list is a table, one row per document, with a checkbox column on the left for bulk actions and these columns across it:

• Source. The vendor, carrier, or job the document is linked to, as a clickable link back to that entity's page, with a small caption underneath reading Vendor, Client, or Job. Documents Verinode could not confidently link show a caption of Unknown, or, when you filed the document under a category yourself, that category's name instead.
• Type. A pill showing the document type in plain English (for example "Renewal Notice" rather than the raw type on file). This is a humanized label, never the underlying database value.
• Amount. The extracted total, formatted as currency with no cents (for example $4,200). Documents where no amount was extracted or applicable show a dash instead of a number.
• Processed. The date the document was parsed, or, if it has not been parsed yet, the date it was uploaded.
• Signals. How many platform signals this document has triggered. A number greater than zero is a clickable link to /signals. Zero shows as a dash.
• Status. Two things live in this column side by side: a status dot (Processed in green, Pending in amber, Failed in red) and, when the document has an extraction receipt, the completeness percentage badge described below.

The completeness percentage and its badge

Completeness is the percentage of extractable fields Verinode actually captured for that document, not the model's own confidence score. If a document type typically carries five extractable fields and Verinode successfully pulled four of them, that document's completeness sits around 80%. It is a measurement of what came through, not a guess about how sure the system feels.

The badge in the Status column is colored against three tiers, and hovering it shows you exactly what the color means:

• 80% and above (green): "We captured most of this document. Click to review."
• 40% up to 80% (amber): "Some fields are missing. Click to review or correct."
• Below 40% (red): "We struggled to read this document. Click to correct it."

Click the badge at any tier to expand the receipt inline, right below that row.

Two different flags: needs review versus a low completeness score

Two separate signals can appear on the same row, and they mean different things:

• Needs review is an amber pill next to the source name, reading "Assign vendor," "Assign carrier," or "Assign job." It means Verinode found a document that clearly belongs to your business but could not confidently match it to an existing vendor, client, or job record. Clicking it (or clicking anywhere that expands the row) opens an assignment panel where you point the document at the right record.
• The completeness badge is about the quality of the read once a document already has a home. It has nothing to do with whether the document is linked to the right entity.

A document flagged Needs Review will always open its assignment panel first when you expand the row. Once you assign it to the correct vendor, carrier, or job, expanding that same row switches over to its extraction receipt, since there is no longer an entity question in the way.

Opening the receipt: "What we read from this document"

Clicking the completeness badge on an already-matched document expands a panel titled What we read from this document, right beneath the row. Inside it:

• We captured: a plain-English, comma-separated list of every field and capability the extraction pulled from that document (for example amount, period start, period end, reference number).
• If any part of the extraction failed outright, a line reads: "Some fields didn't come through cleanly. Edit the values yourself, or re-run if you've updated the source document."
• If nothing technically failed but the completeness score is still under 80%, a softer line reads instead: "Verinode wasn't fully confident on this one. You can correct the fields directly, or re-run extraction."
• Two buttons sit at the bottom: Edit fields and Re-run extraction.

Correcting a document

Edit fields opens (or the pencil icon at the far right of the row opens directly, without going through the receipt panel first) an inline edit form with six fields:

• Document type, a dropdown of the manually correctable types: Invoice, Renewal Notice, Certificate, Estimate, Email, Statement, Assignment, Payment Notice, Job Export, Carrier Scorecard, Incident Report, and Other.
• Total amount ($)
• Period start and Period end
• Document date
• Reference number

1. 1Click the pencil icon on the row, or click Edit fields from an open receipt.
2. 2Adjust any of the six fields. Leave a field blank to clear it.
3. 3Click Save corrections.
4. 4The panel confirms "Corrections saved." and the row refreshes with your values a moment later.

Click Cancel at any point to close the panel without saving.

Re-running extraction

Re-run extraction asks Verinode to read the source document again from scratch. Use it after you have replaced the underlying file (through the eye icon, described below) or when you believe the original read was simply a bad pass. On success, the page reloads and the document's receipt reflects the new read. On failure, you'll see an alert with the error, or, if the request could not reach the server at all, a prompt to check your connection.

Viewing, editing, and removing the original file

Three icon buttons sit at the right edge of every row:

• Eye icon, View original document. Opens the source file in a new tab.
• Pencil icon, Edit extracted fields. Opens the same edit panel described above.
• Trash icon, Remove document. Expands an inline confirmation: "Remove this document? This will also remove any signals sourced from it. This cannot be undone." Confirm with Remove, or back out with Cancel.

Selecting and removing documents in bulk

Check the box on any row, or the header checkbox to select every row currently in view, to select multiple documents at once. A floating bar appears at the bottom of the screen showing how many documents are selected, with Remove selected (which deletes all of them, same permanence warning as a single delete) and Clear (which deselects everything without deleting).

Filtering and searching the vault

Above the table, a search box and two dropdowns narrow what you see:

• Search, placeholder "Search by vendor, carrier, job, or doc type…". Matches against the source name and the raw document type.
• Bucket, "All buckets (N)," "Vendors (N)," "Clients (N)," or "Jobs (N)," where N is the live count in each group. Clients here covers documents tied to a carrier or TPA relationship (scorecards, payment notices) as well as true client records; Jobs covers job exports, assignments, and job-adjacent compliance documents like certifications and team rosters.
• Type, "All types" plus one option per document type actually present in your data right now, listed alphabetically with humanized labels.

If your filters and search leave nothing to show, the table is replaced with an insight preview instead of a blank space. The default heading reads "Your document intelligence builds as files flow in," followed by a short list of what starts appearing once documents flow in: automatic classification into types like invoices and certificates, entity matching to the right vendor, job, or carrier, extraction of amounts, dates, and terms, automatic anomaly and signal detection, and a full audit trail on every document.

If your account has fewer than five documents captured overall, a prompt also appears above the table: "Set up email forwarding, Auto-forward vendor invoices, COIs, and scorecards to your intelligence address. Takes 2 minutes," with a Set up button that goes to /connect. See Connecting your data and Forwarding documents for the channels that feed this list.

The completeness percentage in aggregate: the Vault hero panel

The badge on a single row is the document-level view. The Vault hero panel at the top of /data rolls the same idea up across your whole account:

• The headline number is your total documents captured, with a colored pill next to it reading "N% Extraction Completeness" (green at 80% and above, amber from 40% up to 80%, red below 40%), calculated as the average completeness across every document that has a receipt. The pill is hidden until you have at least one document.
• A line of context underneath adapts to your state: a prompt to connect a channel if your vault is empty, a note about how many documents need a quick review if any are flagged, a note that documents are still processing if any are pending, or, once everything is caught up, a reminder that this vault feeds every section of your business: vendors, clients, jobs, equipment, team, certifications, and more.
• Two secondary tiles sit beside the headline: Signals Fed (documents that triggered platform intelligence) and Needs Review (entities still waiting on an assignment, reading "all matched" once that count hits zero).

The Uploads tile in the Explore row repeats your total document count with a status mix underneath (documents that fed a signal, documents needing review, documents still pending), and the Tracking tile shows how many distinct vendors, clients, and jobs you have documents against, along with the average completeness across that tracked set. Opening Tracking is how you check completeness by entity rather than by individual document, useful when you want to know how well-documented one vendor or job is overall rather than reading receipts one at a time.

Best-practice example

A vendor invoice comes in through email forwarding. Verinode classifies it as an Invoice, matches it to the right vendor, and extracts the amount, the document date, and a reference number, but the invoice's billing period spans an odd date range it cannot confidently parse. That document lands at 75% completeness: an amber badge, tooltip reading "Some fields are missing. Click to review or correct." Clicking it opens the receipt: We captured amount, document date, reference number. No hard failure, just the softer confidence line. Since the period dates are visible on the invoice and simply weren't machine-readable, the fastest fix is Edit fields: type in Period start and Period end directly, save, and the row updates on its own. No need to re-run extraction here, since the source file has not changed.

Related reading

• Connecting your data
• Forwarding documents
• Understanding your margin
• Clients and carriers
• The decision workspace

Data sources