AdemeroHelp

Paige extracted a blank or wrong field

Trace a missing or wrong field value through Paige's pipeline — grouping, recognized text, field guidance, confidence filtering, and export mapping — and fix the stage that dropped it.

Administrators25 minutesVerified 2026-08-30

Before you begin

  • Access to the scan job's settings (some settings areas in Paige appear only for company admins)

By the end of this guide you'll know which stage of Paige's pipeline lost or mangled the value — document grouping, text recognition, field association, confidence filtering, or export — and you'll have fixed that stage or collected exactly what support needs.

Check the stages in order

Paige processes every batch through a fixed pipeline. While a batch runs, the progress tooltip in the Batches grid lists the stages: Page Orientation, Text Recognition, Image Processing, Barcode Detection, Document Separation, Document Classification, and finally Data Extraction.

Field values come last — so a problem in any earlier stage surfaces as a blank or wrong field. Work through the checks below in order and stop at the first one that fails.

If...The problem is in
Pages landed in the wrong documentDocument separation
Paige can't read the text on the pageText recognition (source image)
Text is readable but the value is blank or wrongField guidance
The value is blank and inconsistently soConfidence filtering
The value shows in Paige but not at the destinationExport field mapping

1. Rule out wrong grouping first

If separation put pages into the wrong document, extraction is reading someone else's pages — no field setting can fix that.

  1. In Documents, right-click the affected document and choose View Document.

  1. Confirm every page belongs to this document, and that the page carrying the value is actually in it.

If the grouping is wrong, fix separation before touching anything else — see the separation guide under What's next below, then return here only if the field is still wrong.

2. Can Paige read the value at all?

Extraction can only be as good as the recognized text, so look at what Paige actually read.

  1. In Documents, right-click the document and choose Download.

    What you should see: either Download started! An email will be sent with link(s) when ready. or the file appears in the download panel as Preparing download....

  1. Open the downloaded PDF and try to select or search for the value (search for the amount itself, or a label printed near it). The output PDF's searchable text layer is the recognized text.

  1. Judge the result:

    • The value is selectable and correct — recognition is fine; go to step 3 of this guide.
    • The value is garbled or missing from the text layer — recognition failed. Get a cleaner source (higher scan quality, less skew, a better original) and run the pages through again.
Note

If Skip OCR is enabled on the Processing page of the scan job's settings, the output PDF has no searchable text layer — the AI reads the page images directly, and you can't inspect recognized text this way.

3. Text is readable but the value is blank or wrong: tune the field guidance

Unusual layouts defeat naive extraction even when recognition is perfect. A classic case: an amount printed to the left of its label.

The text layer is clean, confidence isn't the cause — and the field still comes back blank or zero, because the extractor expects the value below or to the right of the label. The fix is telling Paige where the value lives.

  1. Open the scan job's settings and choose Document Types, then edit the document type. The dialog is titled Edit Document Type.

  1. Select the field in the Fields list and expand Advanced Settings.

  1. Fill in AI Extraction Guidance with context and location, not just format — for example: The freight amount is printed to the LEFT of the word "Freight", on the same line.

Figure

Edit Document Type dialog with a field selected, Advanced Settings expanded, and AI Extraction Guidance describing a value printed left of its label

  1. Two more tools in the same Advanced Settings area, when they fit:

    • Custom Data Format Guide — describe the expected format, e.g. Jan 20, 2025 or ABC-7772.
    • Force Selection from Embedded List — match extracted values against an uploaded list of valid values (text, CSV, or XML).

  1. Save, then in Documents right-click the affected document and choose Rerun Auto Entry. This re-runs AI field extraction on that single document.

    What you should see: the field fills in on the Documents grid. Hover the value to see its AI Confidence: X% tooltip — values below 90% render in bold as a low-confidence flag.

Important

If a clearly readable value still comes back blank after two or three rounds of guidance improvements, stop iterating — some layouts need help beyond guidance text. Contact support with the evidence packet at the end of this guide.

4. A blank value may be filtered, not missing

Paige can discard values it isn't confident about. A discarded value is not shown as low-confidence — the field simply looks blank, with no visible trace.

  1. Open the scan job's settings and choose Processing.

  1. Find the card Fields: Confidence Threshold. When its toggle is on, AI-extracted values below the threshold on the Minimum Confidence slider are discarded. The slider marks are Low (30%), Medium (60%), and High (90%); the default is Medium (60%).

Figure

The Fields: Confidence Threshold card on the Processing tab, toggle on, slider at Medium (60%)

  1. Temporarily lower the threshold (or switch the toggle off), then right-click the document in Documents and choose Rerun Auto Entry.

    What you should see: if filtering was the cause, the value now appears — in bold in the Documents grid if it's below 90% confidence, and in Data Entry Mode with a red-to-orange tint on the field box (redder means lower confidence).

  1. Don't leave the threshold low as the fix. Improve the field guidance (step 3 above) so confidence rises, then restore the threshold you want.

5. The value shows in Paige but not at the destination: check export mapping

If the Documents grid shows the correct value but the destination is blank, the value died between Paige and the destination — almost always an unmapped field.

  1. Open the scan job's settings and choose Export. Open the connection card for your destination: Nucleus One, Content Central, or FTP.

  1. For Nucleus One: the setup wizard's Field Mapping step shows Field Mapping for your document type. Confirm the field is mapped — an unmapped field extracts perfectly in Paige and simply never arrives.

  1. For Content Central: in the Field Mapping section (shown once a catalog and Content Central document type are selected), find the field in the Paige Field column and choose its match in the Select Content Central Field dropdown.

    What you should see: Content Central Field Mapping Updated! when you save.

  1. Export the document again and confirm the value arrives at the destination.

Still stuck?

Send support this compact evidence packet — one document is enough:

  • The document itself (right-click > Download in Documents)
  • The exact field name as it appears in the document type
  • The value you expected Paige to extract
  • What the downloaded PDF's text layer shows for that value: correct, garbled, or missing
  • Your Fields: Confidence Threshold setting (on or off, and the percentage)