Fix parsing problems

Updated July 11, 2026

You dropped a carrier quote and the deck that came back is not what you expected. Here is how to tell what happened and how to fix it.

#The exact failure message

When the deterministic parser cannot pull enough fields out of a dropped PDF, you get this toast:

Could not deterministically read [filename]. The parser reads text-layer PDFs. Showing the sample flow instead.

If you see this, the parser tried and failed on that specific file. Nothing about your Vault or account is broken, the document you dropped just was not readable.

#What causes it

A scanned or image-only PDF. The parser reads a PDF's text layer, the underlying characters embedded in the file. A scan, a photo of a document saved as a PDF, or a fax-quality image has pixels, not text, so there is nothing for the parser to extract. This is the single most common cause of a failed parse.

A file that is not the carrier's original download. If a quote has been re-printed to PDF from a screenshot, flattened through certain print drivers, or otherwise stripped of its text layer along the way, it behaves the same as a scan even though it started as a real digital document. The original PDF you download directly from the carrier's or wholesaler's portal is almost always the safest bet.

Too few fields found. Even on a text-layer PDF, if the parser cannot find a named insured or at least one premium line, it treats the read as a failure rather than guessing. This is a deliberate gate: showing you nothing extracted is safer than showing you a deck built on partial garbage.

#What the app does on failure

It does not show you a blank error page. Instead:

  1. It shows the toast above.
  2. It falls back to the seeded sample proposal so you still see a complete deck.
  3. It logs the failure to your activity feed as "Parse of [filename] failed, showed the sample deck," so you (or your agency owner) can see it happened.

This means if a proposal ever looks unexpectedly like a demo client rather than your real prospect, the most likely explanation is a failed parse that silently substituted the fallback. Check your Recent activity on the dashboard, described in The dashboard and activity feed, if a deck ever looks off.

#What to try

  • Get the original file from the carrier's portal. Re-download the quote rather than reusing a forwarded, printed, or re-scanned copy.
  • Confirm it has a text layer. If you can select and copy text out of the PDF in a normal PDF viewer, it has a text layer and should be readable. If you can only select the whole page as an image, it is a scan.
  • Try the sample PDFs to confirm the parser itself is working. The New proposal page's drop zone links to a real sample carrier quote you can download and drop straight back in. If that parses successfully, the parser is fine and the issue is specific to your file.
  • Drop the current policy separately. If only your current-policy comparison PDF fails, the quote can still build on its own. You just will not get the side-by-side comparison slide. See Add the current policy for a comparison.

#How to verify a successful extraction

Once a PDF does parse, do not just trust the deck at a glance. Every extracted number carries the page it came from. Open the Analysis panel from the deck toolbar to see each field's page citation, formatted like "Quote p.5," next to the value the parser pulled. This is internal only, your client never sees it, but it is the fastest way to confirm the parser read the right numbers off the right pages before you present anything. See Page citations for how citations work and What gets extracted from a quote for the full field list.