Features

The complete rule list

Most tools in this space describe themselves as intelligent. This is the whole list of what gets checked, published rather than hidden — you should be able to read the rules before you trust a report.

1 · Missing information

Required RFQ fields that nobody wrote down, and required per-part fields.

  • RFQ number, customer, contact, response deadline, requested delivery date, ship-to, part list, quantities, material, revision, tolerances, surface finish, packaging, inspection, certification and any customer-specified terms.
  • Per part: a part must have an identifiable part number, a quantity and a revision before the rest of the fields mean anything.
  • Cross-checks, which is where this earns its keep: a delivery date with no ship-to is reported as a freight blocker rather than an empty field. A quantity with no unit. A tolerance that is a general note rather than a callout. A certification required for a material nobody specified.

2 · Conflicting information

The same field stated two ways in two files.

  • Nothing is merged and no winner is chosen. Both statements are shown, each against its own source line, and the contradiction is left for you.
  • Restatements that mean the same thing are clustered rather than reported — "5052-H32 ALUMINUM" and "ALUMINUM 5052-H32" are one material, not a conflict.
  • Conflicts are scoped to the part they affect, so one bracket disagreeing with the email does not mark the whole package as contradictory.

3 · Revision mismatch

A part at several revisions across the package, and documents that look superseded.

  • The revision on a drawing title block is compared against the revision named in the email body and in the BOM.
  • A part appearing at Rev A, Rev B and Rev C is reported once, with every file listed.
  • A later revision of a part does not quietly supersede an earlier one. Both are reported; which one you quote is your call.

4 · Attachment problems

Nothing anyone else in this category treats as a first-class finding: whether the files themselves are fit to quote from.

  • Unreadable, empty, and zero-byte files.
  • Scanned PDFs and photographs that carry no text layer.
  • Byte-identical duplicates, and the same content uploaded twice under two different names.
  • Unsupported file types.
  • Drawings with no material block and no revision block — a drawing you cannot identify the material of is not a drawing you can price.
  • Part numbers that cannot be attributed to any file.
  • Files containing instruction-like text, which is treated as data and never as a command.

5 · Ambiguity

Requirements that can honestly be read two ways.

  • "As per previous", "standard finish", "as discussed", "TBD", "refer to attached" with nothing attached.
  • A stated total that cannot be allocated to any part in the package.
  • A tolerance given as a general note with no callout to tie it to a dimension.
  • These are flagged for manual confirmation. They are never resolved, guessed, or picked apart for you.

The rule that outranks all of them

Every value in a report is evidence-gated before it is shown.

  • The cited source segment has to exist in the parsed file.
  • The quote has to appear verbatim in that segment.
  • The value has to actually be supported by that quote.
  • Fails any of the three and the value is dropped. The report shows a dash, which means "not established" — never "this field is absent" and never a guess.

What comes out the other end

  • A per-field report where every value carries a source link opening the exact line, page, sheet or cell.
  • Findings grouped by severity: critical for contradictions inside the package, missing for things nobody wrote down, warnings for file-level problems.
  • A document inventory with a per-file status, so you can see at a glance which attachment is the problem.
  • Processing notes — including a visible notice if the run was degraded because no model provider was reachable.
  • Exports to PDF, Excel and CSV, so a report can be attached to a clarification email without being rebuilt.

See the sample reports for what these look like in practice.