Review mixed numeric conventions within one combined register

A combined column can contain several conventions

Resolve them per source before conversion so a single cleanup rule does not corrupt part of an otherwise useful register.

Test an invoice with the dates, quantities or amounts you find hardest to copy accurately. ParseMill puts extracted invoice fields beside the source for review and correction before export. Open the downloaded workbook in your usual spreadsheet application and check those values there too.

The profile should describe decimal marker, grouping marker, sign conventions and any relevant qualifiers. It should also identify unsupported cases. A source-wide profile may be too broad when a report includes imported attachments or tables from different systems.

Create a per-source notation manifest

These invented examples produce equal derived values only where the convention is explicit. The unresolved C value must not inherit A’s or B’s profile. The appendix demonstrates why file identity alone may be insufficient to determine notation.

Microsoft: Number.FromText supports culture-aware interpretation of numeric text. Choose a conversion from the verified profile and preserve the profile identifier with the derived value.

Normalize by evidence-backed groups

1. Retain source document, section and raw result on every row before consolidation. Create a profile manifest with supporting legend or issuer documentation.
2. Assign profiles to reviewed ranges. Put ambiguous rows and profile conflicts into an unresolved queue with the evidence needed to decide them.
3. Convert copies of the raw strings within each profile. Test representative values with decimals, grouping, signs and qualifiers before applying the rule broadly.
4. Reconcile converted, unsupported and unresolved counts to the input rows. Compare derived magnitudes back to known fixture expectations.

Microsoft: Table.TransformColumnTypes accepts explicit culture for column conversions. If one column contains several conventions, split the processing by verified groups rather than relying on the computer’s default locale.

Detect profile drift

A new monthly report may change notation even if its filename and issuer are familiar. Check the current source before reusing a profile. Record profile versions and keep old outputs tied to the rule that produced them.

Do not use a conversion success rate as the only quality measure. An ambiguous string can convert successfully under several cultures with different meanings. Include source-evidence checks and magnitude comparisons, while treating unexpected magnitude as a review flag rather than permission to guess.

Publish a transparent analysis layer

For the fixture, three supported rows can be normalized under their specific profiles; Report C remains text awaiting clarification. The raw register remains unchanged, and the manifest explains each derived value. Share the unresolved queue with the intended analyst.

Use invoice PDFs from suppliers with different numeric conventions in your ParseMill trial. Compare each amount with its source before export and check the workbook in your receiving application. Keep any broader cross-report normalization as a separately reviewed step.