Re-import CSV output with explicit column types

CSV text does not declare spreadsheet intent

Set the intended column types before loading so identifiers, dates and amounts survive the handoff intact.

ParseMill exports reviewed invoices to Excel with separate summaries, line items, sources and review status. You can import an XLSX or CSV template and save a client setup. Run a real invoice through to the final workbook and check it in the application that receives your work.

Do not assume that quoting a numeric-looking identifier forces every spreadsheet to treat it as text. Test the actual import path, because delimiter parsing and type inference are separate decisions.

Use a synthetic type map

The fixture intentionally includes two representations of the same result. The raw field preserves the source’s trailing zero, while the analysis field supports arithmetic. The date-like batch code remains an identifier and should not be interpreted as a calendar date.

Microsoft: Text/CSV connector exposes type-detection controls and allows imported columns to remain text. Inspect these settings before loading rather than repairing accidental conversions afterward.

Run a controlled import

1. Save the reviewed CSV unchanged and record its encoding, delimiter and expected column names. Prepare the type map alongside it.
2. Use an import workflow with a preview and explicit types. Check generated type-change steps and keep raw fields as text from the first conversion boundary.
3. Compare imported headers, row counts and exact identifiers with the CSV. Inspect numeric fields against independently specified expected values.
4. Save, close and reopen the receiving workbook. Repeat the critical checks and record the application version and import settings.

Microsoft: Keeping leading zeros and large numbers describes text-preserving import for numeric-looking identifiers. If digits were already lost during conversion, return to the original CSV instead of trying to recover them with formatting.

Handle mixed or unsupported columns

A column containing both numeric results and qualifiers such as <0.5 should not be forced into one ordinary number field. Preserve raw text and create a separate reviewed numeric representation where supported. Keep conversion errors visible with their source rows.

When a monthly file changes its schema, review the type map before reuse. A familiar column position can now contain a different field, so map by verified names and meaning rather than position alone.

Check the intended result

The fixture passes when 00127, 03-04 and 42.60 remain exact text and the analysis value is separately available as a number. No unresolved state should disappear into an empty cell. The import settings belong in the handoff instructions. Include a small known-good fixture so the recipient can verify those settings before loading the complete file.

ParseMill’s invoice batches download an Excel workbook. Use this type map if your downstream workflow converts that workbook to CSV, then import the CSV through the recipient’s actual process. Include that extra step in the buying test rather than assuming the invoice workspace exports your final CSV schema directly.