CSV imports: preserve identifiers and dates before automating

Treat the file as data with a contract

Before importing a customer or operations CSV, ask what each column means. An account identifier can consist entirely of digits while still needing exact text preservation. A date requires an agreed order and interpretation. Record the delimiter, expected headers, encoding and the application's handling of empty fields before preparing a full load.

RFC 4180 documents a common CSV format, including quoting conventions, while explicitly acknowledging differences between implementations. It does not supply the business meaning of your columns or promise identical behavior in every spreadsheet and importer. Use it to reason about structure, then check the receiving application's requirements. [1]

Control the spreadsheet import step

If you inspect a CSV in Excel before uploading it elsewhere, that inspection can become a transformation. Microsoft's Text Import Wizard documentation describes choosing column data formats, including Text and date interpretations. Use the import controls available in your version rather than assuming a double-click preserves the original values. [2]

Keep an untouched source copy. Work on a separately named inspection copy and note whether you save it again. If a tool has already changed a value, changing its visual formatting afterward is not evidence that the original information has been recovered. Return to the source when the intended value is uncertain.

Worked example: five rows that expose different failures

Our invented acceptance set includes identifier 00127, ambiguous date 04/05/2026, unambiguous date 2026-05-04, the name River, North, and a blank value. The expected identifier is exactly 00127. The ambiguous date has no accepted interpretation until the source owner confirms the order. The comma-containing name must remain one field.

For the final date, the intended meaning is 4 May 2026. Record the destination's actual value, not merely a screenshot of a formatted cell. A tool may store one representation and display another. The blank case should remain blank unless the agreed mapping explicitly supplies a default; silently turning it into zero changes its meaning.

Preview mappings before importing the full file

Create a small private destination or use the application's documented preview mode. Map each source header deliberately and inspect the fixture. Check whether the importer creates new records, updates matching records or both. The same values can produce very different business effects depending on its matching key.

Compare the number of parsed rows and columns with the expected structure. RFC 4180 describes enclosing fields containing commas, line breaks or quotes in double quotes and doubling embedded quotes. If the preview shifts a name into two columns, repair the export or parser configuration before touching the larger dataset. [1]

Decide what counts as an accepted import

Use one worksheet row per edge case and record source value, expected interpretation, observed destination value and result. Include a sample of non-English characters from your actual authorized dataset if relevant. Never publish real personal information in a teaching fixture or upload it to an unrelated conversion website.

Proceed only when the required cases agree and the update behavior is understood. After the real load, reconcile counts and selected stable identifiers, then check one downstream workflow that uses them. This is a proposed acceptance procedure, not a tested integration or a claim of universal CSV compatibility. If an importer cannot preserve a required value, retain that limitation and choose a supported transfer method instead of guessing.

Sources & verification

Product details and prices can change. Check the linked provider before buying.

1. RFC Editor / Y. Shafranovich: RFC 4180: Common Format and MIME Type for CSV Files Accessed 2026-09-14
2. Microsoft Support: Text Import Wizard Accessed 2026-09-14