Why Excel changes your numbers without telling you.

A postcode, product code or account number beginning with zero is a number as far as a type guesser is concerned, so 01234 becomes 1234. The cell looks fine. The join against another file silently matches nothing.

This is the one to check first whenever a lookup returns fewer rows than you expected. Sort the column and look at the shortest values: if some are one character shorter than the rest and start with a different digit, this happened.

The fix is at export rather than at import. Ask the source system for the column as text, or open the file through a dialogue that lets you set the column type before it is read rather than after.

Numbers are stored as double-precision floats and displayed to fifteen significant digits. A sixteen-digit identifier — a card number, a long order id, some barcodes — loses its last digit permanently, and the replacement is a zero.

The tell is a column of long numbers where a suspicious number of them end in zero. Nothing about the cell indicates that it was truncated, and once saved, the original digit is not recoverable from the file.

Anything matching a date-ish pattern is converted, which is how a product code like 3-10 becomes the third of October and how certain gene names became dates often enough that the field renamed the genes.

It is not reversible by formatting the cell back, because the underlying value is now a serial number and the original text is gone. Undo works; reformatting does not.

Large numbers are displayed as 1.23E+15, which is a display change and looks harmless. The rounding that happens when that is saved to CSV is not a display change.

The general rule underneath all four: a spreadsheet is a calculation tool that assumes its input is quantities. Identifiers are not quantities, and the safest thing you can do with a column of codes is to make sure nothing ever reads it as numeric.