The Clean-Looking Mistakes Are the Ones I Worry About

What this note is about

A messy extraction is easy to distrust. The harder problem is a spreadsheet that looks tidy while a sign, decimal or wrapped description is wrong.

Bad output is not always obvious

The obvious failures are not the ones that bother me most. If the file has empty rows everywhere or page headings in the middle of the table, nobody trusts it.

The more dangerous failure is a neat-looking spreadsheet where a debit became a credit, a decimal moved, or the second line of a description got split into its own row.

I check the boring parts first

The first row and last row tell you whether the converter found the right table boundaries. Balances, when they exist, are useful because they make small mistakes show up quickly.

That is why I keep telling users to review the preview. It is not just a legal-sounding warning. It is how you catch the tidy mistakes.