Start with a rectangular dataset
CSV has no rowspan or colspan. Every record needs the same fields. Resolve merges before serialization, or one wide header will quietly ruin the rest of the file.
Headers matter too. An empty top-left cell can still be the hinge between row labels and data columns. Name every useful column before export.
CSV quoting and encoding rules that prevent broken imports
Any value with a comma, quote, or line break belongs in quotes. Quotes inside that value get doubled. Skip those rules and one product description shoves every later field one column to the right.
Use UTF-8 so names, symbols, and non-English text survive. Some locales open files with semicolons by default, but standard CSV stays comma-separated unless the receiving system says otherwise.
A practical conversion checklist
Run these checks while the captured table is still next to the source page.
- Confirm each row has the expected number of columns.
- Remove page chrome, totals, ads, and notes that got captured as rows.
- Inspect values with commas, quotes, and line breaks.
- Decide whether leading zeroes and long IDs must stay text.
- Open the CSV in a text editor and in the target spreadsheet or import tool.
CSV, TSV, or Excel
CSV travels well for imports and scripts. TSV is handy on the clipboard because tabs rarely show up in ordinary values. Excel wins when people need worksheets and familiar editing.
A good capture workflow produces all three from the same reviewed dataset. Keep formatting choices separate from extraction accuracy.