Cleaning messy CSV data in the browser
A practical pass over duplicate rows, filters, sorting, column selection, and CSV to JSON conversion.
A CSV export from a CRM, a billing system, or a spreadsheet is rarely clean. Rows repeat, columns are in the wrong order, blank lines appear in the middle, and numbers arrive as text. The good news is that cleanup follows a predictable order, and every step can happen in your browser.
Start by understanding the shape
Open the file in the Spreadsheet editor or the Sort CSV preview to check the basics: one header row, consistent column count, and a delimiter that matches the content. CSV supports commas, tabs, semicolons, and pipes; all rows must have the same number of fields.
A cleanup order that works
- Select the columns you actually need with CSV column picker. Fewer columns mean fewer places for errors and a faster workflow.
- Remove duplicates using key columns rather than the whole row, and choose whether the first or last occurrence wins with Remove duplicate CSV rows.
- Filter out noise with Filter CSV rows. Text and numeric conditions are separate because numeric comparison requires finite numbers in every cell.
- Sort deliberately using Sort CSV. Text sorts by Unicode order, so normalize case first if that matters.
- Reshape if needed. Transpose CSV swaps rows and columns when a table was laid out for reading instead of processing.
Convert only after cleaning
Cleaning first makes the conversion predictable. Use CSV to JSON when an API or script needs structured data. Values stay strings by default so IDs keep their digits and leading zeros; enable numeric conversion only when the data really is numeric. Going the other way, JSON to CSV explains how objects, rows, and primitive arrays map to columns.
Limits and gotchas
The browser workspace accepts up to 10,000 rows, 100 columns, and 100,000 cells, with 500,000 characters of pasted input. Large integer identifiers should stay as text. Quote handling follows the usual CSV rules, but spreadsheet programs differ in how they interpret exported values, so reviewing results before opening them elsewhere is worth the minute it takes.