Independent tool / Browser-only analytics
Data quality lab
Profile a CSV. Preview the cleanup. Export the result.
Find missing values and duplicate rows before they distort your analysis. Start with the synthetic sample or use a local CSV. Your file stays in this browser tab.
A file, or a few rows.
UTF-8, comma-separated, first row as headers. Up to 2 MiB, 5,000 data rows, and 50 columns. Quoted commas, quotes, and multiline fields are supported.
Paste or inspect the CSV
Before → proposed cleanup
Missing cells include blank and whitespace-only values. Duplicate rows count additional exact copies across every column; whitespace matters until you choose to trim it.
Changes run in order: trim, drop blank rows, then remove duplicates. The original stays in memory unchanged. Uncheck an option to compare again.
Every change has a reason.
Nothing changed yet. Choose a cleanup above to inspect the affected rows.
Selected row
Data row numbers exclude the header and count quoted multiline records once. This inspector shows cleanup; formula protection is added at export.
Original rows
Proposed rows
Inspect every column
Profile of the proposed dataset before export protection. Numeric-looking values are inferred, never converted. Min, max, and mean use finite numeric values only; mixed columns show how many values were excluded. A numeric-looking identifier may still be an identifier.
Export what you inspected.
CSV quoting alone does not prevent spreadsheet formulas. Export adds a leading apostrophe
to cells and headers beginning with =, +, -, or @, including after whitespace, and to leading tab or line-break values.
This protection can change spreadsheet interpretation, including making negative numbers text. The preview shows cleanup; the export adds these protective prefixes. Treat externally supplied files carefully in the spreadsheet you use.
Independently developed for this portfolio with AI assistance. The sample uses synthetic manufacturing-style records, including deliberate duplicates, blanks, mixed values, and a formula-like note. It contains no employer or customer data. Processing is deterministic: no model calls, uploads, cookies, or browser storage.