Encoding & dialect
BOM, encoding, delimiter, quote/escape and newline consistency.
Validate more than CSV syntax
A file can be valid CSV and still be dangerous to import. CSVProof checks structure, parser assumptions, record identity and values that familiar tools may silently reinterpret.
Private early access · macOS & Windows
DataTransformCompareIssues 6RecipeRun report
Duplicate customer_idChoose the row to keep
Keeprowcustomer_idname
●177001177Meridian Supply
○902001177Meridian Supply AS
From bytes to business keys
Each finding includes its location, the rule it failed and what CSVProof did, not a generic “invalid file” message.
Record structure
Malformed quotes, missing or extra fields, inconsistent width and empty headers.
Value risks
Leading zeros, long identifiers, date-like text, invisible characters and formula-like values.
Data contract
Required columns and values, allowed patterns and ranges, unique or composite keys and lookup rules.
Runs locally in your browser
Get validator accessFree CSV validator
The local integrity checker will share this page with clear, documented tests.
No file upload. No CSV content in analytics. The tool itself is reserved for the dedicated implementation round.
A CSV checker needs rules for the data
Syntax is only the first layer. Save the expectations that make this particular export safe to accept.
CheckExamplesDecision
ColumnsRequired, optional and unexpected headers; order changes and schema drift.Accept or stop
KeysUnique or composite identity, duplicates, nulls and lookup cardinality.Review ambiguity
ValuesRequired values, patterns, ranges, lengths, allowed sets and explicit types.Preserve, fix or reject
ProfileNull and distinct counts, minimums, maximums and suspicious distribution shifts.Investigate drift
OutputEncoding, delimiter, quoting and newline normalization before the file is written.Preview the change
A warning should explain itself
CSVProof distinguishes a malformed record from a value that is valid but easy to damage elsewhere. You decide whether to reject, preserve or transform it.
Interpretation riskRow 184 · account_id
00000000000000000042This 20-character numeric-looking value begins with zeros. Treating it as a number would change its identity.
Preserve as textAdd rule
Turn a good inspection into a contract
After reviewing one delivery, save the parser and validation rules. The next run fails visibly when the delimiter, schema, key uniqueness or required fields drift.
Saved contract · v28 columnscustomer_id uniquecountry required
Incoming file9 columns +13 duplicate IDs !1 blank country !
Catch the exceptions before the next system does
CSVProof is currently in private early access for macOS and Windows.
Get early access