Free tools // 01 · Validator
Data Validator
Drop a CSV or JSON file, get a quality score in seconds — missing values, duplicates, outliers, mixed types, whitespace problems. Nothing uploads. The whole engine runs in your browser.
- 100% client-side — your data never leaves your browser
- Instant results for files up to 5 MB
- Export the validation report as JSON
Validation report
You can't fix what you haven't measured.
What it checks
7 checks · every runSeven checks run against every file. That's not paranoia — 94% of spreadsheets used in business decisions contain errors (Frontiers of Computer Science, 2024). Yours probably isn't the exception.
| Check | Flags as | Triggered when |
|---|---|---|
| Missing values | warning error | Any empty cells; escalates to an error past 50% of a column |
| Mixed types | error | Non-numeric values inside a column that reads as numeric |
| Empty rows | error | Every field in a row is blank |
| Duplicate rows | warning | Two or more rows identical across all columns |
| ID collisions | warning | Repeated values in a high-cardinality column (>90% unique) that looks like an identifier |
| Outliers | warning | Numeric values more than 3σ from the column mean |
| Stray whitespace | info | Leading, trailing, or doubled spaces inside values |
Every column also gets profiled: inferred type (numeric, date, or text), missing and unique percentages, and min / max / mean for numeric columns.
How the score works
100-point scaleSimple arithmetic, no black box. You start at 100 and lose points for what we find. The same scoring backs the before/after reports we ship with client work — so you can check our work with our own tool.
Starting score for every file
Per error — structural problems that break analysis
Per warning — issues worth a look before you trust the data
Per info item — cosmetic problems, easy fixes
Rough reading: 80 and up is workable, 50–79 needs cleanup before you rely on it, and below 50 means fix the data before any analysis — or any AI project — touches it.
Questions
Straight answersDoes my file get uploaded anywhere?
No. The file is read and parsed entirely in your browser with JavaScript — open your network tab and watch: nothing leaves. The only tool on this site that touches a server is the Dashboard Generator, and it says so up front.
What formats does it accept?
CSV (or a .txt file containing CSV) and JSON — either an array of objects or a single object. Files up to 5 MB. Working from Excel? Save as CSV first, or round-trip it through our CSV ↔ JSON converter.
What counts as an outlier?
Any numeric value more than three standard deviations from its column's mean. That's flagged as a warning, not an error — extreme values are sometimes real. The report lists the row numbers so you can judge for yourself.
My score is low. Now what?
Start with the errors — they cost 15 points each and usually point at a handful of broken columns. The deduplicator handles duplicate rows in one pass. And if the file is big, messy, or business-critical, that's literally what we do for a living — the audit is free.
Beyond the score
The validator finds the problems. We fix them.
One-time cleans from $299, with a before/after quality report scored by this same engine.