Loading
Preparing the page
Public context is loading before any interactive tool code.
Preparing the page
Public context is loading before any interactive tool code.
Parse a bounded CSV locally and inspect row width, missing cells, exact distinct values, duplicate rows, conservative lexical kinds and numeric ranges.
Your checked result will appear here.
Rows pass through quote, separator, header and width checks before completeness, distinctness and numeric summaries are calculated. Broken structure never becomes a polished report.
The supported subset follows the common comma, record and doubled-quote conventions. Structural errors surface with a row or character location.
Commas separate fields and CRLF, LF or CR separates records unless the character is inside a quoted field.
Two consecutive quote marks inside a quoted field become one literal quote in the observed value.
Headers must be visible, bounded and unique without relying on letter case.
Every data record must have exactly the same number of fields as the header before profiling begins.
The synthetic fixture includes a quoted comma in the region field, three numeric amount cells and one empty amount. The profile does not substitute or estimate the missing value.
order_id,region,total_cents,refunded ord-1001,North,2400,false ord-1002,"South, East",4100,false ord-1003,North,1300,true ord-1004,West,,false
| Column | Kind | Present | Missing | Distinct | |---|---:|---:|---:|---:| | order_id | text | 4/4 | 0 | 4 | | region | text | 4/4 | 0 | 3 | | total_cents | integer | 3/4 | 1 | 3 | | refunded | boolean | 4/4 | 0 | 2 | total_cents: min 1300 · max 4100 · mean 2600
Observed: 4 rows · 4 columns · 1 missing cell · 1 numeric · 3 complete · 0 duplicate rows
ord-1005,North,900The record has three fields while the header has four, so analysis stops.
The output favors small, inspectable facts over an automatic quality score. A column can be complete and still be incorrect, biased or irrelevant.
Whitespace-only cells count as missing; non-empty cells count as present without a truth-value judgment.
Exact string values are counted after surrounding whitespace is removed; letter case remains significant.
A column is numeric or boolean only when every present cell matches that lexical form; all other mixtures remain text.
Minimum, maximum and arithmetic mean use present numeric cells only and never fill a missing value.
The Markdown report is a starting inventory. Decisions about validity need the source contract, collection process and intended use.
Ask the source owner for units, identifier rules, null policy, permitted values and collection period.
Check the values behind surprising counts before calling them errors or outliers.
A bounded sample may not represent the complete export, later periods or other populations.
Turn each approved quality requirement into a test instead of relying on an unexplained composite score.
The tool reports visible structure and descriptive values. It does not diagnose causes or decide fitness for use.
001 remain text so identifiers are not silently converted to numbers.