ARZENTIQ
DEVELOPER & DATA

CSV Column Profile & Stats

Profile every CSV column: inferred type, empty and distinct counts, min, max, mean, median, standard deviation, text lengths, date range and top values.

A profile table (one row per column) as CSV, plus totals: rows, empty cells, duplicate rows, ragged rows and columns whose type has outliers.

Example: id,name,score,joined,ok over 4 rows → id integer 1–4, mean 2.5; name text, 3 distinct, top "Ann" ×2; score decimal, 1 empty, mean 8.25, stdev 1.25; joined date 2023-12-31 … 2024-03-01.

v0.1.0 · last reviewed 21 September 2026
Loading the workspace…
BUILT TO BE UNDERSTOOD

Every column,
described.

How a column’s type is decided, which statistics are shown, and what counts as empty.

Type inference

Each non-empty cell is classified from its text: an integer or decimal (plain digits, an optional sign, thousands commas only when a decimal point is present), an ISO 8601 date, a boolean word (true/false/yes/no/y/n/t/f) or text. A column takes the one kind it contains; a mix of integers and decimals is decimal; a ≥ 90 % majority keeps the majority type and counts the rest as outliers; anything else is mixed.

Statistics

Numbers get min, max, sum, mean, median and the sample standard deviation (n − 1); text gets the shortest and longest length; dates the earliest and latest; every column gets the empty count, the distinct count and its most frequent values. Statistics are computed over the majority-type cells only, so one “n/a” does not turn a numeric column into text.

Empty, ragged, duplicate

A cell is empty when it is blank after trimming; a row shorter than the header counts as empty in the missing columns and is reported as ragged; duplicate rows are rows identical in every column. Numbers with currency symbols, percent signs or a decimal comma are text here — the Number & Date Normaliser can clean them first. Nothing leaves the browser; the same four anonymous usage counts as the rest of the site apply.

SOURCES

  • Type = the one kind seen, or a ≥ 90 % majority with outliers counted, else mixed; plain decimal numbers, ISO 8601 dates, true/false/yes/no booleans; sample standard deviation (n − 1)

Last reviewed 21 September 2026. How results are checked: How we verify.