CSV Data Profiler

Before analysing, importing or sharing a dataset, it pays to know what is really in it. How many cells are empty? Are the numbers really numbers? Is there an amount ten times larger than all others? Which values occur most often? A data profile answers these questions for every column at once, and this tool builds one from any CSV or Excel file.

The overview

At the top, the profile counts rows, columns, empty cells with their share of the whole table, and how many columns are numeric. A high share of empty cells or fewer numeric columns than expected is the first sign that something needs cleaning.

Column types

Each column’s type is detected from its filled cells: whole numbers, decimals, dates, true/false, or text. A type is assigned when at least 90 percent of the filled values agree, so a single typo does not turn a numeric column into text, while a column that is half text is labelled text. Numbers written like 1,234.50, €9.99 or (12) are recognized; switch on the decimal-comma option for files that write 1.234,50.

Missing and distinct values

For every column you see how many cells are empty, as a count and a percentage, and how many distinct values it has. An ID column should have as many distinct values as rows; a status column only a handful. The five most common values with their counts reveal spelling variants, placeholders like “n/a”, and default values that were never changed.

Numeric statistics

Numeric columns get minimum, maximum, mean, median, the first and third quartiles and the sample standard deviation. Comparing mean and median is a quick check for skew: a mean far above the median means a few large values pull it up.

Box plots, histograms and outliers

A box plot shows the middle half of the values as a box, the median as a line and whiskers up to the most extreme values within 1.5 times the interquartile range. Values beyond that are counted as outliers, a common rule for values worth a second look. The histogram shows how values spread across twelve equal bins, revealing gaps, clusters and long tails.

Codes with leading zeros

ZIP codes, account numbers and product codes often start with zeros. Such values are counted and treated as text codes, not numbers, because averaging ZIP codes makes no sense and converting them would drop the zeros.

Downloading the profile

The full profile downloads as a CSV or Excel table with one row per column: type, counts, the top value, statistics and length ranges. Keep it as documentation of the dataset, or compare profiles of two exports to see what changed.

Privacy

Datasets are profiled on your device; nothing is uploaded.

A quick quality checklist

After profiling, ask a few questions. Is every required column free of empty cells? Does the ID column have as many distinct values as there are rows? Do numeric columns have plausible minimums and maximums, without negative quantities or prices of zero? Are dates within the expected period? Do category columns contain only the expected values, without spelling variants? Each “no” points to a clean-up step before the data is used.

Related tools

Summarize groups with the CSV Pivot Table, find repeated rows with Dedupe CSV, and mask personal columns before sharing with Anonymize CSV Data.