CSV Data Profiler
Understand a dataset before you use it. The profiler reads CSV or JSON and, for every column, infers the type, measures how complete it is, counts distinct values, shows ranges and means for numbers and dates, lists the most common values and flags data quality problems: missing values, mixed types, leading or trailing spaces, outliers and duplicate rows. Numeric columns get a histogram.
- Runs in your browser
- No sign-up
- Free to use
Processed entirely in your browser – your data is not uploaded. Files up to 50 MB.
How to use CSV Data Profiler
- Paste data or open a file.
- Click Profile data.
- Read the column table and findings.
- Fix issues with the cleaning tools.
CSV Data Profiler features
Type inference
Integer, decimal, date, email, URL, boolean, text.
Completeness
Filled share per column.
Quality findings
Mixed types, spaces, outliers, duplicates.
Histograms
For numeric columns.
CSV and JSON
Same report for both.
Private
Runs in your browser; nothing is uploaded.
When to use CSV Data Profiler
- Checking exports before analysis.
- Data migration projects.
- Reviewing third-party data.
- Teaching data quality.
CSV Data Profiler FAQ
How are outliers found?
Values outside 1.5 × the interquartile range (Tukey’s rule).
What counts as missing?
Empty cells and placeholders like NULL, N/A or “-”.
How big can the file be?
Up to 50 MB, depending on your device.
What is a high-cardinality column?
A column with many distinct values, such as IDs, emails or timestamps. They are poor choices for grouping in charts.
How is the column type inferred?
From the values themselves: if nearly all non-empty values are numbers, dates or true/false, the column gets that type. Mixed columns are flagged.
What does completeness mean?
The share of rows with a non-empty value in that column. A completeness of 60% means 40% of rows have no value there.
Can I profile JSON data?
Yes. Arrays of objects are read like a table, with one column per key.
Profile first, analyse second
Most analysis errors come from data that is not what you assumed. A two-minute profile shows gaps, wrong types and duplicates before they reach a report.
Profiling a dataset before using it catches problems early: a date column stored as text, a numeric column with “N/A” values, an ID column with duplicates, or a category column with spelling variants such as “UK”, “U.K.” and “United Kingdom”.
For every column the profiler infers the type, measures completeness, counts distinct values and lists the most common ones. A column where almost every value is distinct is likely an identifier; one with a handful of values is a category you can validate.
Your data is processed in the browser and never uploaded. Use the findings to decide what to clean before importing the data into a database, BI tool or model.
Profile every new data source before building reports on it, and again when the source changes. A profile from last month compared with today’s shows quickly whether a column suddenly has more missing values, new categories or a different type – often the first sign of a problem upstream.
For very large files, profile a representative sample first – for example the first 100,000 rows – to get quick answers, then profile the complete file once the cleaning steps are clear. Remember that a sample may miss rare values that only appear in later rows.