dataviz
Colour theme

CSV column profiler

Know what is actually in your CSV before you trust it

Upload a spreadsheet export and get a per-column profile in one pass: what type each column really holds, how much of it is missing, how many distinct values there are, and how the numbers are distributed. No formulas, no pivot tables, no sampling.

Free to use. CSV files up to 100 MB.

How it works

  1. Create an account

    Your name, an email address and a password of at least 8 characters. Nothing else to set up — you land straight on the upload screen.

  2. Upload a CSV

    Drop the file or pick it from disk. Your browser uploads it straight to storage — the file never passes through our server. If it is not UTF-8 we detect the encoding and convert a copy.

  3. Read the profile

    Every column is analysed in full — not a sample. A small file is done in seconds; a large one can take a few minutes, and the page tells you while it works.

What you get, for every column

The same set for all of them, plus extras where the type allows it.

Always

  • Its storage type as the engine read it — BIGINT, VARCHAR, TIMESTAMP.
  • An inferred semantic type: integer, decimal, boolean, date, date & time, categorical, text, or identifier.
  • Null count and percentage, so you can see what is missing before you build on it.
  • An exact distinct count — counted, not estimated.
  • The minimum and maximum value.

Where the type allows

  • Numeric columns also get mean, median and the 95th percentile, plus a histogram of the distribution.
  • Categorical and boolean columns get their most frequent values with counts.
  • Date and date-time columns show the span from earliest to latest.
  • Text and identifier columns have no meaningful distribution, so they show none rather than an empty chart.

What it looks like Example — not your data

Three columns from an imaginary 28,450-row orders export, drawn with the same components the real profile screen uses.

1order_idIdentifiernear-uniqueno nulls
2amountDecimalExample histogram: amount is right-skewed, with most orders between 100 and 175 and a long tail out to 500.0.4% null
3countryCategoricalExample top values: JP is the most common country with 8,913 rows, then US, ID, AU and DE.2.1% null

On the real screen each row opens to show the exact counts, the full histogram, the top values with their shares, and mean, median and the 95th percentile for numeric columns. Everything is also available as a plain table.

Files that do not start with a header row

Plenty of real exports open with a title line before the column names. That wrecks a naive profile: the title becomes a column name and the real header gets counted as data.

We look at the first rows and tell you which one we think holds the column names — and how confident we are. It is a suggestion, never applied automatically. You see the raw first rows yourself, pick the header row, and re-profile. You can also declare that the file has no header row at all.

Limits, and one thing to know before you start

  • CSV only. Excel and Parquet files are not supported.
  • 100 MB per file. Up to 500 MB and 20 datasets per account in total.
  • Profiling a large file can take several minutes. A run is given up on after five minutes.
  • Datasets cannot be deleted yet. There is no delete button, so every upload is permanent and counts against those two limits for good. Upload accordingly.
  • No dashboards, saved views, sharing or export yet. What you get is the column profile on screen.

Ready to look at your data?

Create an account and upload your first CSV — it takes about a minute.