Online Tool Store
CSV Statistics
Paste or load a CSV to profile every column — type, blanks, unique values, min, max, mean, median, and standard deviation. Runs entirely in your browser.
🔒 This tool runs entirely in your browser. Your files are never uploaded to a server.
Example shown — replace it with your own data. Nothing is uploaded.
How it works
- Paste your CSV, or load a .csv file from your device. Neither is uploaded.
- The delimiter is detected for you — override it if a file uses something unusual.
- Every column is profiled as you type: type, blanks, unique values, and the numeric spread.
- Download the profile as its own CSV to keep alongside the data.
The statistics, and which ones to trust
For a numeric column the mean, median, and spread are computed as:
mean = Σx ÷ n
median = middle value once sorted (mean of the two middles when n is even)
std dev = √( Σ(x − mean)² ÷ (n − 1) )
The n − 1 in the standard deviation is deliberate. A CSV is nearly always a sample of something bigger, and dividing by n would make the spread look smaller than it is; n − 1 corrects that bias. It is left blank for a single-value column, where spread has no meaning.
Reading mean against median is the fastest way to spot a skewed column. In the example data the units column has a mean of 120.5 but a median of 103 — the mean sits well above the middle value because one large row pulls it up. When those two numbers drift apart, the median usually describes a typical row better than the mean does.
The blank and unique counts answer different questions: blanks tell you where data is missing, while a unique count equal to the row count suggests an identifier, and a very low one suggests a category.
FAQ
Is my data uploaded anywhere?
No. The file is read and analysed entirely by your browser, which is the point — you can profile a customer export or a payroll extract without it leaving your machine. Loading a file uses the same local read; nothing is sent.
Does it handle commas inside quoted fields?
Yes. Quoted fields may contain the delimiter, line breaks, and doubled quotes for a literal quote character, all following the CSV standard. A row like "East, coastal",212 is read as two fields, not three.
How does it decide a column is numeric?
Every non-blank value in the column has to parse as a number. Thousands separators and a trailing percent sign are tolerated, so 1,234 and 45% count. If a single value is text — a stray "n/a", for instance — the column is treated as text and the numeric statistics are left blank rather than computed from a subset.
Which standard deviation is used?
The sample standard deviation, dividing by n − 1. Data in a CSV is almost always a sample rather than a whole population, and using n − 1 corrects the underestimate you get otherwise. It is left blank for a column with only one value, where it is undefined.
What counts as blank?
An empty cell or one containing only whitespace. Blanks are excluded from every statistic and counted separately, and the count is highlighted when it is above zero — an unexpected number of blanks in a column is usually the first sign of a problem upstream.
How large a file can it handle?
The limit is five million characters, roughly a 5MB file, because the whole thing is held in the tab while it is analysed. For anything larger, profile a representative extract or use a database or a script instead.
How we compare
| Feature | Online Tool Store | Upload-based CSV sites | Excel / pandas |
|---|---|---|---|
| Your data never leaves your device | ✓ | ✗ | ✓ |
| Whole-file profile in one glance | ✓ | Varies | A few lines of code, or manual work |
| Nothing to install, no sign-up | ✓ | Sign-up is common | ✗ |
| Handles quoted fields and odd delimiters | ✓ | ✓ | ✓ |
| Files above a few megabytes | ✗ | Often, since the server does the work | ✓ |
| Charts, pivots, grouping, joins | ✗ | Some offer basic charts | ✓ |
This is the first look you take at an unfamiliar export — how many rows, which columns are numeric, where the gaps are — on data you should not be uploading to a stranger's server in the first place. It profiles; it does not transform. Once you know what you have, Excel or pandas is where the actual analysis belongs.