Count Unique Values in a Column
The first thing to do with an unfamiliar column, and the fastest way to find the problems in a familiar one. Sorted by frequency, a value-counts table shows you the real categories, the typos hiding among them (Germany 4,102 and germany 3), and whether a column you assumed was unique actually is. It's also the check to run before any join or dedupe — a count above 1 on your key column is what will multiply your rows.
How to use
- Paste or drop the file and name the column by header or 1-indexed position.
- Read the status line. It gives you the distinct count, the total rows, and whether any value repeats — often the whole answer.
- Set Top N to keep the output readable on a high-cardinality column.
- Toggle case and space normalisation to see how much of your apparent variety is really formatting noise.
Three questions it answers immediately
"Is this column unique?" If the status line says all values are unique and the distinct count equals the row count, it's a candidate key. Anything less and a join on it will multiply rows.
"What are the real categories?" Sorted by frequency, the genuine values are at the top and the data-entry accidents are at the bottom — the long tail of counts of 1 or 2 in a column that should have five values is your cleanup list.
"How much of my data is missing?" Empty cells are counted and shown as (empty) rather than being skipped, with their share in the percent column. That's a more honest null-rate figure than a summary statistic, because you see it next to the real values.
Reading the normalisation toggles
Turning on ignore case and ignore surrounding spaces collapses variants into one row. Compare the distinct count with the toggles off and on: if it drops from 40 to 12, you have 28 rows' worth of formatting inconsistency and the fix is trim columns plus a case normalisation. The displayed label is taken from the first occurrence, so you still see the spelling as it appears in the data rather than a lower-cased version of it.
FAQ
Can I count combinations of two columns?
Use group and aggregate with both columns in the By field and count rows as the aggregate — that's the same operation over a composite key. Or join the columns first with concatenate columns and count the result.
How do I get the rows behind one value?
Filter rows with column = "value". For a per-value file rather than one at a time, split by column value writes one file per distinct value.
What's the shell equivalent?
cut -f3 f.tsv | tail -n +2 | sort | uniq -c | sort -rn. Add | head -20 for a top-N. This page adds percentages and normalisation, and doesn't make you count the field number.
Does it work on numeric columns?
Yes, but a frequency table over continuous values isn't very useful — every value appears once. Stats gives you min, max, mean, and null counts, which is the right summary for numbers. Value counts is for categories and identifiers.
Privacy
100% client-side. No upload. See the privacy policy.