Why you'd use this
Say you run a report of students by country. The pivot table shows UK 412, United Kingdom 380, U.K. 95, Untied Kingdom 12. That's one country counted four ways, and every chart built on it is wrong. The same thing happens with city names, product categories, job titles and department names, anywhere people type a value instead of picking it from a list.
Fixing it by hand means find-and-replace, one spelling at a time, and hoping you caught them all. This tool shows you every distinct spelling at once, maps the obvious ones for you, and asks you only about the ones it isn't sure of. Then the whole column comes back clean.
How to use it
- Paste the column you want to clean on the left, or upload the CSV and pick the column.
- Paste your canonical list on the right: the only values you want to end up with.
- Add alias rules for the mappings no algorithm will guess, like UK = United Kingdom.
- Hit Standardize. Every distinct value is mapped, with the number of rows it covers.
- Open Needs a decision, pick the right value from the dropdown, and download.
The order it tries things in
Each distinct value goes through three checks, and the first one that hits wins:
- Exact. The value is already in the canonical list once case, spaces and punctuation are ignored. "INDIA" and "india" both land on India.
- Alias. One of your alias rules covers it. That's how "Bharat" becomes India and "U.K." becomes United Kingdom.
- Fuzzy. The same smart matching as the fuzzy matcher. Typos like "Untied Kingdom" and "Nethrelands" score high and are mapped automatically if they clear the threshold.
Anything left over is marked needs a decision, with the three best suggestions in its dropdown. "Republic of India" will usually show up here: India is the top suggestion, but it's your call, not the tool's.
Why work on distinct values
A column with 200,000 rows often has only a few hundred distinct spellings. Decide each spelling once and every row that uses it follows. It's also how controlled vocabularies are run in practice: a short mapping table that a person owns and reviews, applied to everything downstream. Download the mapping table and you can reuse it as the alias list next time.
Where this comes from. I've spent a lot of time on controlled vocabularies, getting product, operations and counsellor teams to use the same names for the same things. At scale, an AI drafts the mapping and a person approves it. The Subject Validator case study shows that version.
Questions
What's the difference between this and the fuzzy matcher?
The fuzzy matcher compares two lists row by row. This one works on distinct values: if "UK" appears 4,000 times, you decide it once. It also adds alias rules and a dropdown to override any mapping, because standardizing a column is about getting every value right, not just finding a likely match.
What goes in the alias box?
Mappings that no string similarity would ever find, like UK to United Kingdom or Holland to Netherlands. One per line, written as alias = canonical value. The canonical value must be in your canonical list. Aliases are matched ignoring case, spaces and punctuation, so one UK line also covers U.K. and u k.
Will it change values I didn't approve?
Only fuzzy matches at or above the threshold are mapped automatically, and you can see and change every one of them. Anything below goes to "needs a decision" and stays as it was unless you pick a value.
Is my data uploaded?
No. The page does all the work in your browser. Nothing is sent to a server.