Replace a column with plausible people, and count what collided.
What it's for
Swap the real people in a spreadsheet for made-up but realistic ones, to test a system, train someone or take a screenshot without showing anybody's real details.
Before you start
Have the spreadsheet open, or the .csv file in a text editor, and know which columns hold personal details. You paste the rows in; nothing is uploaded.
Set up how the rows are read
In Delimiter, pick Tab if you will copy from Excel or Google Sheets, or Comma for most .csv files. Leave First row is a header ticked if the first row holds the column names.
Choose the kind of names
In Names, Filipino gives common Philippine names and International gives common English-language ones.
Keep repeats consistent
Leave Keep equal values equal ticked so a person who appears in many rows gets the same fake name each time, and counts still add up. Untick it for a one-off screenshot. If it stays ticked, type any word in Salt (optional).
Paste the rows and pick a fake for each column
Paste into the CSV box, replacing the sample. Under Columns to replace, set each private column to a kind of value, such as Full name or Mobile number, and leave the rest on Leave as it is. Number in the same range keeps amounts believable.
Generate and copy
Press Generate to bring the result into view — it also updates as you change things. Press Copy to take the new rows.
Reading the result
If it goes wrong
Masking proves the column was handled; it does not give you anything to show. xxxx xxxx gives a vendor nothing to demo with, a developer nothing to load into a staging database, and a screenshot nothing to be a screenshot of. Rosa Bautista gives all three, and the original is just as gone.
But the usual way a fake-data generator works is wrong here. Draw a fresh name for every cell and one customer appearing in twenty rows becomes twenty people — changing every count, join and group over that column. So the fake value here is derived from the real one: same input, same output.
The price is that it is reversible — anybody holding this page can run it forwards. The salt closes that and costs reproducibility.
And collisions are the failure nobody looks for. Sixty given names and sixty surnames is 3,600 combinations, which sounds ample and is not: the birthday bound puts an even chance of the first clash at about seventy rows. Two real people given one fake name are one person to everything downstream.
No. There is no upload and no file picker — you paste the rows in and the replacement happens in your browser, in this site’s own calculation engine. Nothing is transmitted, logged or stored, which matters more here than on any other page: what people paste into this tool is a staff list or a customer export.
Because otherwise the file stops meaning anything. One customer appearing in twenty rows of a sales export has to become one fake customer, or every count, join and group over that column changes. So the fake value is derived from the real one rather than drawn fresh each time, and the same input gives the same output. You can switch that off, and the page says what you lose when you do.
Every cell is drawn independently from the browser’s cryptographic random source. Nothing can be reversed, which is the strongest privacy this tool offers — and one customer in twenty rows becomes twenty different people, so no count over a replaced column is meaningful any more. It is the right setting for a screenshot and the wrong one for a test database.
With consistency on and no salt, partly — the mapping is a function of the value, so anybody with this page can put a name in and see what it becomes, then check whether that name is in your file. Put anything in the salt field and that stops working. The cost is that the fakes will be different next time unless you use the same salt again, which matters if two exports have to line up.
Because two real people given the same fake name are one person to everything downstream, and nothing about the file looks wrong. The pools here hold a few thousand name combinations, which sounds ample: the birthday bound puts an even chance of the first clash at about seventy rows. The page counts them per column so you can see it rather than discover it later. A column that has to stay unique should use Reference.
They are random, but inside the range the real column already occupied. A salary column replaced with values between 1 and 1000 will not load into anything that validates it, and a hire-date column full of dates in 1970 makes every tenure calculation nonsense. Staying inside the real minimum and maximum keeps the file usable and still contains nobody’s salary.
They cannot be. Every generated address uses example.com, example.org, example.net or a subdomain of them — domains reserved by RFC 2606 precisely for this, which nobody can register. That matters when a fake dataset gets loaded into a real system that then tries to email it.
They are the commonest given names and surnames of the place you pick, which is what makes them look plausible and also means a generated row may match a living person by coincidence. Nothing links that person to your data, but it is worth knowing before a fake dataset ends up in a screenshot.
The range is measured before anything is replaced. Measured as it goes, an early fake value would widen the range the later ones are drawn from and the column would drift away from itself row by row.
Important
Nothing you paste is transmitted or stored. The generated names are common ones, so a row may match a living person by coincidence, and everything you did not replace is untouched — a fake name beside a real job title, branch and start date can still identify somebody. Read the output before you send it. Nothing you type here is sent to our servers — the calculation runs entirely in your browser.
