Follow five short steps: name your columns, choose the kind of data, set a few rules, check everything — then download a CSV with up to 10 million rows.
Plain answers about how the rules work.
No value appears twice in that column. Number and date ranges are shuffled so each value is used once. Name-like columns get a number added (e.g. “Manila-42”). If the range is too small for your row count, you'll see a message before you continue.
The column is never left blank. Without it, the Blank cells slider (under More) controls how often a cell is empty.
The row's ID — like a primary key in a database. It's unique, required and always clean. Only one column can be the Key. A Row number or Unique code works best.
Choose My own list as the data type, then type the values separated by commas — for example Bronze, Silver, Gold, Platinum. Only those values are used, picked at random. Typing a value twice makes it appear more often.
Realistic mistakes for practising data cleaning: extra spaces, random capitals, typos, wrong formats (john@@mail.con), placeholders like N/A, unusual values and broken characters. Unique columns only get extra spaces so they stay unique.
Never checks every row and replaces any repeat. Add on purpose copies earlier rows but keeps Unique and Key columns different — like the same customer entered twice.
No — Excel stops at 1,048,576 rows. For larger files, use a database, Power BI / Power Query, or Python. On Chrome or Edge, turn on Save straight to disk in File format for very large files.
No. Everything is created inside your browser. Nothing is uploaded, and no cookies or browser storage are used. Close the tab and it's gone. Only the feedback form sends what you type into it.
What's being built next.
Generate several related tables at once — customers, orders, payments — linked by matching keys, ready for JOIN practice in SQL, Excel or Power BI.
Upload a sample of your own spreadsheet and the tool detects every column, type and pattern — then creates a safe, look-alike dataset with zero real customer data.
Get the same data as an Excel workbook, JSON, or ready-to-run SQL INSERT statements — one click, straight into the tool your team already uses.
Turn your dataset into a mock REST API with JSON responses, paging and filters — so front-end and QA teams can build before the real backend exists.
Lists come from text files in the data folder (one item per line). A new .txt file appears as a new data type.
| File | Used for | Items | Source |
|---|
Give each column a name and pick the kind of data it should hold.
How many rows, how clean the data should be, and whether rows may repeat.
Anywhere from 1 to 10,000,000.
Clean data, or a little mess to practise cleaning? Fine-tune any column later with More in Step 1.
Whether the same whole row may appear more than once.
How the CSV file should be written. The defaults work for most tools.
Everything in one place. Use Edit to change anything before you preview.
The first rows of your actual file.
You can keep using the page while this runs.
A matching CREATE TABLE statement, plus the command to load your CSV.