CSV generator

How to generate a CSV file

1. Set the number of rows (1 to 1,000,000) and a seed. The starting setup has six columns: a sequential id, a full name, an email, a signup date in 2024–2025, a plan chosen as Free, Pro or Team in a 6:3:1 ratio, and an amount from 0 to 500 with two decimals.
2. Add, remove or rename columns and pick a field type for each. Every column name must be filled in and unique.
3. Optionally set a share of blank cells or require unique values per column, then generate and download the file.

Rows are written in a background worker in parts rather than as one huge string, so a million-row file does not freeze the page, and cancelling stops the job. One output file is capped at 4 GB.

Field types

Seeds and reproducibility

The seed is any text. Each column draws from its own random stream derived from the seed and the column name, so:

• the same seed, row count and column settings give a byte-for-byte identical file every time, in every browser;
• adding, removing or reordering other columns does not change a column’s values, but renaming a column changes them;
• asking for more rows keeps the earlier rows unchanged and appends new ones;
• the interface language changes names, cities, companies and the default phone format, not numbers, dates, choices or emails.

Blank cells use a separate stream. A blank cell does not consume a value, so changing the blank percentage shifts which values appear in the later rows of that column.

Blank cells and unique values

Blank cells is a percentage from 0 to 100 per column; each cell is left empty with that probability. Use it to test how an import handles missing values.

Unique values redraws a value until it has not been used in that column, and gives up after 1,000 tries with an error naming the column. The settings are rejected up front when uniqueness is impossible: an integer range with fewer whole numbers than rows, or fewer choices than rows. Other types can still run out partway, for example a unique full name after about 144 rows in English (12 first names × 12 last names). Sequence, UUID and constant columns skip the check, and blank cells are not counted as values.

Settings errors

• Row count must be a whole number from 1 to 1,000,000, and there must be 1 to 100 columns.
• A column name is empty or used twice: every header must be filled in and different.
• Integer and decimal columns need both a minimum and a maximum, with the minimum not above the maximum.
• Date ranges must be written YYYY-MM-DD, such as 2024-01-01; 2024/01/01 or 01/02/2024 is rejected.
• A pattern must contain at least one #, ? or *; otherwise use a constant column.
• A choice list needs at least one value with a weight above zero.

Example

Five rows with the seed demo-2026: a sequence starting at 1001 padded to 6 digits, a full name, an email, a weighted choice, a decimal from 0 to 250 with 40% blank cells, and the pattern SKU-####-??. Run it again with the same settings and you get exactly these rows. Line endings are shown as plain line breaks.

rows: 5    seed: demo-2026    language: English
customer_id  sequence  start 1001, width 6
name         full name
email        email
plan         choice    Free:6 / Pro:3 / Team:1
mrr          decimal   0–250, 2 places, 40% blank
sku          pattern   SKU-####-??

customer_id,name,email,plan,mrr,sku
001001,Mary Brown,casey.maple1@example.test,Free,,SKU-6836-LU
001002,Ethan Davis,taylor.copper2@example.test,Pro,,SKU-8060-HR
001003,Ethan Clark,morgan.falcon3@example.test,Team,114.71,SKU-1401-TZ
001004,Olivia Davis,jordan.harbor4@example.test,Pro,91.82,SKU-8840-PT
001005,Mia Taylor,casey.prism5@example.test,Free,6.99,SKU-6878-OZ