FREE WEB TOOL
Randomly sample rows without replacement from CSV, with an optional deterministic seed.
Pull 50 rows for a quick review or select 100 records for a spot check without duplicates. Leave the seed blank for a fresh random sample; provide a seed when you need the same sample again.
The data rows are shuffled with a Fisher–Yates process, then the first N rows are returned. A source row cannot appear twice in one sample. If N is larger than the available row count, the tool simply returns all rows.
Use the same nonblank seed with the same CSV and conditions, and you can reproduce the same result. That is useful for audits or reviews where someone else needs the exact same set of records.
Leave the seed blank and the tool starts from fresh randomness, so blank does not mean a hidden fixed sample.
If 90% of your rows are group A and 10% are group B, a 10-row sample may contain one B row—but it can also contain none. This is simple random sampling, not stratified sampling.
The first header row is always kept at the top. The data-row order is randomized. If original row order matters for later tracing, add an ID or row-number field before sampling.
For a 100,000-row file, start with a fresh 100-row sample to look for obvious problems. If you find something worth sharing and need others to inspect the same records, rerun with a fixed seed. That gives you both exploration and reproducibility.