Data tool

Dataset Corruptor

Damage CSV, TSV, JSON, or plain-text datasets on purpose so you can test importers, validators, ETL jobs, and cleanup workflows against messy data.

Runs Locally

This tool runs in your browser. Files never leave your device.

Drop a JSON, CSV, TSV, or TXT file here or paste data below.

Drag and drop a dataset file anywhere inside this input card.

Corrupted output

Detected CSV · 5 source lines · 242 B output

id,na,email,plan,active,score,created_at
1002,Jane Doe,jane@example.com,Free,,74,2026-06-03
1003,Sam Smith,sam@example.com,Team,true,88,2026-06-04
,Avery Stone,avery@example.com,Pro,true,91,
1001�,false,laramie@example.com,,true,98,2026-06-01

Corruption report

Rows dropped
0
Rows duplicated
0
Rows shuffled
4
Values blanked
4
Values mutated
1
Types changed
1
Fields removed
0
Fields renamed
1
Junk inserted
1
Structure damage
0

What this breaks

Imports: missing rows, duplicated rows, changed delimiters, blanks, and broken structure.
Schemas: removed fields, renamed keys, changed value types, null-ish values, and malformed output.
Cleaning: junk characters, weird casing, truncated strings, fake nulls, and inconsistent values.
Testing: seeded output lets you reproduce the same messy dataset during debugging.

What this tool does

Dataset Corruptor adds controlled noise, missing values, and bad rows to a dataset so you can test how your pipeline handles imperfect data. It is built for data and ML teams who need realistic dirty data to validate cleaning, preprocessing, and quality checks before production.

How it works

Paste or upload your data, choose what kinds of problems to add and how much, and the tool generates a corrupted version in your browser. Copy or download it for testing. Nothing is uploaded.

When to use Dataset Corruptor

Validating a data-cleaning pipeline
Testing how a model handles missing values
Building realistic dirty data for QA
Stress-testing data quality checks