Learn to clean messy data by actually cleaning it

Most data work starts with a file that is not quite right: padded identifiers, ambiguous dates, duplicate lines, amounts in mixed formats. Data Cleaning Exercises gives you realistic, fictional datasets with those problems and asks you to fix them under clearly stated business rules.

What makes it different

• Real execution. SQL runs on DuckDB and Python runs on pandas. Nothing is simulated.
• Rules first. Every exercise publishes the rules that decide correctness, so you are never penalized for guessing differently.
• Graded on new data. Your work is re-run on hidden variations, so a hardcoded answer or deleting hard rows does not pass.
• Evidence kept. Invalid rows are quarantined with a reason, not silently dropped.

What it is not

This is a learning product, not a data-cleaning service for confidential data, and completion certificates are not professional accreditation. Product and tool names are used descriptively; there is no affiliation with their vendors.