Data Cleaning Exercises

Guide · updated 2026-11-04

Outliers: flag, don't delete

How to handle unusual values without changing the population you report on.

Extreme is not wrong

A very large sale or a refund can be real. Deleting it changes totals and hides events that matter.

Flag with a reason

Add a flag column with a declared threshold or rule and let someone with context review. Report how many rows were flagged.

Sentinels are different

Codes like 9999 or -999 that mean no reading are missing values in disguise. Convert them explicitly with a documented rule.

Practise it