Skip to content
Data Science & Analytics

Data Cleaning

The 80% That's Janitorial

7 min+65 XPStep 1/10

The part nobody photographs

Ask working data scientists where the time goes and you get a sheepish number: 60–80% is cleaning. Not modeling, not insight — reconciling formats, chasing missing values, arguing about what 'revenue' means in this table.

That isn't a flaw in the profession — it is the profession. The glamorous output (the model, the dashboard) is a thin layer over hours of unglamorous verification:

  • Does this column mean what I assume it means?
  • Did the definition change when the vendor system migrated in March?
  • Are 'sales' counted pre-refund or post-refund?

Every broken dashboard you've ever seen was a cleaning failure that shipped.

analogy

Data cleaning is mise en place — the chopping and measuring before the cooking. Nobody photographs it, but it's the difference between a kitchen and a fire.