What is normalization, and when might denormalization be justified?

Normalization organizes relational data to reduce unnecessary duplication and update anomalies. For example, storing a customer’s address once avoids inconsistent edits across many order rows when the business meaning is a current customer address. Historical order delivery addresses may legitimately be stored separately because they represent a different fact. Denormalization can improve measured read workloads, but requires a plan for consistency and updates. Start from correct data meaning and actual access patterns rather than treating more tables or fewer joins as an automatic sign of better design.