Normalization

Normalization is the practice of organizing tables so that each piece of information is stored in exactly one place, with related data split into separate tables connected by keys rather than repeated wherever it is needed. A database is said to be in a given "normal form" — first, second, third, and beyond — each stricter than the last about eliminating a specific kind of redundancy.

The classic example: instead of repeating a customer's full address on every one of their orders, a normalized design stores the address once in a customer table and references it by key from the orders table. This prevents update anomalies — updating the address once, in one place, rather than hunting down and fixing every repeated copy — and enforces data consistency automatically, since there is only one version of the fact to be right or wrong.

Normalization is standard practice for transactional, write-heavy systems, where consistency and storage efficiency matter more than read speed. Analytical modeling frequently goes the other direction, applying denormalization deliberately, because heavily normalized schemas like a snowflake schema require many joins at query time, which slows down the read-heavy, ad hoc queries typical of reporting — a warehouse's star schema design is a conscious trade of some redundancy for query simplicity and speed.