Normalisation: Why Should One Fact Not Be Stored Everywhere?

Normalisation organises data around facts and their dependencies, giving each fact one authoritative home while other places refer to it by an identifier. It is not the same as making text formats uniform, which is standardisation, nor does it mean splitting tables as far as possible.

Suppose an accounting system copies a customer's address into both the customer record and every unsent invoice. If the customer moves and only one copy is changed, the system now contains two competing “current” addresses. Keeping the address in the customer record and storing the customer identifier on each invoice leaves one place to update and joins the information when it is needed.

Normalisation reduces update conflicts caused by repeated facts, but repetition is not always wrong. Deliberately retaining a copy to speed up a fixed report is denormalisation; it also requires a stated way to synchronise copies and an acceptable period of staleness. The central judgement is simple: first give facts clear ownership, then accept controlled duplication only for a measured reading need. A clear structure matters more than the slogan “do not repeat yourself”.