Normalisation

The forms matter less than the property they produce. 1NF — atomic values, no repeating groups. 2NF — no partial dependency on part of a composite key. 3NF — no transitive dependency; every non-key column depends on the key, the whole key, and nothing but the key.

What you are actually buying is that a fact lives in one place, so it cannot disagree with itself. Storing a customer's address on every order means an address change must update every order, and any missed path produces silent divergence that no constraint will catch.

The reason to know the forms explicitly rather than by instinct: they name the specific anomalies. An update anomaly is changing one copy and not another. An insertion anomaly is being unable to record a fact because unrelated data is missing. A deletion anomaly is losing a fact as a side effect of deleting something else.

Normalise first, denormalise deliberately and locally where a measured read path requires it — and when you do, write down which writer is responsible for keeping the copies in step, because that obligation is now permanent.