Database normalization is the practice of organizing a database into related tables so each fact is stored only once, which reduces errors and duplicated data.
Data · Reviewed by Ray Rasouli
Ask about database normalization
Plain answers, with your setup in mind.
Ask anything about database normalization. For example, how it works in your current tools.
A scripted guide for now. It does not store what you type.
In a poorly designed database, the same fact gets typed into many places. A customer's address might be stored on every order. When the address changes, some copies get updated and others do not.
Normalization splits data into tables by subject: customers, orders, products. Each order just points to the customer by ID. Change the address once and every order sees the new one.
A small manufacturer running orders from one big spreadsheet faces this every day. Moving to a normalized database, or a system built on one, ends the hunt for which row is correct. Reporting databases sometimes undo some normalization on purpose for speed, but the main system should stay normalized.
In practice
Spreadsheets are used as databases.
One wide sheet repeats the same details on every row, and edits drift apart.
It is taken too far.
Splitting data into too many small tables can make simple reports slow and hard to build.
Copied fields are added for convenience.
Developers add a duplicate field to save a lookup, and it falls out of sync over time.
They are a series of rules, such as first, second and third normal form, that describe progressively stricter ways to remove repeated data from tables.
It is the deliberate copying of some data back into fewer tables to speed up reading, often used in reporting databases and data warehouses.
A 30-minute discovery call with Ray, then a clear list of what to fix first. No obligation.
With Ray Rasouli. No obligation.