The relational model is built on the principles of data normalization. This improves data quality by avoiding duplication and preventing data errors.

This video shows the problems that can occur when you store unnormalized data. It then walks through normal forms 1, 2, and 3. For each Chris discusses their criteria, how to fix your model to meet each form, and how this avoids further problems.

Highlights from this session include:

1:10 - Errors that can occur in unnormalized data
3:00 - Overview of normal forms 1-3
4:30 - First normal form
6:10 - The problems with storing comma-separated lists
11:00 - Converting comma-separated lists to columns
14:30 - Avoiding duplicate rows with primary keys/unique constraints
16:50 - Surrogate keys
17:30 - Candidate keys
19:30 - Converting comma-separated lists - rows or columns?
32:20 - Functional dependencies
34:10 - Second normal form
36:20 - Foreign keys
43:45 - Third normal form
47:50 - Derived values - storing vs calculating at runtime
51:30 - Using virtual columns to store formulas
52:50 - Check constraints to keep stored derived values in sync
55:15 - Quiz: is this design in 3NF?
59:50 - Recap
1:01:50 - Further reading