Describe 1NF, 2NF, 3NF, normalization's purpose, and its performance trade-off.

WHAT THIS TESTS: The interviewer wants to see if you understand normalization as a mechanical process with specific goals, not just academic definitions. They are checking whether you can connect each normal form to a concrete data-integrity problem and whether you understand that normalization is a design trade-off rather than a universal rule. Senior candidates should show they can reason about when to break normalization for performance.

A GOOD ANSWER COVERS: First, 1NF: every column contains atomic indivisible values and there are no repeating groups. Second, 2NF: the table is in 1NF and every non-prime attribute depends on the entire primary key, eliminating partial dependencies. Third, 3NF: the table is in 2NF and there are no transitive dependencies where non-key attributes depend on other non-key attributes. Next, state the problem normalization solves: it reduces data redundancy and prevents insertion, update, and deletion anomalies so that a single fact is stored in one place. Finally, name the performance trade-off: a highly normalized schema requires more joins to reconstruct entities, which increases query complexity and read latency, often leading to intentional denormalization in read-heavy systems.

COMMON WRONG ANSWERS: A red flag is listing the normal forms as one-word concepts like atoms, keys, and dependencies without explaining what anomaly each form prevents. Another mistake is saying normalization makes the database faster; it actually slows down reads by spreading data across more tables. Some candidates also claim every database should be in 3NF or BCNF by default, which signals a lack of practical experience with large-scale read-heavy workloads where selective denormalization is standard.

LIKELY FOLLOW-UPS: The interviewer may ask how you would model a read-heavy analytics workload, which should lead to a discussion of denormalization, star schemas, or materialized views. They might also ask for the difference between 3NF and Boyce-Codd normal form, or when you would intentionally violate normal forms, such as storing a computed column or duplicating a customer name in an orders table to avoid a join.

ONE CONCRETE EXAMPLE: Imagine a table with columns order_id, product_name, product_price, and customer_zip. If product_price depends only on product_name, not on the order_id key, that is a transitive dependency violating 3NF. Splitting this into an orders table, a products table, and a customers table removes the redundancy, but fetching an order summary now requires three joins instead of one. In a high-traffic e-commerce site, you might keep product_price denormalized in the orders table to serve the query in a single row lookup.