Step 1: Imagine a single giant unnormalized table.
Suppose all information, customer details, order details, and product details, were crammed into one huge table. Every time a customer places a new order, their name and address would be repeated all over again, wasting space and creating many identical copies of the same fact.
Step 2: See what normalization does to fix this.
Normalization systematically splits this giant table into smaller, well structured related tables, so each fact is stored only once and linked using keys. This directly cuts down redundancy, and because each fact lives in exactly one place, updating it becomes simple and consistent, which is what we mean by improved data integrity, no more risk of one copy being updated while another copy is forgotten.
Step 3: Rule out the opposite sounding options.
Increasing redundancy, decreasing performance, and cramming everything into a single table are all things normalization actively works against, they describe the unnormalized starting point, not the goal of the process.
Step 4: Conclude.
The true purpose of normalization is
\[ \boxed{\text{To eliminate data redundancy and improve integrity}} \]