Data platform · Storage & Modeling

Normalized vs denormalized

EasynormalizationdenormalizationOLTPanalytics

Interview question

When do analytics systems denormalize data?

How to answer

The short answer

systems often normalize (3NF) to avoid update anomalies: customers in one table, orders in another.

Analytics systems often denormalize selected attributes into wide facts/marts so BI queries avoid heavy multi-join graphs.

Output
OLTP:  customers >< orders >< order_items >< products
BI mart: fct_order_items with product_name, category already joined

Denormalization trades storage and some update complexity for read speed and simpler dashboards.

Caution

If a product name changes and you denormalized it into historical facts, decide whether facts keep the name at order time (usually yes) or always show the latest name.

Interview tip: "Normalize for writes; denormalize carefully for reads." Tie it to and mart layers.