Normalization vs denormalization — what’s the trade-off?

Deep dive

**Normalization** organizes data to reduce redundancy and anomalies (classic normal forms like 1NF/2NF/3NF). Benefits:

• one source of truth,
• fewer inconsistencies,
• simpler updates,
• smaller rows.

Cost:

• more joins,
• more complex queries,
• sometimes slower reads for read-heavy workloads.

**Denormalization** intentionally duplicates or precomputes data to make common reads fast:

• fewer joins,
• simpler read queries,
• can match access patterns (feeds, dashboards, analytics).

Cost:

• harder writes (must update multiple places),
• risk of inconsistency,
• more storage.

Practical guidance

• Start normalized for OLTP systems.
• Denormalize selectively based on measured bottlenecks (use `EXPLAIN` and real traffic).
• Consider alternatives: materialized views, caching layers, read replicas.