CQRS -- separate read and write models

A single data model optimized for writes (normalized) is inefficient for reads (requires complex joins). Or vice versa: a denormalized read model is difficult to update consistently.

Command Query Responsibility Segregation: use separate models for reads and writes. Write model: normalized, validates business rules, processes commands. Read model: denormalized, optimized for queries, may use different storage (e.g., Elasticsearch). Sync models via events or change data capture. Start simple: same database, different query and command handlers. Scale: separate databases when needed.

Read and write workloads have fundamentally different requirements. Reads need speed and flexible queries. Writes need consistency and validation. One model cannot optimize for both.