6. Operational Efficiency and Optimization (12%)

Domain 4 is the domain where all prior architectural decisions become measurable outcomes. Model selection choices become cost-per-query numbers. Caching decisions become latency improvements. Chunking strategy choices become retrieval performance metrics. The exam tests whether you understand the levers — which architectural changes produce which operational improvements — and whether you can read a scenario's constraints and select the right optimization strategy.

Three sections cover this domain in sequence: cost optimization (where money is spent and how to reduce it), performance optimization (where time is spent and how to reduce it), and monitoring (how to know when either is degrading). They are interdependent — the most common exam trap in this domain is recommending a cost optimization that silently degrades performance, or a performance optimization that doubles cost without commensurate benefit.

⚠️ Common Misconception: Switching from on-demand to provisioned throughput is always the primary cost optimization lever. In practice, prompt optimization, caching, and model right-sizing typically deliver larger cost reductions than throughput pricing changes — and they don't require capacity commitment. The exam tests whether you tackle cost at the right layer.