Operational cost

Architecture choices incur ongoing cost: monitoring dashboards, on-call rotations, security patching, capacity planning, incident response, and the expertise to run each component. A microservice that saves dev time but triples pager load may be a net loss.

Include toil, mean time to recovery, and the bus factor for niche infrastructure (Kafka, service mesh, custom operators).

On interviews: quantify operational cost alongside dev velocity; explain what you would simplify if on-call burnout appeared.

Common pitfalls: adopting Kubernetes or a mesh without staffing; no SLOs so every alert pages equally; operational work unpaid because it is "not a feature."

Checklist:

• Estimate on-call and runbook burden per component.
• Staff platform expertise or buy managed services.
• Tie architecture reviews to SLO and incident data.
• Sunset components that cost more than they save.