Choose between a platform build and a managed service under lock-in risk

• Identify which capabilities differentiate the business, required scale and compliance, acceptable vendor dependency, team expertise, and decision horizon. Separate price negotiation leverage from true portability.

• Model three demand scenarios and total cost including egress, support, engineering, on-call, upgrades, incident loss, and migration. Validate projections with a representative workload, not list price alone.

• Meet delivery, durability, regional, security, and support goals with a named owner. Preserve data export, contract clarity, and a tested exit path proportional to lock-in risk.

• Define a thin domain event contract, portable schemas, identity and observability owned by the company, and an adapter boundary around vendor-specific controls. Keep data retention and export mechanisms independent enough to rehearse recovery.

• Applications publish domain contracts through a stable client layer; platform-specific routing and operations remain behind it. Periodic exports and replay tests prove data can move, while not pretending engines are interchangeable.

• Choose based on required delivery, ordering, and replication semantics; an abstraction that hides meaningful differences creates false portability. Document which semantics would change on exit.

• Compare real incident responsibility, support response, regional failure, quota, and control-plane behavior. Build requires staffing and game days; buy requires vendor escalation, isolation, and service degradation plans.

• Evaluate key ownership, tenant isolation, audit, supply chain, data location, and privileged vendor access. Contractual controls complement, not replace, technical evidence.

• Own end-to-end delivery, lag, correctness, cost, and vendor-dependency telemetry. Track adapter exceptions because each one increases exit cost.

• Use scenario ranges and cost per business event. A managed premium may be rational if it avoids years of undifferentiated operations; internal build is rational when requirements or scale create durable strategic advantage.

• Abstraction increases option value but can constrain the product to a lowest common denominator. Abstract stable domain contracts and operational evidence, not every feature.

• Run a time-boxed proof with failure and export tests, negotiate exit terms, and define reevaluation triggers. If buying, rehearse partial migration; if building, earn scope with a narrow production cohort.

• Principal candidates connect architecture to operating capability and option value. They quantify switching friction honestly and recommend a decision with explicit triggers for revisiting it.

Interviewer evaluation rubric