Interview prompt
Problem context
Skills being evaluated
Use the sequence below to surface constraints, choose boundaries, test failure behavior, and defend trade-offs. Concrete numbers are interview assumptions, not claims about a real production system.
Clarify the decision
- Identify actual workflows, contractual obligations, business value, missing target capabilities, security risk, and who can accept continued exposure. Separate active use from comfort and undocumented dependency.
Establish scale assumptions
- Measure users, queries, data, custom features, support incidents, full run cost, and target migration effort by cohort. Attribute the cost of exceptions visibly.
Functional and non-functional requirements
- Provide a credible replacement or explicit product decision, preserve required data and audit, communicate dates, support migration, and stop indefinite silent exceptions.
High-level architecture
- Instrument every dependency, create compatibility exports or adapters only for bounded gaps, offer migration tooling and shadow reports, and freeze new legacy feature development. An exception registry records owner, cost, risk, and expiry.
Data model and flow
- Migrate data with checksums and report parity, preserve stable historical identifiers, and publish target freshness and semantic differences. Archive legacy output under defined retention.
Consistency and transaction boundaries
- Run dual reports only through a validation window; after authority moves, corrections occur in the target. Avoid permanent synchronization that turns sunset into a second architecture.
Failure modes and recovery
- Canary business units, retain rollback for a short accounting period, and rehearse historical retrieval. A missed deadline escalates to the named executive risk owner rather than silently extending.
Security and privacy
- Block new sensitive data and users on legacy, isolate access, patch within a reduced scope, and disclose residual risk. Exceptions cannot override legal mandates.
Observability and SLOs
- Track active use, parity, migration blockers, support load, security exposure, exception age, and avoidable spend. Publish cohort progress to sponsors.
Capacity and cost
- Compare building missing target features, maintaining adapters, and ongoing legacy run cost. Charge exception cost to the sponsoring portfolio where governance permits.
Alternatives and trade-offs
- Perfect feature parity may never justify itself; redesign workflows or accept limited loss when business value is lower than perpetual cost and risk.
Evolution and migration
- Freeze, instrument, migrate easy cohorts, resolve true target gaps, set final dates, disable writes, archive reads, then decommission infrastructure and credentials.
What Staff and Principal candidates should emphasize
- Principal candidates make the business own exceptions and provide respectful migration support. They use architecture, economics, and risk together rather than issuing unenforceable dates.
Decision trade-offs
Feature parity
Option A
Rebuild every legacy feature
Option B
Migrate the valuable outcome and redesign low-value workflows
Recommendation:Fund parity only for capabilities whose business value exceeds long-term target complexity; offer export or process change for the tail.
Exceptions
Option A
Allow informal deadline extensions
Option B
Executive-owned, costed, expiring exceptions
Recommendation:Make exceptions explicit decisions with cost, risk, owner, and expiry so delay competes transparently with other investments.
Follow-up interview questions
- 01What if one executive refuses the deadline?
- 02How much feature parity is enough?
- 03When can legacy reads be archived?
- 04Who pays for an exception?
Common weak answers and mistakes
- 01Publishing a sunset date before understanding workflows and target gaps.
- 02Rebuilding every obscure feature for nominal parity.
- 03Extending deadlines quietly and making the platform team own business risk.
- 04Decommissioning runtime but forgetting data, credentials, exports, and audit retrieval.
Interviewer evaluation rubric
Sends deprecation notices or mandates migration without usage evidence, parity plan, or accountable exceptions.
Inventories cohorts, freezes features, supplies tools, validates outputs, sets dates, and archives required history.
Costs exceptions, aligns sponsors, redesigns low-value workflows, manages security exposure, and removes all legacy assets.
Changes portfolio incentives so powerful stakeholders make transparent business decisions rather than externalizing perpetual platform cost.