The 12-question system design loop

Before going deep, answer these questions in order:

1. What problem are we solving, for whom, and what is explicitly out of scope?

2. What are the FRs, measurable NFRs, fixed constraints, and still-unproven assumptions?

3. What is the source of truth for each important piece of data?

4. What can happen at the same time, and can those operations conflict?

5. Where are the transaction boundaries?

6. Which operations can be retried safely, and how are duplicates handled?

7. Which interactions must be synchronous, and which can be asynchronous?

8. What happens when every dependency is slow, unavailable, or partially successful?

9. What happens at normal load, peak load, forecast growth, and an agreed stress scenario?

10. Who is allowed to do what, and where are the trust boundaries?

11. How will we know the system is healthy and meeting its SLOs?

12. How do we deploy, migrate, roll back, and evolve without breaking users?

Evidence: S2, S3, S23, S24, S46, S53