How would you justify choosing request-response architecture over event-driven architecture for a new application, and when would you choose one over the other?

Request‑response is a synchronous, point‑to‑point pattern that returns a result immediately. It is simple to reason about, has low latency, and is ideal for CRUD operations or services that must reply within milliseconds. Event‑driven, on the other hand, is asynchronous and many‑to‑many; producers emit events to a broker and consumers react independently. This decouples services, allows horizontal scaling, and provides resilience through eventual consistency. A circuit breaker can be added to a request‑response chain to prevent cascading failures, but it does not solve the coupling problem. You would choose request‑response when the client needs an immediate answer, the interaction is simple, and you can tolerate a single point of failure. You would switch to event‑driven when you need to handle high throughput, bursty traffic, or when services must evolve independently. For example, a payment service that must notify inventory, analytics, and email systems can publish a "payment.completed" event; the inventory service updates stock asynchronously, while the API still returns a success status to the caller. In that scenario, event‑driven gives scalability and resilience, whereas request‑response would become a bottleneck.