With the widespread adoption of Microservice Architecture and Domain-Driven Design (DDD), establishing a test strategy that aligns with software architecture is crucial. However, existing test classifications often suffer from ambiguity or from a lack of explicit alignment with software architectural concerns, leading to increased communication overhead and inefficient test structures in practice. This paper proposes a method that formally redefines well-known test levels—Unit, Integration, and E2E—based on “process boundaries” and “external I/O,” and deductively maps them to the responsibilities of a typical layered architecture. Furthermore, we present a case study applying this method to the development of a large-scale, mission-critical SaaS in the financial domain. Through an empirical evaluation in a production environment, the results demonstrate that the proposed approach significantly expanded the scope of assurance while maintaining execution speed and reliability, contributing to the overall optimization of the testing strategy in complex distributed systems.