Improving Software Maintainability Through Modular Enterprise Design

Abstract

Maintenance, not initial construction, consumes the majority of the lifetime cost of enterprise software, and the architectural decisions that most strongly predict future maintenance cost are the decisions about how a system is decomposed into modules and how those modules are permitted to depend on one another. This paper synthesizes the classical module-decomposition and software-metrics literature with the contemporary microservices and cloud-native literature to propose LATTICE, a Layered Architecture Technique for Tracking Interface Cohesion and Enterprise-maintainability. LATTICE treats modularity not as a single design-time decision but as an ongoing, measurable property of a system, combining (a) information-hiding-based module boundaries, (b) a quantified coupling and cohesion baseline derived from a design structure matrix, (c) change-impact tracking across releases, and (d) governance thresholds that flag boundary erosion before it accumulates into unmaintainable coupling. The architecture is illustrated through a design structure matrix representation of a disguised enterprise order-management system, a comparative table contrasting LATTICE-guided modular design against ad hoc and strictly layered design on four maintainability-relevant dimensions, and an illustrative case scenario tracing a single change request through a LATTICE-governed codebase. All quantitative figures are explicitly labeled as illustrative estimates constructed to demonstrate the framework rather than as measurements from a production deployment. The paper concludes with a discussion of the limits of metrics-driven modularity governance and directions for empirical validation.