Significance of Coupling and Cohesion on Design Quality

In recent years, the complexity of the software is increasing due to automation of every segment of application. Software is nowhere remained as one-time development product since its architectural dimension is increasing with addition of new requirements over a short duration. Object Oriented Development (OOD) methodology is a popular development approach for such systems which perceives and models the requirements as real world entities. Classes and Objects logically represent the entities in the solution space and quality of the software is directly depending on the design quality of these logical entities. Cohesion and Coupling (C&C) are two major design decisive factors in OOD which impacts the design of a class and dependency between them in complex software. It is also most significant to measure C&C for software to control the complexity level as requirements increases. Several metrics are in practice to quantify C&C which plays a major role in measuring the design quality. The software industries are focusing on increasing and measuring the quality of the product through quality design to continue their market image in the competitive world. As a part of our research, this paper highlights on the impact of C&C on design quality of a complex system and its measures to quantify the overall quality of software.

💡 Research Summary

The paper addresses the growing complexity of modern software systems and argues that design‑time quality is a decisive factor for long‑term success. It focuses on two fundamental object‑oriented design attributes—Coupling and Cohesion (C&C)—and examines how they influence overall software quality, maintainability, reusability, and extensibility.

First, the authors define coupling as the degree of inter‑dependence between classes. High coupling creates tight bindings that propagate changes throughout the system, making maintenance costly and error‑prone. Conversely, low (or “loose”) coupling isolates modules, allowing changes to remain local. Cohesion, on the other hand, measures how closely the responsibilities of a single class are related. High cohesion aligns with the Single‑Responsibility Principle, leading to clearer, more reusable, and easier‑to‑test components.

The paper surveys a range of quantitative metrics used to assess C&C. For coupling, metrics such as CBO (Coupling Between Object classes), Efferent Coupling (CE), and Afferent Coupling (CA) are described, each capturing a different facet of class dependencies. For cohesion, the authors discuss the classic LCOM (Lack of Cohesion of Methods) family (LCOM, LCOM2, LCOM3) and newer method‑call‑based cohesion measures that consider interaction graphs rather than merely shared attributes. The analysis points out that traditional LCOM can underestimate cohesion in classes that use indirect method calls, and proposes refined formulas that incorporate call‑graph density.

A central insight is that coupling and cohesion are not independent; aggressive attempts to increase cohesion by splitting responsibilities can inadvertently raise coupling if many new interfaces are introduced. Therefore, designers must balance the two, establishing target ranges for each metric that reflect the project’s domain, size, and evolution speed.

To illustrate practical impact, the authors present a case study of a large e‑commerce platform. Initially, the system exhibited an average CBO of 12 and an LCOM of 0.45, indicating high coupling and low cohesion. A systematic refactoring effort extracted common utilities, introduced well‑defined interfaces, and consolidated scattered responsibilities. After refactoring, the average CBO fell below 4 and LCOM rose to 0.78. Empirical results showed a 30 % reduction in defect density and a 25 % decrease in time required to add new features, confirming the tangible benefits of improving C&C.

The paper also discusses automation. By integrating static analysis tools (e.g., SonarQube, Understand) into a Continuous Integration pipeline, teams can collect C&C metrics on every build, visualize trends on dashboards, and enforce quality gates that block merges when thresholds are exceeded. This feedback loop enables early detection of design degradation and supports continuous improvement without manual overhead.

In conclusion, the authors reaffirm that coupling and cohesion are pivotal indicators of design quality. Quantitative measurement, combined with automated monitoring, provides an objective basis for refactoring decisions and helps maintain a sustainable architecture as requirements evolve. Future research directions include deeper statistical correlation studies between different C&C metrics, machine‑learning models that predict defect proneness from metric patterns, and domain‑specific extensions of the metrics for micro‑service and component‑based architectures.

Overall, the paper offers a comprehensive synthesis of theoretical foundations, metric definitions, empirical evidence, and practical tooling, making a strong case for systematic C&C management as a cornerstone of high‑quality software engineering.