To make software better in view of its maintainability, its software development process must be controlled and continuously observed. Researchers and software managers have stressed on the early measurement of maintainability starting from design phase itself so that timely steps could be taken for producing maintainable software. This paper evaluates and compares several methodologies for improving the numerical stability of a fuzzy-logic-based maintainability metrics system. Fuzzy parameters are adjusted using heuristic methods. A number of alternates were considered, in which training data sets were generated using different methods and these sets were used to evaluate objective functions in GA and accordingly fuzzy parameters were tuned. After conditioning, real projects’ maintainability data is used to show that fuzzy model performance is increased, however marginally, after conditioning