It is clear that the burden of software maintenance increases proportionately with our inventory of software systems. Our inventory is increasing because we now recognize existing code as reusable assets rather than liabilities. With this recognition comes an understanding of the importance of evaluating code quality and maintainability. We show how maintainability metrics can be used to gauge the effect of perfective and adaptive maintenance on large industrial software systems. In a series of four studies we measure the maintainability of the source code before and after a prescribed maintenance activity. These measurements permit us to analyze the effect of the maintenance activities as well as evaluating the model's sensitivity to code change.