The planning of software maintenance is a major problem that information systems managers are facing in many organizations. The problem stems from two basic root causes: the competing demands (services, repairs, improvements, etc.) for the same limited maintenance resources and the uncertainty associated with these demands. Managers lack reliable tools that could enable them to handle the uncertainty and thereby proactively plan the maintenance to better respond to user requests. This study provides a quantitative method to estimate the uncertainty and to help managers forecast the demands of maintenance. Since the accuracy of maintenance request forecast depends on predictable regimes of demands, the method first hypothesizes a non-stationary, life-cycle model for the basic regimes of different types of maintenance requests. Each regime features a probabilistic dominance of certain types of requests in the overall distribution of demands. The method then includes ways to characterize the distributions of different types of requests, monitor their evolution, and determine the shifting of regimes on which the managers base their planning decisions. This study performs a laboratory experiment to validate the quantitative method that determines the software maintenance life cycle. The findings confirm the hypothesis and substantiate the appropriateness of the quantitative method proposed.