In order to continue to make progress in software measurement, as it pertains to reliability and maintainability, there must be a shift in emphasis from design and code metrics to metrics that characterize the risk of making requirements changes. By doing so, the quality of delivered software can be improved because defects related to problems in requirements specifications will be identified early in the life cycle. An approach is described for identifying requirements change risk factors as predictors of reliability and maintainability problems. This approach can be generalized to other applications with numerical results that would vary according to application. An example is provided that consists of 24 space shuttle change requests, 19 risk factors, and the associated failures and software metrics.