Each module included in a software system poses a differential risk of failure in operation. However, neither traditional software reliability measurement techniques nor traditional development and testing methodologies consider the fact that the consequences of various software failures will be very different. This paper demonstrates how to measure the risk of failure, or expected value of loss resulting from errors in different modules. Module risk is a meaningful measure of quality that can be exploited throughout the software life cycle. We describe how the risk of software failure was estimated for modules in a commercial lending system. Analysis of the application and its use was employed to determine the financial exposure associated with the system. The potential exposure of each module in the system was measured by relating the individual modules to external system use. Module failure likelihood was estimated prior to use and updated with operational results.