User configurable software systems allow users to customize functionality at run time. In essence, each such system consists of a family of potentially thousands or millions of program instantiations. Testing methods cannot test all of these configurations, therefore some sampling mechanism must be applied. A common approach to providing such a mechanism has been to use combinatorial interaction testing. To date, however, little work has been done to quantify the effects of different configurations on a test suites' operation and effectiveness. In this paper we present a case study that investigates the effects of changing configurations on two types of test suites. Our results show that test coverage and fault detection effectiveness do not vary much across configurations for entire test suites; however, for individual test cases and certain types of faults, configurations matter.