This paper argues that software testing can be less thorough yet more efficient if applied in a well-managed, empirical manner across the entire software development life cycle (SDLC). To ensure success, testing must be planned and executed within an earned value management (EVM) paradigm. A specific example of empirical software testing is given: the Empirical Bayesian Stopping Rule (EBSR). The Stopping Rule is applied to an actual Department of Defense (DoD) software development to show potential gains with respect to archaic testing methods that were used. The result is that a considerable percentage of the particular testing effort could have been saved under usual circumstances, had the testing been planned and executed under EVM with the Empirical Bayesian Stopping Rule algorithm.
展开▼