Constructing a Stability Index for Software Using Probability Ratio Sequential Testing

In system-level software testing, one practice that is commonly used is stability testing. Stability testing is a "best practice" that supports the decision that a product is field ready. Typically, a stability test will last a few days. In this paper, we explore a technique that supports quantitative evaluation of product readiness for field service. It is also possible to apply this analysis to nonoverlapping subsets of the product, such as subsystems or user types. A stability index constructed in this way is useful for determining sensitivity to variable customer profiles, balancing the mix of categories of test cases, and optimizing total duration of stability testing.