Statistical quality control: how to detect unreliable software components

Abstract

If "classical" testing-strategies are used (no usage-testing) the application of software reliability growth models may be difficult and reliability predictions can be misleading. Nevertheless statistical methods can be successfully applied to failure data. This paper presents an approach which allows the detection of unreliable software components and the comparison of the reliability of different software versions-even if testing is done in a classical manner. A simple to use graphical method-which is mainly based on the sequential test of Wald (1947)-is described. The methodology was successfully applied to a software system for tax consultants.