Controlling software reliability during development

The application of both classical and state of the art tools of quality control to the software development process are described. These tools and methods use actual test results and determine the range of failure rate expected, based on the anticipated customer use. These tools properly take into account the natural "clustering" of orthogonal defect discoveries during stochastic testing, consistent with Mandelbrot's work in Poisson processes. These tools are robust, have a history of success in other fields, and are easily implemented. They clearly define the bounds of the software reliability which are supported by the testing, and the probable risk that the software will be less reliable than expected. They enable the timely introduction of software without unreasonable risks of uncertain reliability.