Statistical interpretation of absolute system performance requirements

For critical functions within a system, many times a customer specifies stringent performance requirements parameters in absolute terms. Strictly interpreted, these requirements must be met 100% of the time. Ideally, the system has to be tested for all possible combinations of input parameters for such functions. This is cost-prohibitive, especially in the case of complex systems for which it is not possible to identify all input scenarios. A statistical method based on percentile and confidence interval is developed to interpret these requirements. Under this method, a statistical sample is tested using hypothesis testing to establish system compliance/noncompliance to a specification within an agreed-upon confidence level. The method is developed under the assumption that the customer and the contractor have already negotiated to apply it to mutually agreed absolute requirements. The method was actually applied in establishing test plans for a real system. It is the authors´ intent to share the statistical method and test plan considerations with readers, and hopefully initiate the use of statistically bounded requirements in the specification of future systems.