Statistical comparison of final weight scores in quality function deployment (QFD) studies

Purpose – Quality function deployment (QFD) is a methodology to translate the “voice of the customer” into engineering/technical specifications (HOWs) to be followed in designing of products or services. For the method to be effective, QFD practitioners need to be able to accurately differentiate between the final weights (FWs) that have been assigned to HOWs in the house of quality matrix. The paper aims to introduce a statistical testing procedure to determine whether the FWs of HOWs are significantly different and investigate the robustness of different rating scales used in QFD practice in contributing to these differences. Design/methodology/approach – Using a range of published QFD examples, the paper uses a parametric bootstrap testing procedure to test the significance of the differences between the FWs by generating simulated random samples based on a theoretical probability model. The paper then determines the significance or otherwise of the differences between: the two most extreme FWs and all pairs of FWs. Finally, the paper checks the robustness of different attribute rating scales (linear vs non-linear) in the context of these testing procedures. Findings – The paper demonstrates that not all of the differences that exist between the FWs of HOW attributes are in fact significant. In the absence of such a procedure, there is no reliable analytical basis for QFD practitioners to determine whether FWs are significantly different, and they may wrongly prioritise one engineering attribute over another. Originality/value – This is the first article to test the significance of the differences between FWs of HOWs and to determine the robustness of different strength of scales used in relationship matrix.