At present, full sampling inspections of quality characteristics are used in the automatic assembly of switches, but even this does not eliminate all defective articles. Use of the Mahalanobis-Taguchi (MT) system to detect defective articles more reliably was studied and introduced into the production process. The measured characteristic was the operating-force-returning-force curve, which reveals many defective trends. Independently developed support software was used to convert raw data to optimum characteristic quantities (by grouping the data) and set factors. We also used a proprietary approach to signal typing and calculation of Mahalanobis distance from signals in studying conditions and evaluating appropriate scaling. The system we developed is able to respond to a change in switch model by deriving new optimum conditions immediately. After this system was introduced, there was a corresponding reduction in complaints; this was considered to be a result of use of the MT methodology.