Maximizing customer satisfaction by optimal specification of engineering characteristics

Abstract

The House of Quality (HOQ) has been widely discussed as a mechanism for capturing the Voice of the Customer and guiding the process of converting the customer's voice into internal engineering specifications. However, the discussion of this process to date has tended to be qualitative rather than quantitative. Various heuristic practices have evolved with little substantiation of their value or evaluation of their impact other than illustration in a few case studies. Examples of practices promoted by several authors and practitioners include the inclusion of the relative magnitude of effects, but not direction, in the body of the matrix; the restriction of relationships to an integer scale of subjective, ratio values; and allocating available engineering resources to the hard-to-improve characteristics (because the easy ones will presumably take care of themselves). The process for setting target specifications given the HOQ data has not been well defined either. Design teams are required to develop their own ad hoc rules. When multiple engineering characteristics affect a single customer attribute or when an engineering characteristic impacts multiple customer attributes, the design problem becomes difficult and current procedures are of little use to the design team. This is particularly true when some interactions are negative. The result may be lost opportunity from the selection of non-optimal designs. This research takes a scientific approach to optimally specifying the target values for engineering characteristics in a competitive marketplace. A methodology for building linear or nonlinear statistical models of customer preference (value) for each customer attribute followed by an optimization routine that will obtain the optimal specification of engineering characteristics is proposed. The optimization model considers economic and technological constraints as well as customer preferences. The process results in optimal, feasible design specifications. In addition the effect of variability in parameter estimates on the ability to correctly identify the key engineering characteristics for optimizing customer value is investigated. Model performance is also analyzed through random generation of 320 problem instances of varying linearity, size, density, monotonicity and correlation. Finally, the methodology is applied to a problem from industry and the results are compared with an actual, subjectively-derived design.