RELATING PROCESS MEASUREMENTS To CUSTOMER DISSATISFIERS

Through the work of W. Edwards Deming, manufacturers have seen the benefits of improved quality and variation reduction on customer satisfaction. These manufacturers have sought to eliminate variation, expecting that all variation reduction would affect customer satisfaction similarly. Yet, there appears to be little understanding about how specific process variation leads to customer complaints. Thus, manufacturers have not been able to tune their quality improvement methodology This internship focused on understanding what plant processes and critical dimensions, measured through existing plant systems, were most significant in predicting customer complaints. By targeting variation reduction efforts on these critical locations, overall customer satisfaction could be improved without the expense of reducing overall variation. A technique is presented to correlate plant measurements with customer survey data. When combined with detailed measurements, this approach could be used to predict the probability of a customer complaint, given the critical dimensions. However, such data is not currently available in most plant measurement systems, so a second model shows how critical measurements can be obtained that can be used to focus variation reduction efforts. Although the current data available did not allow for such detailed analysis, the overall methodology is shown to be sound through correlation with existing data and surveys. Such a model could be used to accelerate critical design tradeoffs and the process of allocating tolerances. The requirements for obtaining data useful for this technique are discussed, along with current inhibitors in the company studied and recommendations for implementation. These inhibitors include the push for high production numbers, lack of the information and organizational infrastructure required to distribute this data, and the primitive collection and storage processes for measurement data. Inhibitors to variation reduction in general within plants are discussed as well. These include mismatched process ownership, where responsibility is given without the required authority, and incomplete benchmarking, where industry-leading plants are studied, but the improvements required are not filtered throughout the company's plants. Thesis advisors: Roy E. Welsch, Professor of Statistics and Management Science Daniel E. Whitney, Senior Research Scientist