Abstract

Taguchi has defined "loss to society" as those losses incurred when a product's characteristics do not meet the customer's tolerances, performance measures, serviceability, expected life, etc., for which it was designed. Often these characteristics are not clearly known nor clearly specified to the supplier.

This study was undertaken with the cooperation of a manufacturer of temperature control devices to determine whether the concept of Taguchi's loss function could be applied to a specific product, and how the use of this cost function would compare with the "conventional" method of edstimating quality costs.

The "conventional" method assumes that there are no "losses" unless the product characteristic falls outside the customer's specified tolerances whereas the Taguchi cost function assumes that there is a "loss" (quality cost) whenever the product characteristics deviates even an iota from a stated target value.

A specific model temperature control device was selected (one of a large family of such devices) for which product history was available. This data included an average year's production of this device as well as the number of items returned by customers as not performing to warranty specifications, the average factory cost of manufacturing the product, and the average rework costs per unit for items rejected at final inspection.

Samplings of specific temperature measures were analyzed and found to be approximately normally distributed. The population mean and standard deviation was estimated, and using these estimates projections were made as to the expected number of devices to be reworked based on the manufacturer's acceptance sampling plan used at the final assembly stage.

The total expected annual in-house costs due to rework (not meeting customer tolerances) plus those of the expected costs of warranty returns as well as the expected cost of loss of customer goodwill for products not meeting specifications and not identified by the acceptance sampling plan were compared to the expected rework and return costs using the Taguchi cost function and his proposed production tolerance desired from his cost function.

It was found that by using the Taguchi quality tolerances (deviation from a target value) almost 98% of all proudcts would be rejected in-house and subject to rework. Almost no out-of-tolerance products would be shipped to the customer. The total expected annual cost thus incurred would vastly increase the cost of producing each unit over the procedure using the customer tolerances wherein no expected losses would occur unless the product exceeded the customer's specifications.

It was proposed that experiments be conducted to reduce the standard deviation of the temperature control device since six standard deviations slightly exceeded the customer's tolerances. At this time the company is actively considering such experiments.

Conclusion: Even though the exact form of the "societal" loss function may never be known, it is possible to follow a simple step-by-step methodology to determine:

1. Whether the product performance natural tolerance limits exceed or fall within the customers tolerance
2. Whether a formal investigation of customer tolerances should be undertaken.
3. Whether it would be economically feasible to conduct experiments to identify the significant causes of the natural variance of performance measures
4. Whether it is economically feasible to reduce those sources of variation.

This study suggests that the use of the Taguchi loss function provides an excellent base to provide a manufacturer with facts to enable him to continue striving to meet the goal of continuing quality improvement while reducing the costs of providing the improved product, thus truly meeting the general challenge for the latter eighties as well as the nineties.