[SysOp note: The electronic version of this file contains no figures or graphics. --Tom Glenn, the TQM BBS, 301-585-1164.] ----------------------------------------------------Business Index & ASAP------ AUTHOR(s): Cavinato, Joseph TITLE(s): SPC - getting serious with quality. (part 1)(statistical process control) Distribution p65(2) May 1992 v91 n5 It wasn't too long ago that the practice and image of quality consisted of a person walking the factory floor with a clipboard. True quality was left to precision industries like pharmaceuticals and engine aircraft manufacturers. For most others, like basic chemicals, textiles and steel, quality was a give-and-take between purchasing, manufacturing and the supplier's sales personnel. Today, the typical management of "quality" looks like an unbelievable waste. Yet, the following were standard business practice until just a few short years ago. * Delivery quantities with wide industry standard overage/ underage norms. * Specification tolerances with ever-widening brackets of what was acceptable to make and sell to customers * Delivery uncertainty and quality variation that led to large materials management and work-in-process inventories. * Regulated rail and motor practice did not allow for much tailoring of service for specific customers, so everyone got a fairly low average level of service on inbound deliveries. Entire engineering courses and a textbook body of knowledge about inspection grew up between the 1940s and 1960s. The options were simple: do a complete inspection of inbound goods, force the supplier to do it and hope that they follow through and do it, calculate a sampling scheme for inspecting only some of the items (needed if inspection destroys the unit), or don't do any inspection all. The problem with these approaches is that the inspection only catches bad goods or services after they were made. The act of inspecting rarely, if ever, got the message back to the original production line for change. Costs of Quality There are four key costs of quality. Appraisal is the act of detecting poor goods and services, but more important is the detection of the machine, person or process that is doing it. Prevention is the positive follow-up act of altering the system so that future goods and services aren't undesirable. If the time between production, appraisal and prevention is short, then the chance of producing many poor items is greatly reduced. Third are the costs incurred by the company for having produced poor quality items. These might be returns, rework and associated wasted inventory and processing costs. Fourth are the costs incurred by customers for poor quality. These might be their own downtime, extra efforts and their own customer service harm. In any case, these represent loss of future customer service and sales. The problem was, and still is, that only actual inspection costs are captured in standard budgeting and other costing systems. These have not caught up with the reality of the need to directly measure or accept the many actual and implicit costs of quality as a path to lower long-term costs and higher revenue. Logistics was a problem, too. Once the service was performed it was either good or it wasn't. Inspection was merely to measure what you got, but there was no way of making that shipment or service any better. But with the philosophy of statistical process control, the opportunity arises for measuring it while it is happening with the link to correction and preventing a future one from being out of kilter. The Heart of SPC There are a couple of basic logics to statistical process control. One is what is called "capability." This utilizes what the production equipment or service can normally produce in its regular course of activity and the customer either accepts or rejects this performance from the start. Another logic is to measure production while it is happening, checking to see if it is still producing acceptable "quality" items, and to stop and correct it if it is not. This attempts to produce a real-time feedback to the production equipment. It eliminates the problem of having to inspect goods produced months ago without any feedback ever getting back to production. A buyer's initial specification consists of a nominal spec (a length of 100 millimeters, for example) with an allowed tolerance of 2 millimeters. This spec indicates that most of the goods should be 100 millimeters, but an allowance of up to 102 and down to 98 would be allowed. In an SPC context, this would be seen as 100 mm being the average of all units but 99+ percent of them will vary within 102 and 98. This is owing to the fact that statistically this will appear in the form of a mean of 100 mm but a "bell curve" of actual items will fall within the 2 mm each way band. The mean and standard deviation are used to construct the idea of capability as illustrated in Figures 1 through 4. If all the units in an initial test run of, say, 500 units average 100 mm and the standard deviation is such that 3 standard deviations above and below are within 102 mm and 98 mm then the system is said to be "centered and capable." A system that is not centered (the average of all the units are, say 99.5 mm) but three standard deviations above and below this average still fall within the 102 mm and 98 mm requirement, then the system is said to be "capable but not centered." See Figure 2 page 65. A system that produces an average to the nominal spec of 100 mm but three standard deviations of that initial run of items span a range of, say 102.5 mm and 97.5 mm is said to be "centered but not capable." See Figure 3 above. Lastly, a production run that produces neither to the desired average nor within the tolerance is said to be "not centered nor capable" as shown in Figure 4 above.