Statistical Process Control - Case Study Example

Running Head: STATISTICAL PROCESS CONTROL Statistical Process Control [Name of the Writer] [Name of the Institution] Statistical Process Control Introduction Statistical techniques are playing very important role in the quality control process in every organization; it helps to improve the quality of processes. Statistical process control (SPC) tools are using widely in order to measure and analyze the variation that may present in the processes.  Most frequently used for manufacturing processes, the objective of Statistical process Control is to examining and monitoring the quality of product and maintains processes to predetermined targets.  Many experts consider the American quality revolution of the 1980s a response to an American quality crisis that had reached major proportions. This quality revolution led to the overall integration of statistical process control (SPC) and total quality management (TQM) in many U.S. manufacturing sectors. Although manufacturers have used statistical process control (SPC) methods successfully for decades to find causes of poor productivity and to improve product quality, only recently has SPC caught the attention of motor carriers (Mundy, 1986, 25). Background The purpose of the original study initiated in 1987 was twofold: (1) to demonstrate the application of statistical process control to delivery performance and (2) to explore the potential of SPC in this area of operations. The research plan included three tasks. The first task was to select the hub-and-spoke terminal networks for analysis and the performance measures of interest. The second task involved data collection and processing from various mainframe databases to create dispatch records. These records contained date-time values for arrivals and departures at selected points along the route from origin to destination. A total of 55,989 complete dispatch records were collected for the period from January 30, 1987, to March 13, 1987. A computer program, written in Statistical Analysis System (SAS) language pieced together dispatch records to create 6,822 completed-shipment records. Other SAS programs generated random samples and computed the sample statistics needed to build control charts. The third task was to construct control charts, evaluate delivery service quality, and assess the potential of SPC techniques. Analysis Constructing process control charts and assigning causes of unnatural process variation are indispensable steps in any statistical process control program. Although these steps may appear uncomplicated, they present both conceptual and practical problems. The analysis of these problems, as well as the proposed solutions, is presented in two parts. The first part examines the problems related to charting delivery service. The second investigates issues involved in the search for assignable causes (Richards, 1989, 512). The purpose of this paper is to study the SPC tools applied by the Yellow Freight System, Inc., on their Less Than Truckload (LTL) delivery service. Charting LTL Delivery Service To construct useful control charts of Less than Truckload (LTL) delivery service, carriers must (1) define and accurately measure delivery service and (2) define a service lane in a way that meets both theoretical and practical requirements. The kinds of problems that arise when undertaking these efforts are discussed below. Service Measurement Yellow Freight System, Inc. considers the consistency of transit time to be the most critical element of LTL service quality. Transit time is defined as the number of days from pickup until delivery and is measured by subtracting the pickup date from the delivery date. Appointments and Refusals Appointments and refusals create two methodological issues. The first issue involves the selection of the proper delivery date to use: the actual, the tender, or the first-attempt date. The actual delivery date indicates when die shipment was delivered to the consignee (Michael, 1986, 5). Appointments and refusals make the use of actual delivery dates alone misleading. Fortunately, different kinds of control charts can be developed to satisfy each perspective. The R chart, which examines process variability such as transit-time consistency, would use the service-cutoff approach to measure transit time. The P chart, on the other hand, can show the percentage of shipments meeting (or failing to meet) expected delivery dates (Eugene, 1980, 154). Closing on Weekends Many customers close shop or reduce operating hours on weekends. These practices complicate the task of charting delivery service for several reasons. First, the practice of closing on weekends may severely bias measurements of the total days required for delivery (James, 1984, 60). Second, the practice of closing on weekends by customers leads many carriers to set delivery standards and measure performance in terms of weekdays. The use of weekdays to measure transit time, however, will severely bias the control chart results. Lane Definition The nature of LTL operations makes lane definition a challenging task for statistical process control of delivery service. The large LTL carriers use hub-and-spoke networks to organize activities such as pickup, dock handling, re-handling, linehaul, and delivery (Myron, 1988, 36). On the other hand, a precise definition of a lane would narrow the focus on the network to a specific shipper-receiver pair and a movement itinerary such as the following: shipper-satellite hub-hub-satellite-receiver. This study considered the following three approaches to this task. Service Standard Approach Yellow Freight handles the lane-definition task by organizing the hub-and-spoke network into service lanes. A service lane includes all the terminal pairs that are assigned the same delivery standard. For die network studied, this approach placed 257 terminal pairs into the three major service lanes shown in Table 1. Both X-bar and R charts were developed for the shipments that moved in the same direction on these service lanes and excluded shipments bypassing a hub terminal. The charts produced results, like those shown in Figures 1 and 2, indicating badly out-of-control delivery processes. This apparent lack of control primarily emanates from measurement and lane-definition problems. The high proportion of points at or near the lower and upper control limits (2.03 and 3.34) and the recurring weekly cycles in Figure 1 are signs that the chart incorporates different processes in the same plot. Specifically, the use of weekdays to measure transit time creates the following two processes: (1) the shipments picked up during the later part of the week that benefit from weekend periods not counted in the service standards and (2) the shipments originating on Mondays and Tuesdays that do not have that benefit. In addition, the range of distances among the terminal pairs assigned to a lane appears wide enough to create different (shorter- versus longer-haul) processes within that lane (see Table 1). (Control Charts: Montgomery, Douglas C. Introduction to Statistical Quality Control International Edition. 5. Edition - September 2004) As Figure 2 illustrates, changing die measure of transit time to total days does not solve either problem. A high proportion of points remain near to or beyond the control limits. Further, recurring weekly cycles appear in which shipments picked up on Wednesdays and Thursdays now show the worst performance. Figure 3 shows that many satellite-terminal pairs with the same service standard generate shipments moving considerably different distances, while many other pairs about the same distance apart have different service standards. Turn Approach Yellow Freight operations staff initially used the turn approach to set the lane standards. A turn represents the maximum hub-to-satellite distance that a driver can make in a round trip without exceeding the ten-hour legal driving limit. Yellow Freight uses a turn distance of 2S0 miles (see Figure 4). This approach first groups origin-destination pairs of satellite terminals by the number of turns that a shipment requires and then assigns lane standards to each group. (Control Charts: Montgomery, Douglas C. Introduction to Statistical Quality Control International Edition. 5. Edition - September 2004) To analyze the turn approach purely from an operations perspective, satellite pairs were grouped by die number of turns only. The local shipments processed through hub terminals (no satellite-terminal activity) were assigned zero turns. The shipments picked up or delivered from a terminal pair consisting of one hub terminal and one satellite terminal located within 250 miles of a hub were assigned one turn. The shipments moving through two satellite terminals — each located within 250 miles of the hub — involved two turns, and so forth. Table 2 summarizes the results of this lane-definition methodology. Analysis of variance (ANOVA) was used to see how well this grouping methodology worked. Specifically, a fixed-effects randomized block design was used to determine whether statistically significant differences in delivery service exist among the four (0, 1, 2, and 3) turn groups. The block design provided a way to control the variation in delivery service that results from weekend closings. The shipments were blocked into two groups: (1) Monday delivery and (2) non-Monday delivery. A random sample of forty shipments from each turn delivery combination was taken for the analysis. The results shown in Table 3 indicate that die block design was effective and mean transit time is significantly different for at least one of the satellite-terminal groups. The Tukey method of multiple pair-wise comparisons was used to examine differences among individual group means. As Figure 4 indicates, it can be inferred, with a 9S percent family confidence coefficient, that the mean transit time is significantly different for each pair of groups. In other words, the turn approach has created four homogeneous groups of satellite pairs in terms of delivery service. When die same kind of ANOVA approach was replicated on several other lanes, however, the results were mixed. Specifically, the mean transit time was not significantly different for some turn groups. Except for the originations and terminations through hub terminals, which generally had significantly shorter transit times than the other terminal groups, no clear patterns emerged. Modified Turn Approach The mixed results for die turn approach suggest the need for further refinement. One possibility is to define a turn in terms of time (ten hours) rather than distance (250 miles). Each turn would include only die satellite terminals that could be served in a normal manner within the ten-hour, round-trip limit. This modification would make a turn more sensitive to elements such as road conditions, topography, and traffic congestion. Another refinement is to use traffic density to adjust the terminal-pair groups that are created by using the turn approach. High traffic density enables the carrier to move shipments faster and more efficiently. Thus the terminal pair that regularly generates a relatively high volume of traffic might be upgraded by moving that pair from its original group to a group with fewer turns. These solutions, as well as accurate measurements, will enable die carrier to develop useful charts of delivery-service quality, which die carrier can use to: (1) assess delivery-service capability, (2) set lane standards commensurate with capabilities, and (3) monitor overall lane-service performance for deterioration or improvement over time. In addition, sales representatives can use die charts as evidence of service quality when dealing with shippers. Assigning Causes of Un-natural Variation Although control charts of delivery service performance have useful applications, these charts will not enable die carrier to identify, assign, and then eliminate causes of poor service quality. Control charts alone do not improve quality (Joseph, 1988, 49). SPC program managers will need to identify the proper unit of observation for linehaul dispatch, local dispatch, and dock operations. Since the basic task of local dispatch and dock operations is to process individual shipments, the shipment represents the proper unit of observation. By contrast, the correct unit of observation for the linehaul function is the trailer. In LTL transportation, each road trailer may contain twenty-five or more shipments. By using the transit times of individual shipments to construct control charts of linehaul service, the SPC program manager runs a very high risk of sampling identical observations from multiple shipments loaded in the same trailer. This outcome artificially reduces the process variation in the sample and thus introduces a measurement bias that distorts control charts, especially the R chart. Conclusions and Recommendations Charting LTL delivery service requires that special attention be given to lane-definition and service measurement. The carrier has to define a traffic lane that is both conceptually sound and practical for statistical process control of delivery service. The modified turn approach appears to be an effective way to create homogeneous subgroups of satellite pairs (Tickle, 1987, 41). The practice of closing on weekends by customers can severely bias measurements of total delivery days. Further, this practice leads both carriers and shippers to measure delivery-service standards and performance with business days or weekdays, and that measure can severely bias control chart results. The recommended solutions to these measurement problems are as follows: 1. For X-bar and R charts of transit time, two solutions are available. The first is to use total days and cut off service on the earliest of the following dates: tender, first attempt, or actual delivery. For internal purposes, the actual delivery is accomplished when a shipment is ready for delivery on weekends but must wait until Monday when the customer will open for business. The second solution is to create separate charts for the shipments that span weekend periods. 2. If P charts of shipments meeting required delivery dates are needed, develop one chart for the shipments that involve the weekend period and another chart for die shipments that are completed entirely within the weekday period. SPC program managers must give special attention to determining the proper unit of observation (Lisa, 1986, 72). . References C.E. Richards. "Monitoring Rail Transit Time Using Statistical Process Control," Logistics and Transportation Review 20 No. 4., 1989, pp. 512-513 Eugene L. Grant and Richard S. Leavenworth, “Statistical Quality Control”. 5th ed. (New York: McGraw-Hill. 1980). pp. 152-168. James H. Foggin, "Improving Motor Carrier Productivity with Statistical Process Control Techniques," Transportation Journal (Fall 1984), pp. 58-74: Joseph V. Barks. "Shipper Quality Survey." Distribution (August 1988). pp. 48-56. Lisa H. Harrington, "The Quest for Quality," Traffic Management (July 1986), pp. 71-74 Michael S. Galardi and Thomas Sanderson, A Service Quality Program for Motor Carriers (Lexington, Mass: Temple, Barker & Sloane, Inc., 1986), pp. 5-7 Montgomery, Douglas C. Introduction to Statistical Quality Control International Edition. 5. Edition - September 2004 Mundy et al., "Applying SPC.” 1986, p. 25 Myron Tribus, "What Can Private Fleets Learn from the 'Japanese Management Style?'," The Private Carrier (August 1988), pp. 32-37. W.R. Tickle and Cort J. Dondero, "Blueprint for Quality." Annual Conference Proceedings, Council of Logistics Management I (September 1987), pp. 33-44 Appendix Quality Control Morup (1992) said that the quality is one of the most significant and effective factor that a company can utilize in the battle for customers. To be viable, we should satisfy the customers. In order to be more competitive, we should please the customers. We can say that the Quality is the measure of customer satisfaction. Total Quality Management (TQM) is an organizational process that actively involves every function and every employee in satisfying customers’ needs, both internal and external. TQM works by continuously improving all aspect of work through structured control, improvement and planning activities that are carried out in concern with guiding ideology that focuses on Quality and Customer Satisfaction as the top priorities. TQM stress of the importance of zero defects and achieving the right target the first and every time. Variances in product are not acceptable and methods such as the Statistical Process Control (SPC) is use to achieve the objective. Zero defects are the result of an emphasis on prevention and diligent use of measurement, process control and the data driven elimination of waste and error. As Crosby said, "The purpose of quality management is to set up a system and a management discipline that prevents defects from happening in the company's performance cycle." Read More