Process_Improvement_Plan

Process Improvement Plan Charles “Chris” Bozue OPS 571 March 14, 2011 Michael Kline Introduction In this paper the subject to discuss is the effects of any seasonal factors and confidence intervals based on the process of “getting to work by 6:45am.” Earlier investigation revealed bottlenecks and overcoming them through Goldratt’s Theory of Constraints (TOC). This paper will consist of discussing any control limits along with the calculations and data used to determine them. This discussion takes into effect the very limited collection of process performance data over an initial four-week period. Also to be presented will be the measures of confidence usefulness on the number of data points collected. Statistical Process Control The bottom line goal of Statistical Process Control (SPC) is to “arrive at” and “keep” process control. The process identified in week one of getting to work by a certain time is monitored through control chart tracking using the process flowchart identified in Attachment 1 and an Excel spreadsheet as shown in Attachment 2. The top of the Excel spreadsheet chart identifies the “in-control” requirements of the process. Daily measurements were laid into the chart. Comparisons between the in-control and “actual” measurements were used to identify any variations. After week one analysis, investigative analysis showed variations identified as potential process problems which were then modified. Because the process was time based, measurement will be accomplished by a process known as sampling variables that is measuring the amount of deviation from the set standard identified. The use of these statistical techniques of measuring and analyzing the variations recognizes the measurement of statistical process control. According to Chase, Jacobs, & Aquilano (2006), the standard practice is that statistical process control for variable is to set control limits out to three standard deviations for variable sampling. This particular collection of data does not conform to typical variable measurement, although concepts are considered. Seasonal Factors A seasonal factor is also called an Index and according to Chase, Jacobs, & Aquilano (2006), is the amount of correction needed in time series to adjust for the season of the year. Seasonal factors affected this process performance on only a few occasions. Unlike a production process, seasonal factors in excess of a norm do not have an adverse affect on the getting to work process because the arrival time is adjusted by upper management. This would be like an increased last minute order being cancelled at the last minute, there would be no affect to the production process that should be viewed negatively. It was identified though that seasonal factor over a longer period, like that preceding this charted data time period, would influence process performance negatively. Because of this, adjustments would have to be made based on further data collected over the seasonal period. Taking into account what is known currently, adjustments having to be made would be the amount of manual intervention allowed. This will be discussed further during the confidence interval section. Because this period of time will be considered under demand, multiplicative seasonal variation will be used and thus taken into account. This will be considered during seasonal (weather) changes as there can be an expected larger variation. During winter months, travel time increases resulting in earlier wake up times and less alarm reset ability as mentioned earlier. If information was collected over the past year or more, the seasonal information could be broken out into a seasonal factor that could be analyzed. Analysis would come through a simple proportion that would provide a factor. This factor could then be multiplied into the time value taken from wake up to work arrival. This would provide the seasonal increase or decrease in time needed to arrive by the appropriate time. According to the U.S. Census Bureau, “series whose seasonal effects come primarily from weather, the season factors are estimates of average weather effects but not for abnormal weather conditions.” Confidence Intervals Confidence Intervals, based on a percentage factor, provide an estimated range of values from which a probability that the numbers provided contain the true value or reliability of an estimate. In this process plan, the average arrival to work is 8:21 minutes ahead of time. Because a confidence interval is a range spanning from Lower Confidence Limit to Upper Confidence Limit, it encompasses the population parameter of interest with a degree of certainty as specified here as 95%. As shown in the graphs below, the red center line is the mean or control line of the process. Deviations are calculated based on the center line and are shown by the Upper and Lower control limit lines. By using a 95% confidence interval, calculated at 6.34, our arrival time upper limit is 1:47 minutes while the lower limit is 14:55 minutes. It is forecast that with a 95% confidence interval, arrival time to work will be made every time. This is just based on arrival time. Focusing on the bottleneck area, it was found that manual manipulation of the alarm reset was a major cause of process failure. Taking the same information, it was found that on average the alarm was reset 2.79 times with an upper and lower confidence limit that based on 95% of 2.79+/-0.8 which equated to two and four alarm resets.   | Arrival | Alarm Reset | Total Deviations sum | -156.00 | 53.00 | Count | 19 | 19 | Mode | -25 | 4 | Median | -7.00 | 3.00 | Mean | -8.21 | 2.79 | Standard Deviation | 14.11 | 1.78 | 95% Confidence Int. | 6.34 | 0.80 | Upper Confidence | -1.47 | 3.59 | Lower Confidence | -14.55 | 1.99 | Based on the information supplied here, it is forecast that as long as the manual intervention alarm shutoff is allowed no more than three resets, arrival to work will occur prior to established times with 95% confidence. As stated previously with the short time of analysis, this process can now be repeated enough times to ensure quality of data. Continuous Improvement will result as of this ongoing study. Conclusion Alarm reset had been identified as the main contributor to any deviation to the process. After applying Goldratt’s Theory, a decision was made that instituted a modification to the manual intervention by increasing alarm volume. Increasing volume improved on time rate from 40% to 86% using rigid time values. Taking acceptable deviation +/- 5 min into account, this equates to a 100% on time rate. The process of flowcharting, identifying bottlenecks, providing control limits and seasonal factors considerations improved this process by over 50% from the initial week making this a successful implementation. References Bozue, C. (2011). Process Flow Diagram (Rev ed.). Dayton, OH: Self. Chase, R. B., Jacobs, F. R., & Aquilano, N. J. (2006). Operations management for competitive advantage (11th ed.). New York: McGraw Hill/Irwin. U.S. Census Bureau. (2008). Manufacturing, Mining, and Construction Statistics. Retrieved from http://www.census.gov/const/www/faq2.html#two Atch1 Previously provided Flowchart Atch2