In my last posting I said that I would present a case study using what I originally called the Ultimate Improvement Cycle (UIC) but is now referred to as TLS. This case study involves a company that had been manufacturing truck bodies for the transportation industry since 1958 and had been one of the recognized industry leaders. The company had a staff of 17 full time Engineers and Engineering performance was measured by the number of hours of backlog waiting to pass through Engineering, which seemed odd to me that such a negative performance metric was being used. I had been hired in as the VP of Quality and Continuous Improvement because the company was losing market share as well as delivery dates being missed. In addition, morale within Engineering was apparently at an all time low. Upon arriving at this company I was informed that I was also to have responsibility for Engineering. Because of the poor performance in Engineering, the company had fired their VP of Engineering and I was the “lucky” recipient of this group. Apparently the backlog of quotes had risen from their normal 300 hours to approximately 1400 hours just in the previous 2 months.
My first step was to create a high-level P-Map and a VSM to better understand what was happening. I was in search of the constraint or that part of the process that was limiting throughput. It was clear immediately that the constraint was the order entry system in that the process to receive a request for a quote and deliver it back to the customers was consuming 40 days! And since it only took 2 weeks to produce and mount the truck body, it was clear to me why market share was declining. I then created a lower level P-Map of the quoting process to better understand what was consuming so much time.
My next step was to develop a run chart to get some history of what was happening within this process. Figure 1 represents the number of backlog hours from February 1999 through April 2000. The data confused me because I first observed a steady decline in backlog hours beginning in May of 1999 through October 1999 and then a linear increase from November 1999 through April of 2000. My investigation revealed that the decline was associated with mandatory Engineering overtime and the ascent occurred when the overtime was canceled. The Engineering VP hadn’t defined and solved the problem, he had just treated the symptoms with overtime.
My next step was to create another run chart going back further in time. If I was going to solve this problem, I needed to find out when this problem actually started. Figure 2 is the second run chart I created and it was clear that this backlog problem was a relatively recent change, so the key to solving it was to determine what had changed.Figure 1In fact, somewhere around January of 1999 the incumbent VP of Engineering had decided to move on to another position outside the company. His replacement apparently didn’t like the way the Engineering Group was arranged. The company produced three basic types of truck bodies and historically had three different groups of engineers to support each type. These groups were staffed based upon the ratio of each type of truck body ordered and produced. Between June of 1996 and December 1998 the company had maintained a very stable level of backlog hours. The new VP of Engineering consolidated the three groups into a single group and almost immediately the backlog grew exponentially. The company fired him in May 1999 and hired the new VP that used overtime to bring the backlog hours down. Part of the problem was the performance metric in place which caused abnormal and unacceptable behaviors.
For me, the fix was clear. Return to the previous Engineering configuration, identify the system constraint, focus on it and drive out waste and variation, to drive the backlog hours lower. The results were swift and amazing in that the backlog decreased from 1200 hours to 131 hours in 5 weeks and has remained within an acceptable maintenance level ever since. In fact, because we also eliminated much of the waste within this process the time required to process orders through Engineering decreased from 40 days to an astounding 48 hours! Figure 3 represents the final results of our efforts. Not only had the amount of backlog hours decreased to levels never seen before (from a historical average of 300 hours to less than 150 hours), the number of quotes completed were at all-time highs for the company.
The key to this success was to first identify the operation that was constraining throughput (the system constraint), second, decide how to exploit the constraint by applying various Lean and Six Sigma tools to the constraint, third, subordinate everything else to the constraint and then if need be, break the constraint by spending money. Thanks to Lean and Six Sigma, we did not have to spend any money. And the good news is, Engineering overtime was virtually eliminated while the market share rose to unprecedented levels!