In the last blog we discussed the elements of the CRT, with the promise to share a CRT from a real situation.
Without revealing company names, let’s start with some background on the particular company. They are a major producer of electronic components, mostly in the form of circuit cards. They are major supplier to other companies in the electronics industry. The plant was configured with seven (7) major assembly lines. Most lines were dedicated to certain types of boards, but there was also several with cross-functionality. In other words, the same type of board could be produced on more than one line. The most notable problem, and the reason they called us, was they were suffering from very high levels of WIP and not being able to meet on-time delivery demands from the customer.
We started our analysis with them by interviewing the workers on the line. We were first looking for the perceived UDE’s that existed.
The UDE’s provide a very important piece of the puzzle you are trying to solve. But, BEWARE: Not ALL UDE’s are really UDE’s. It’s important when you collect UDE”S to have people write them down in the form of answering a question. For exmple; “When I think of the current system, it bothers me that…” The “that statement” becomes the UDE. The more people you talk with, the better the UDE list will become. Another important factor is to note the commonality between statements. Five or six different people might all say something different, but all six mean exactly the same thing. When you find an UDE that fits this category – you’ve found an important UDE. It is also important to filter the UDE’s – to separate those emotional statements from logical statements. As an example, suppose during the UDE collection someone responds back to the statement with “It bother me that my boss is an idiot!” No matter how true that statement may, or may not be, it is an emotional statement and not a logical statement. Spending the necessary time on the front end to filter the UDE’s can translate into a much smoother process when constructing a CRT.
With that said, here is the final UDE list developed for this company:
1. The front of the line is measured in utilization minutes.
2. The back of the line is measured in boards per day.
3. RM’s are sometimes not available for production runs.
4. Testing takes too long to complete for some boards.
5. FTC and CQA perform the same function.
6. Some test equipment is not effectively used.
7. 100% of the boards are tested.
8. Boards can be rejected for cosmetic reason and not functionality.
9. Some batch sizes for some boards are too large.
10. Some FG’s sit in testing waiting for transfer to FG inventory.
11. Testing is not considered part of the production line.
From the 30 or so different UDE’s collected the list was reduced to the above list. Each UDE seems to be a separate problem with no clear correlation between them, and each is causing its fair share of Undesirable Effects in the system. So, the hunt was on to discover correlation between the UDE’s and surface a probable Root Cause.
Constructing the CRT
With the UDE list we are trying to build correlation between the entities. In other words, are there any two of these entities where one can cause the other? When you find those two it becomes the starting point to build the rest of the CRT. Continue building until all, or most of the UDE’s have been connected. Figure 1 shows how these entities were connected to show the CRT. You’ll notice that the entity boxes each contain a number at the top. This is nothing more than an entity address. This method helps when scrutinizing using the CLR’s to be able to point out entities quickly in order to make a connection. Those entity numbers with an asterisk (*) were entities from the original list. You will also notice some entity numbers without an “*”. These entities surfaced during development of the CRT as predicted effects and additional causes from the CLR’s
Using the CRT to formulate the sufficiency based logic you can see that from the original UDE list we were able to show cause-effect-cause relationships between all of these undesirable effects. The root cause in this example WAS NOT an entity listed in the original list, but rather a root cause that was exposed because of the CRT. In this case it was policy constraints. I say constraints as plural because this company has so many measurements they were trying to record, and some of these measurements were in direct conflict with each other. You’ll notice at the bottom of the CRT the two measurements – one for minutes and one for boards. In their mind high machine utilization was equal to producing lots of boards. They had very expensive equipment and the only way they could justify the equipment was to keep it busy all the time. Because of this measure they continually loaded the system with work, which created a vast work-in-process (WIP) inventory, much longer lead times, and consequently missed due dates.
In the end, policy measurements were changed, or eliminated, and the system was structured as a Drum-Buffer-Rope system with the test equipment being the drum. By using the test equipment as the drum we were able to release (“pull”) work into the system at the correct rate. This was a much different environment than trying to “push” work into the system for the sake of efficiency. The overall WIP reduced dramatically, the lead-times were shortened to hours rather than days, and on time delivery skyrocketed. Revenue jumped $350M in 6 months time and all because a CRT helped them understand what the real root cause was.