Improving process capability is difficult, especially when one party assumes all the responsibility.
Recently, I had to support a team getting ready to launch a product. But the customer held up approval since a part feature did not meet specification.
Evaluating Process Capability
We often use a pre-production run to check the quality of parts. In most cases these parts get made under ideal conditions. When parts get made under ideal conditions we say such parts came from a system of common causes. When we measure a feature of parts made by a system of common causes each data point is indistinguishable from any other data point. This just means we cannot know why each data point has a specific value thus the reason we say it is random.
Don’t Tweak the Process Based on Random Data
What happens when a non-random pattern is present in the data? We investigate it and try to find a cause. So what happens when there is nothing worth investigating? We can tinker with the process but that will only add more variation. Deming demonstrated this in his famous funnel experiment. As the saying goes – don’t touch it if it ain’t broke.
Has this Ever Happened to You?
Imagine the following. A part feature displays a random pattern of data and fails to meet specification. What element of the process do we improve? The truth is we need to improve several elements of the process not just one. Sometimes this is possible with effort and in other cases we need a technological leap. But there is another alternative. The customer could open up their specification. This, in my experience, can often be a reasonable course of action.
The Joint Venture Statistics – Cp
One measure of process capability is the Cp statistic. I call it the Joint Venture Statistic. It’s a ratio of the tolerance spread (USL-LSL) over the process spread (6s). The customer controls the tolerance spread and the supplier controls the process spread. If both parties want to improve Cp, they need to ask these questions.
1. Does the process spread display a pattern of random variation. If the answer is YES then the next question is:
2. Can the customer increase the tolerance spread to yield a desired level of capability? If the answer is YES then problem solved! If the answer is NO then the next question is:
3. Is a reduction in the process spread realistic? If so, is there TIME to research and deploy alternatives?
Compromise is Possible
If we can stay true to these questions there is always a reasonable compromise. In my case, the customer removed the special characteristic from the print. This dropped the need to prove capability. In the meantime, they issued a deviation to begin production. This bought more time for the supplier to improve the process. In time, research efforts identified a solution that doubled the capability of the process.