One of the most important commonalities I see in successful DMAIC projects is being able to find and eliminate the contrast. Contrast, in my opinion, is one of the best ways to solve problems because there is a natural and readily available comparison. Contrast indicates that either a shift in the mean or a reduction in the variation (or perhaps both) will solve the problem--after all, not every part is not performing as intended.
To illustrate this, I am going to tell you the story about one of our suppliers that produced a part for us called a gap hider. A gap hider is a hinged part that is situated between the back of the rear seat and the cargo load floor of our SUVs. The purpose of the gap hider is to keep the seat hinges covered when the rear seats are laid flat, for when you are carrying larger items in the back end. The gap hider, in order to properly function, uses two torsion springs. These torsion springs are attached to both hinged sides of the gap hider by simple insertion holes. When I was asked to help this team, they were telling me that the torsion spring was ripping through these holes on some parts. It was clearly a type of shear failure of the material.
After touring the plant and asking questions about the manufacturing process, I asked to see failed parts and non-failed parts, about five of each. I got some really strange looks with that request, but I told the team that we can learn as much from parts that are functioning properly as we can from parts that are not functioning properly. Even when every part is not functioning properly, there is usually contrast between the worst of the worst and just barely bad. Clearly, if the torsion spring was tearing through the material, either the material was too weak or the force from the spring was too large. I chose to look at the torsion springs first, and found that the spring force on the bad parts was equal to the spring force on the good parts. This indicated that the torsion spring force was not the critical input variable. Therefore, my attention shifted to the material. My team tried to tell me the solution was to make the material thicker or to change its composition, but I wasn't convinced. I told them "why should I change the design if not every part is failing?"
When I performed pull tests on the good and bad parts, the bad parts had highly variable strengths, while the good parts were very consistent. And that's when I saw it--literally. When I looked at the torsion spring insertion holes on the good parts, they were sharp and well-defined. On the bad parts, they were rounded--almost like they were sucked down instead of punched. This was leading to variable localized material thickness near the holes, which was causing the highly variable pull strengths. Remembering the manufacturing process parameters, I spoke with the process engineer and we designed an experiment with the processing parameters (vacuum level and time, temperature, and punch force magnitude) and used a Likert scale to rate the "sharpness" of the holes in the DOE parts. We were able to change two of the process parameters to get consistently sharp holes and the pull force performance and variability significantly improved with it.
Contrast can be identified in many ways, but don't forget to include your senses--sight, sound, and touch. Also remember that the key to solving with contrast is to find a contrast in the input variable that creates the contrast in the output variable. This is how I was able to eliminate the torsion spring as the input variable--there was no contrast in the torsion spring force, for example.
In about two weeks, I will continue the series with the next topic for effective DMAIC problem solving--focus on physics and logic, leverage the math. Until then, I wish you success in your problem solving.