"I've only been asked to troubleshoot a production process one time. We had a resistance welding line joining the two halves of a stainless steel, sheet-metal clamshell with a perimeter weld. I wasn't an expert on resistance welding, but I got on the plane and went. When I got there, I looked at the parts, which had molten metal squeezed out the ends, cracks where the weld started, and periodic blow holes through the thickness of the sheet metal. I couldn't understand how an existing process could be this far out of whack. Not knowing what else to do, I asked for the welding machine manuals. I read them. I compared the settings on the machine to the settings in the book, and they had nothing in common. I reset the feed speed, the electrode pressure, the current and a few other things, then we sent some more parts through the line. I had them prepare a few quality check samples, and they looked good. We ran some more, checked again, and those were fine, too. I asked a few questions about how the machine settings had gotten so far off, and it was clear that this had been a slow process of adjusting the settings based on superstition and feelings until it was finally so far out that the process blew up. I posted the correct settings at the machine, got on the plane and went back home. In the end, all I did was read the instructions.

I realize that most production problems are much more complicated. I have watched colleagues struggle with them. If I had responsibility for a process that wasn't producing the desired result, I would first verify the basics - like I did with the resistance welder. I would do the obvious things: figure out when the problem started, look for some other factor that I could correlate with that. I would look to see if other machines performing the same process were still giving good results. I would quantify where we were at and where we needed to be - maybe it was a dimension that wasn't right, for example. And if nothing simple popped up very quickly, I would be looking to apply my Shainin Red X training to structure the approach and resolve the problem. Often, these kinds of problems are not simple and involve a lot of interactions that can only be found when you apply a structured and statistical approach, once you see that the answer is not obvious. This takes time, but produces better results."