I recently attended the Advanced Problem Solving and Root Cause Analysis Workshop provided by Worchester Polytechnic Institute. Jim Leonard taught a structured root cause methodology for problem solving, which I found very useful. The steps in this comprehensive method include:

• to name the problem by creating the problem statement
• to gather information by asking a series of problem specifying questions
• to note distinctions and changes in the information gathered
• to brainstorm possible causes
• to test the possible causes against the problem specification, verify the true cause, and fix the problem

Emphasis was placed on gathering information to specify the problem through a series of control questions related to the "what, where, when, and size" of the problem. Multiple case studies demonstrated that the problem is usually half solved once the control questions are answered (which reminded me of the ending of the G.I. Joe cartoons from my childhood—"Now you know, and knowing is half the battle"). I have studied other root cause techniques including the 5 Whys,Ishikawa fishbone, events and casual analysis, and the informal interview. While I still believe these techniques are valid, I see each one as alternatives to the different steps of Leonard's problem solving process, but none as a complete structured methodology.

Another portion of the workshop was devoted to understanding when to use this problem solving methodology. In layman's terms, Leonard described the basic statistical concepts of variation and the difference between special-cause and common-cause. In a variation of the classic "how many 'F's are in this sentence," Leonard had each participant read and count the number of 'F's in a projected sentence. As we tabulated the number of F's counted by each person, it was interesting to see a range of answers. He refused to tell us the true number of F's in the sentence, claiming, "Too late, it shipped!" It was a clear demonstration of how variation is a reality in our industry, present in everything from measurements taken at incoming inspection to in-process test results. It was easy to see that the technique we were learning worked for finding the special cause of results falling outside of the normal variation of a process in control. If only the statistics course I took in college had been so clear!