Finding Answers

1/ Go back to basics - physics, chemistry, maths etc. to really understand the problem.

2/ Decide how you would solve it, with out cost/timing/complexity considerations.

3/ Now ask various suppliers how they'd do it. Steps 1 & 2 will allow you to ask intelligent questions, evaluate their solutions, raise potential problems and filter out supplier biases.

4/ Modify the original solution so it conforms with the "best" solution.

5/ Bask in glory at another job well done.

1. Talk to the floor personnel, the people involved with the problem, get real feedback

2. Bias is a part of feedback, expect the floor, management, salesfolk to all give a certain biased report;

3. Now, decide which tests have been run, which tests should be run and lastly, how to accurately derive at a reasonable result or conclusive non-result

4. Each report will consist of part or all of the following: people, process, machines, materials, measurement and the environment;

5. Test problems may have resulted from a special cause of variation: power spike, cable cuts, change of personnel on a machine or function of duty, etc.

6. A Pareto Chart, actually two, a bar chart and a cumulative line graph, are Big guns to help begin to Identify a problem, be it a test problem or a product problem;

7. I believe we as problem solvers are unindated with just Too Much data for most rework or problem solving situations and that is a problem unto itself, so Just Use the Available Data and apply that data to a Pareto chart to begin to analyze the most active contributor of a possible problem (or answer - depending on your point of view)

8. "Drill down" is a science, insulating out "Noise" and "Clutter" from decision making is key, and in my case, Time is always of the essence, so without a precise list of data to Drill, substitute a Checklist Analysis;

9. The resulting "Cluster" of instances, occurances and the like resulting from a Checklist helps pinpoint a desired area of concentrated measurement or focus;

10. Ishikawa (cause and effect) Root Cause Analysis is a "work backward approach" used to define root causes of frustrating process results, i.e., Weird Data, etc. CAUTION avoid typical think tank Tar Pits - avoid getting caught in "circular analysis" C=B=A=C, by getting enough feedback from each point of the identified problem!

Traditionally 4% of any business error is resulting in 50% of the total business waste.

This can also apply to a problem solving mantra (You do not require boatloads of info!)

Problem solving can only be accomplished when Focusing upon the correct problem area, then begin to break down the components of the problem (test results: people, machines, environment, changes, etc.) Avoid Missing The Vital Few while combing through the Trivial Many!

Really, this question might beg Have We encountered a true anomaly? How can we be sure?

STABILITY ANALYSIS - upper control limits, lower control limits. Apply Standard Diviation and really Use Your Control Charts! (I see them peppered across a war room wall,yet few can logically and quickly explain what I am seeing) Yes, you must esimate the average and standard deviation usig theaverage and range of your sample or problem area data, but due to the very lack of huge amounts of data (huge costs and huge info disagreements) we use smaller samples (1k lots yielding 2 to 10 samples, or 1 sample or 11-25 sample sizes, depending on your choice of charting)

Each chart must be adjusted for formula an compensate for varying samples and populations - enter Dr. Shewhart famous set of constants developed back in the 1930's ... crazy huh? Forget the science behind the science and just use the available Monitoring Process to Analyze Your Charts, and that is how you might begin to look at unusual or weird test results or problems "unidentified fluctuating order" UFO's.

OK - Now you might consider the input from a Vendor. He better know what he is talking about however!