Six Sigma Management

Real SCM Challenges. Real SCM Experts. Real SCM Solutions.™

HOME
Business Management
Career Management
Design Management
Environment
Inventory Management
Lean Management
Logistics Management
Project Management 1
Project Management 2
Purchasing Management
Quality Management
Risk Management
Six Sigma
Supply Management

SIX SIGMA MANAGEMENT
Six Sigma—Part I Learning to Look at the Data

By Jim Lamprecht
Jimlamprecht@earthlink.net

There are several unforeseen consequences or even deleterious side effects that result for the heavy-handed used of statistical software favored by Six Sigma courses (or course developers). Certainly one cannot deny that statistical software packages are a convenient means to analyze data but there lies the crux of the problem. Indeed, if one is repeatedly told that, in order to be an efficient (Black Belt or Green Belt) problem solver one must know how to analyze data then obviously, one feels compel to collect data and analyze them (preferably to death) using the newly purchased software. I do not deny that in many occasions the collection and proper analysis of data is one of the essential means to reach an informed decision but that is not the nature of the problem I wish to explore. What I wish to suggest is that the over emphasis on statistical methodology driven by the perceived need to use statistical software all to often unduly complicates or even introduces a bias during the analysis phase.

The problem

The problem is in part due to the fact that each of the half a dozen or so suppliers of statistical software tries to outdo the others. Over the past five to seven years this competition has led suppliers of statistical software packages to add more and more statistical power to their software. This drive to offer what is perceived as a better, more sophisticated products, has led to the introduction of unending improvements and hence, increased complexity. This increased complexity is evidenced by the seemingly unending number of statistical coefficients, statistical tests and statistical techniques that can be invoked by the, no doubt, overwhelmed user. This phenomenon would not necessarily be a problem were it not for the fact that Six Sigma course developers and hence instructors feel compel to introduce the unsuspecting participants to an avalanche of exercises which are partly designed to familiarize users with the software as well as expose users to as many statistical tools as possible. We are thus faced with the absurd scenario whereby more and more time (usually at least 25-30 percent) is spent explaining software commands and another 25 percent of the time is spent statistical tables and coefficients simply because they are part of the statistical package! Proponents of such an approach would undoubtedly reply that they only cover a portion of what the software offers and that is thankfully true but, I suspect that the reason why other methods are not presented to the participants is simply because the course developers do not know how to apply these “other techniques” within the context of a Six Sigma course. If they did, they would likely talk about it. But there are other problems.

Why not just look at the data

Once participants have collected data (not an easy task and at times a costly proposition), they must next learn to analyze the data using the appropriate statistical technique(s)—I say “techniques” because there often is more than one way to “look at” data. This brings me to another problem. In their haste to statistically “analyze” the data, participants are not told to first look at the data. Often, by simply looking at a print out of the data, or by plotting the data using a series of plots (e.g., Box-plot or histogram) or graphs (e.g., scattergrams or 3D plots), the ‘answer’ can be seen or better yet, an error can be uncovered. In their haste to statistically analyze numbers, users are often put in a position to produce several tables of probability coefficients that all too often prove little more than what was already evident by simply looking at the data or examining a series of plots. Indeed, in many cases, once the data are collected the results are for the most part already known, the statistical analysis only confirms what can be seen from the graphs or from the data itself. If the reader is skeptical, go back to your Six Sigma notes or to some statistical textbook and examine the data presented in some exercises and see if you cannot “see” the results in the data. Naturally, if you have ten variables to analyze over 1,000 observations, “seeing” the data will not be practical or easy but then again, analyzing the 90 possible interrelationships between all pairs of variables will not be any easier with a computer (I am ignoring three ways interactions).

I am not suggesting that running the data through some statistical package is a waste of time (although sometimes it is), I am only suggesting that the act of simply looking at what was collected can be very revealing. In Part II of this article we will contrast Quantitative vs. Qualitative methodology.

Previous Articles:

Six Sigma Problem Solving
Six Sigma Training Challenges
ISO 9001:2000 Meets the DMAIC Method

SME Solutions

ISM Solutions

CSUH Solutions

LeanSCM
Solutions