A Deliberate Provocation

Avoiding Man Made Chaos, Donald J. Wheeler, Making Sense of Data p190-192, 2003.

I'm sorry. I can't help it. This insight moves me so strongly that I really need to share it and see if I can get feedback from other people.

Here it is. A series of statements so strong that I can do little more than reproduce them:

"..description is not analysis. Description may allow you to make comparisons, some of which may even be meaningful, but it will not provide the needed insight into the system that produces the descriptive values. In fact, the usual descriptive measures tend to make the underlying system invisible.

Yet the essence of management is predictions, and while prediction requires knowledge, explanation does not...Which is why a new approach to presenting and interpreting data is needed. The old ways simply do not do the job.

The use of process behavior charts for presenting and interpreting data represents a new paradigm in data analysis. While the technique is simple, the revolution in thinking is profound...

While it may be interesting to compare the sales volumes for different regions, such a comparison is mere description, not analysis. ...The sales volumes for one region simply do not provide the proper context for interpreting the sales volumes for another region. When it comes to analysis, the only meaningful comparisons are comparisons of a measure with itself over time.

Regions are different. No data are needed to support this claim. It is self evident. Data may be used to describe the difference, but not all differences in the numbers are due to differences between the regions. Since cross-region comparisons are essentially "apples to oranges," it is difficult to determine how to filter out the noise in such descriptive uses of data. However, when a measure is compared with itself over time, process behavior charts provide a way to separate the routine variation from the signals of exceptional variation.

...process behavior charts provide the insight you need to improve both good and bad processes. When a process is unpredictable it will be worthwhile to eliminate assignable causes of excessive variation. When a process is predictable any improvements will require fundamental changes in the process. While this is directly contrary to much conventional wisdom, it has been proven to work time and time again, in all types of endeavors all over the world.,

This resonates so strongly with all my experience in building and maintaining databases for business and non-profits that it rings me like a bell. Yet, outside of the 6 sigma world I see almost no mention of it on the web or on forums. It really makes me wonder.