May 2001

Vol. 3, No. 5

Squeezing the chart:
Are you getting everything you can?

Origins often shape future developments. The bicycle seat, for example, was patterned after earlier technology--the horse saddle (and in fact is still referred to as a "saddle"). Keyboards are based on typewriter keyboards, derived from typology practices in newspaper publishing and the demands of linotronic typesetting rather than the electronic requirements of computers.

Charts, such as those that illustrate processes, derive from many of the conventions of the visual media of which they are a part. Because of these origins, it is easy to think of charts in the same way that one considers photographs: they represent a kind of snapshot of a process, after all. Furthermore, charts can be printed--a process that creates a version of a chart that can no longer be changed. Charts provide information, sometimes specific information such as that relating to the stability of a process. One prints the chart, considers the stability, and for more information, prints another chart.

To capitalize on the potential for each chart means really taking advantage of the electronic media, rather than bringing only those expectations related to visual media. Rather than printing a chart and taking it "over there" to show to someone, the way we show pictures of our kids and dogs, we can derive much greater information by taking advantage of the electronic medium itself. How is this possible?

When one looks at a chart generated by a statistical program such as SQCpack, the chart itself may look like hundreds of similar charts. It will have its plotted line of data, computed control limits, and dotted line indicating the mean of the data. So much for that. One can glance at such a chart, see that it reflects a stable process, and move on.

But wait, there's more. When a control chart appears on a computer screen, it is not a printed photo. Instead, it invites interactive information gathering and provides a variety of ways to "see" data. If an out-of-control point appears on the chart, for example, one can see what is responsible for that condition, simply by clicking on the out-of-control point itself. Already, additional information has been provided by the same chart. When one explores further, the same chart can yield subgroup information that might provide insight into different shifts, different operators, alternative dates, a Pareto chart with results of out-of-control tests, etc. With CHARTrunner, limits can be computed right on a chart. For example, a user could compute a set of limits on the first half of the data, note the mean, then compute another set on the second half of the data to see if the mean has shifted.

One normally thinks of our charts like this: over there is a pile of data. Take a Polaroid camera over there and take a snapshot of the data, then bring it back to look at. If someone wants to look at it some other way, ask a number of questions, then just go back over there and get a different snapshot.

New technologies always insist on a new way of seeing (see Joel Barker's tape, "Discovering the Future"). Rather than seeing charts as snapshots, their capabilities demand thinking of them as flexible tools through which one can easily and interactively view a data set from many different perspectives with very little effort. The user sets up one chart, the software leads a guided tour of the data set as the user points the way towards something that he or she wants to explore, and the software finds the treasure.

Seeing software with new eyes will lead to greater control of data, better understanding of its implications, and improved ability to ask the right questions in data analysis.



Copyright 2001 PQ Systems.

Please direct questions or problems regarding this web site to the Webmaster.