Operation Improvement Inc.

It seems like everyone is talking about “Data” and “Data Analysis”. Of course, what everyone wants is actionable data!

A “play it where it lies” approach to data simply treats any data as a math problem. There are just numbers to be crunched and fed into statistical algorithms with only a loose reminder of their connection to the real world. We then have results and answers without conviction – UNactionable data.

Actionable data always has an explicit logical chain of thought beginning with first-hand observation of something. Then we sharpen our observations with attribute or variable measurement by making a systematic comparison to an objective standard. (We count and quantify.)

With physical characteristics this is often straightforward, but do we all agree on the objective definition of: a NEW customer? an EXCELLENT customer service call? full COMPLIANCE to policy or contract deliverables? a DEFECTIVE part? an IMPROVED process? an OUTAGE in our IT systems?

With even the best measurement system in place, we still have two recurring measurement quality issues: outliers and missing data. Have we investigated outliers as opportunities to learn and integrate new information or do we pretend that they don’t exist? And, what about missing data? Some missing data can be interpolated, but other missing data should also be treated as outlier information.

GPS position and location tracking data might be used to extrapolate an in-between location and time, but missing refrigerator temperature data might indicate an equipment malfunction or failure!

If, without grouping errors*, we correctly begin to reduce good data with descriptive statistics; then human ideas/abstractions will emerge from the numbers. We begin to confidently speak and accurately describe the typical “new customer”, “service call”, “warranty return” and so on; and we can graphically illustrate the analytical results that back up these sweeping generalizations.

Our confidence in these conclusions rests on our ability to trace all the way back to what could be seen, heard and experienced, and that is what makes this data actionable.

(A “very bad” example of interpolated values: