In order to manage the emerging complex enterprise, managers must dramatically upgrade their sensor technology.
What does that mean?
It means that we must develop new ways to measure management performance variables. Most importantly, we must develop new ways to measure those aspects that are most dynamic in the organization's operations and environment.
This topic breaks down into three specific initiatives.
In a series of studies dating back to my dissertation, I sought new ways to measure the characteristics of complex management systems. If these measures are to be comparable across different products, firms and industries, they should fall into a class of measures that does not depend on overly specific units of analysis. In the beginning, I took my cues from the precedent in science and engineering of using dimensionless variables (e.g., mach number and reynolds number) to overcome differences in scale and materials. The price/earnings ratio (P/E) is a similar measure that is probably more familiar to businessmen.
Unfortunately, I had a hard time finding measures of operations that were both dimensionless and useful. This was really not surprising, since the few that exist in engineering and business are widely regarded as classics. As a result, I have directed my research to look for the next best thing: omnidimensional measures that can be applied the same way to any unit of analysis. For example:
These measures are enriched because, in business, it is possible to look at any measure from at least two perspectives. The first uses the measure in directly quantifiable terms. The second takes account of the fact that metrics are often seen as perceptions by the parties involved. Thus, on the one hand, it is possible to state a formal definition for product variety and apply it to actual product families. At the same time, it may be just as valid to ask customers for their perceptions of the degree of choice that a given manufacturer offers.
About a year ago, David Nembhard and I began to look at a lot of empirical data on worker performance at a large US manufacturer. This data was collected as newly-hired and reassigned individuals tried to acquire the skills to perform a specific manual assembly task. Our experience with this data has led us to consider a fundamental research question:
Is there some general way that mathematical functions can be used to characterize and describe the behavior of large populations of individual units?
We believe that there are many situations where a fitted mathematical function is a very reasonable alternative to a massive tabular database. We continue to look for ways to follow up on this belief.
We found that conventional learning curve models were unsatisfactory because they tended to fall into two categories. The first category viewed learning as a purely individual activity and tended to ignore the systemic effects that are so common in large organizations. The second category tended to view learning as an aggregate characteristic of the organization. This results in mushy, overall measures of learning that have a huge amount of unexplained variance.
David Nembhard and I have developed and tested a method for directly measuring the shape characteristics of the individual learning curves of large populations of learners. We have applied this approach to data on thousands of individuals in a large US manufacturer (involving 3,800 distinct episodes of learning). The resulting ‘learning maps’ suggest that there is a substantial and regular pattern of variance in the shapes of individual learning curves. This paper is now being revised for publication. (I have a working paper on this subject at my web site)
