February 09, 2003
Quantum Data
I've had this saying for years, I call it 'quantum data'. It's based on the idea of Heisenberg's uncertainty principle and Schroedinger's cat. The concept behind quantum data is that some information cannot be examined. And that if is were examined it would destroy the values being observed.
A more concrete example perhaps?
Let's say you have a database system that collects information for order taking. Handheld devices are in the field being used to eliminate many of the errors and delays found in previous methods. One result here is much more efficient and accurate orders. Along the way, however, it seems like other data is being collected that could be useful. This is where it gets problematic.
The goal of the effort was to decrease delays and reduce errors. There are some subtle behavioral issues that were involved in the process. By having the salesmen in the field directly entering the orders it eliminated a number of places where delays and errors could develop. Along the way the systems became much more aware of other information about the process itself. This is the quantum part.
The process of selling, ordering and supplying customers is a complex one. There's a whole host of subtleties that on the face of it seem calculable. When, in fact, they're not. They're much more involved than mere tabulation would indicate. Using the data collected about the process will change the process.
The salesmen in this situation will rebel if management thinks it can use the data collected about the selling process. This is where observing the process will change the process. One can't directly tabulate time on task for things like this. Oh you can certainly try but you run the risk of the decisions made being fed back into the system and seriously altering it. In that particular example, management wanted to make an immediate jump from having horribly long delays, fraught with numerous errors, directly into doing time-to-completion analysis on salesman performance. It was my estimation that this would violate the rule. The salesmen would end up paying their child to sit at home and punch the buttons very slowly just to keep management from forcing them into Catch-22 situations. As a result, the old situations of errors and delays would crop back up again. All because someone looked at data in ways that harmed it's development.
Time on task and other human behaviors are not as directly calculable as you'd think.
So my saying here is that you can't look at the data collected about the process without disturbing the process. It's 'quantum data' and can't be used without recognizing that will cause the very values it presents to change.
The next time a system evolves, ask yourself, what part of it can't be observed without harming the system as a whole? Think about over-analyzing web log reports or employee utilization reviews. See where paying too much attention to the process leads can lead to decisions seemingly disconnected from the process end up feeding-back into it and ruining it?
Now, do I have all the science figured out here? No, probably not and there's doubtless some other name for this observation. I've found mine works pretty well and that people understand it.
