It is a Monday morning. I am supporting a junior colleague, Maria. We are both fresh and energized. She is proudly showing me the improvement in the team actual performance indicators and suddenly her expression changes in a heartbeat. I asked her what’s in her mind and she points at a team velocity chart showing the team throughput over time and she says: “we have tried to get the same velocity every week for more than a quarter now, and no matter what we do, there is always variation. We have worked really hard at the quality of our planning. I don’t know what else to do.”
I asked her what she would expect to see instead, and she looked at me with a bewildered expression. “Don’t ask me stupid questions like that”, she says, “a flat line, or even better a straight line with a slope showing we are getting better over time.”
Maria, who is smart and competent, was suffering a severe case of “idealism”, she was expecting to see in the actual behavior of a real life system the ideal results she imagined could happen. We needed to have a conversation about variation in systems.
Everything I am sharing here is my admittedly poor understanding of the ideas on variation of Dr. Deming and Dr. Shewhart, any good idea is stolen from theirs and any useless bit is, of course, my own. If you haven’t yet go read them directly, they are giants. You better spend time with them than with me, you can thank me later. If you are more of a watcher than a reader Deming Institute has a YouTube channel with great videos of Dr. Deming too.
First important and big idea: variability is inherent to the behavior of any system. There is no such a thing as a perfect result all the time. There is variation in the mere act of measuring, whatever you are measuring or observing will never be single value all the team, even if you are measuring a universal constant like the speed of light. There is variation in all complex systems behavior, as the many interrelated parts interact, and everything that can be managed is a complex system. Many systems, while still being deterministic and not random, show complex behavior in which minuscule differences in initial conditions produce large differences in the future results. Add humans to the system and it won’t get any better, add human interactions and the emergent social behavior and… well, you get the point. You better expect variability in everything we do, because it will be there.
Second important idea there are 2 different types of variability in the behavior of a system:
- Common cause variability, which is originated within our system and inherent to it, like the quality of our estimates and the variations in our human performance.
- Special cause variability, which is created by changes in our environment that affect our system, like several team members getting sick at the same time, a lab fire, and a market shift requiring us to integrate with a type of backend we haven’t found or expect to find before. This source of variability is also present in all systems, but unlike common cause variation, it can often be attributed to the single event that created it.
A third idea is the different ways in which these sources of variation create system adaptation and respond to it. The more mature a system is, the lower the common cause variability can be, but it will always exist. Please notice I said mature, not old. This does not get any better with time, only with adaptation. If and when the system adapts to its environment better we can expect common cause variability reduced. On the contrary, special cause variability does not depend on the system maturity at all, so it does not tend to get better or worse. it fluctuates. No matter how mature the system, we can always expect challenges being thrown at us by special causes every now and then. They won’t be the same special causes all the time. Right, that is what makes them special after all, and the amount of their effects won’t be constant or predictable. Special cause variability will throw odd balls at our system unexpectedly and often to which the system will have to react.
That means Maria should never expect a flat throughput chart, but she can expect common cause variation to be controlled through adaptation and improvement. That means the difference between the bad weeks and the good weeks gets smaller as the team gets more mature. That reduction on the effects of common cause variation will also help Maria identify the special cause events so that her team can manage them and avoid some of the risks they poise for them.
The Adaptive Executive 