Understand Variation

Last time I talked about the mistake of attributing performance to individuals rather than the system of work. One of the things that comes from an understanding of variation due to a system is why it is a mistake. But that is not the only thing. Understanding variation is one of the major pointers to what you should do next to improve, and sometimes more importantly, what you should NOT try to do.

In the 1920s Walter A. Shewhart worked a Bell Laboratories on trying to improve quality in the equipment they produced. He wanted to find a way of telling when you should act on a measure and when you should not. I.e. if you were measuring something in a system when would it be worth your while to chase down that figure to find the source of the problem and when would it be better use of time and resources to do nothing. He came up with a technique called Statistical Process Control (SPC).

The SPC tool comprises of charting a run of measures and also calculating, from the measures, the average and Control Limits. Usually there will be an Upper Control Limit (UCL) and Lower Control Limit (LCL). The theory says that if all the points lie within the control limits then the system that produced the points is “in control”. Another way of saying this is that the system is predictable. You can predict that given the system does not change, future points will also lie within the limits. If there are points outside the limits then the systems is “out of control” and therefore unpredictable.

The points inside the limits are said to be from “common causes” and those out side the limits are “special causes”. The common causes come from the system and should NOT be acted on. The specials causes come from out side the system and should be investigated and the causes removed. This gives a formula on when it is economic to act and when not.

If you have special causes you need to work to remove them to get the system in control. If the system is in control, you need to work to reduce the causes of variation.

Here is the thing, all systems produce variation. Think of stock markets, sales figures, time between accidents, the number of calls a contact centre receives each day. They all vary and you need to know if that variation is predictable and when something special has happened that needs investigation.

There will be more on variation as we go. But the thing to remember is that variation is everywhere, but that we shouldn’t act on every single point until we know that it is a special cause and then we shouldn’t rest until the cause is removed.



If you found this interesting