Statistical Process Control (SPC) is a method of analysing real-life data (typically the output of a process), that is taken over a period of time (and is therefore called a time-series) and automatically determining if the resulting process is under control or “out of control”.

A good description of what it is, is presented on this Wiki page

http://en.wikipedia.org/wiki/Control_chart

An example of the web application output is shown below

The blue line are your data points, the green line is the median point for your data, and the red lines are positioned a certain number of standard deviations away from the median. If the blue line crosses the red line at any point is means the process went out of control at that moment in time and you should try to analyse why that is (perhaps in a brain-storming session). The data for the above graph is shown below.

2010-05	62
2010-06	46
2010-07	22
2010-08	42
2010-09	84
2010-10	43
2010-11	39
2010-12	25
2011-01	22
2011-02	31
2011-03	33
2011-04	33
2011-05	21
2011-06	20
2011-07	20
2011-08	17
2011-09	34
2011-10	30
2011-11	60
2011-12	26

You can copy this data directly into the NavWin web application, into the box marked Paste Data Below. Set the Chart Width and Chart Height to the size in pixels. Press the Create Chart button and finally press the Update Control Point button.

Although it sounds like this type of analysis would be extremely complex, the actual calculation is extremely simple. It is more a leap of faith as to whether you are going to accept the information as useful or not. Personally I find it generally reliable.

The calculation used on this website is called the I-MR or sometimes X-MR (or Imr or Xmr). X stands for the set of individual data points {x} and MR stands for moving range. This technique was established in the 1920’s and used empirical data from thousands of real-life processes to work out an inherent rule of stability that applies to all complex processes (in particular complex processes involving people). The technique is meant for any discrete data. The data distribution (e.g. Gaussian, Poisson) is irrelevant.

The actual technique is meant to utilise two graphs (the X and the MR). However, for the sake of simplicity we only calculate the X part. If you want to get really serious and try to predict the moment when a process is about to go out of control; then use the MR chart also.

Is your process under control?

The key question this chart tells you is whether or not you are in control of a process you run. The results apply to any process; you just need to provide the time data. In addition, the chart can tell you when you lost control of a process so you can analyse what went wrong at that time.

Can you improve your process?

If you accept the results of the chart (and many managers will accept them), then you can easily demonstrate whether or not certain process changes are actually effective or not. In particular, you can demonstrate whether someone’s great new idea, which seemed to show big improvements in output, was actually well within the normal output fluctuations of the underlying process. Unfortunately, the person who you just brought down to earth will not thank you for it, especially if that person is your boss. Some forward thinking bosses are the exception.

Changing a process is hard

In general, once you have your data and your chart, it is very hard to make significant long term improvements by making small, weak-willed, low commitment changes. To break an established process out of a long term cycle usually requires some drastic effort and some big change to the way you work. You have to think out of the box. To break the cycle, it forces you to evaluate every part of your team behaviour, and strip away the ineffective activities and keep the good ones. It usually takes months so be prepared for the long haul.

The main thing now is that you can finally measure the impact of each process change as you apply them. Try not to do them all at once!