# The Manager's Toolkit - Part 4: Scattering the problem

## Scatter Charts

The Cause and Effect chart was introduced as a method of finding out what causes could possibly influence an effect you see. The chart is really a list of possibilities and you will naturally want to check if the cause in the chart really does influence the effect. In other words: are the cause and effect related?

This is the step that many people leave out, they sometimes list the causes but then they get their exercise for the week by jumping to conclusions and assuming that the first cause they identify is the main one.

Just because Event A comes before Event B it is not necessarily true that Event A must cause Event B.

In the Western world we have become very hooked on the idea of causality. If I change the oil in my car and one week later the engine seizes up then it must he the oil that caused it! The fact that the car is 25 years old and was on its last legs anyway has nothing to do with it. This is also true of places like the USA where everything has got to be somebody's fault so you can first blame them, and then sue them for it. The idea of accidents and coincidences seems to have lost favour in the West. Event A does not necessarily cause Event B and they may be totally unrelated.

The scatter chart is a quick and dirty test for possible cause and effect relationships. It does not prove that one variable causes the other but it does make it clear whether a relationship exists and the degree of scatter gives a good idea of the strength of the relationship.

## Power cuts cause babies!

For instance it has been shown that where there is a power cut in a major city then nine months later there is a rise in the birth rate. This is not to say that the babies are caused by power cuts and that you must keep your wife indoors during a power cut, in fact..., but that is another story. There is a probably a weak relationship between power cuts and the birth rate that would be revealed by a scatter chart. Two examples of scatter charts are shown. In the first the two variables are probably related and in the second the two variables are probably independent.

## Essential tips

• If the variables are related but the spread or scatter is very wide then the relationship may not be direct. There is probably another factor that you have not thought of that is varying and affecting the result but you are not measuring or controlling it. Finding this other factor (usually by cause and effect charts) is generally important in reducing the variation and scatter.
• The line does not need to be straight; it can be curved as well. After all you are normally only interested in increasing or decreasing one of the variables.
• Get as many points on the diagram as possible. The more the merrier is the rule for scatter charts.
• You can do sophisticated multiple regression analysis (?) to calculate correlation coefficients and statistical significance using a computer but in practice it is simply necessary to establish that there is a relationship. Don't go near a computer until you have tried the figures out on some graph paper!

Scatter charts are useful to find if a relationship exists and to allow you to solve a problem in the quickest and easiest way. They are a natural extension to cause and effect charts and should be used to get the best result.