Scatter Diagrams And The Connection With The Stratification Principle


A scatter plot or scatter-graph is a type of mathematical diagram using Cartesian Coordinates to display values for two variables for a set of data. The drawback to using this method is that employees at the shop floor level very seldom have the necessary background for using such an ‘advanced method’.



  • The basic reason for dealing with stratification is that it enables an effective causal analysis to be carried out and so improves the design of effective prevention methods.
  • An effective causal analysis will only be effective if measurements of production results are supplemented by data on the most important causes, e.g. by data on people (which operator), materials (which supplier), machines (type, age, factory), time (time of day, which day, season), environment (temperature, humidity) etc.
  • Without such data, it can, e.g. be impossible to determine whether the cause of a particular quality problem can be narrowed down to a particular operator, or whether it is due to something completely different.


Example 1:

  • A scatter diagram which has been constructed by a QC circle in Hamanako Denso, a company which belongs to the Toyota group.
  • The QC circle at the coil-winding section, which consisted of a foreman and nine female workers, had problems with a high break rate for coils.
  • The break rate was equal to 0.2% and the circle members decided on an aggressive goal which should decrease the break rate to 0.02% within six months.
  • By using flow charting, cause-and effect diagrams, data collection (check sheets), Pareto diagrams, histograms and scatter diagrams they succeeded in decreasing the break rate to 0.01% within the six months.


Example 2:

  • The relationship between the ordinary profit and the company size measured by the number of full-time employees in the Danish printing industry is presented.
  • This is an example of where there is no direct causal relationship between the two variables.
  • It was not possible to construct and measure a causal variable in this example and hence it was decided to use a simple variable instead which correlated with the cause system.