What is DOE?

Design of Experiments (DOE) is a large subject. It’s a general phrase that refers to a broad range of experimental techniques. So, when someone says that they’re going to use DOE, what they mean is that they want to learn more about a process via an experiment, and that they’re going to use one of the techniques that comes under the umbrella of ‘DOE’.

There are many different levels of experimentation, from small, localised experiments on production processes, to large scale, global clinical trials. In Six Sigma, we generally focus on the use of controlled experiments, in which a set of input factors are modified in a controlled and structured manner, throughout a sequence of trials. The overall aim is to understand which process inputs have a statistically significant affect on the process output.

What are the different types of a DOE used within Six Sigma?

In general, the types of DOE used within Six Sigma projects fall into three categories:

1) Full Factorial Designs involve very possible combination of the input factors (with each input factor having two or more distinct levels). So, if the experiment involves four input factors, each of which can be set at two levels, there will be 16 different combinations in the experiment. By incorporating every possible combination, these experiments gather lots of data, but can also become very large (and costly).

2) Fractional Factorial Designs do not use every possible combination of input factors, but instead use a carefully selected sub-section of the Full Factorial Designs. By selecting the trials carefully, these experiments can provide more information or use fewer trials – in other words, more ‘value for money’.

3) Response Surface Designs are more complex experiments that can also detect and quantify non-linear relationships between the process inputs and outputs. The results can then be used to develop optimal process settings.

Where does DOE fit within the DMAIC Framework?

In Six Sigma training, DOE is sometimes positioned in the Improve phase, because it can be used to optimise a process.

Our view is that DOE is predominantly an Analyse tool. Designed experiments help you to understand which of your process inputs are important and which are not. They also help to model the relationship between those important inputs and the process outputs – much like regression. Once the important input factors have been identified, they can be adjusted and better controlled, in order to improve the process output.