Using Design of Experiments (DOE) in Lean
The Design of Experiments technique is used for statistically determining ways to improve an existing process while limiting the risk of a wasted effort. Experimental Design, as it’s also called, analyzes the relationship between the process factors and their results, in other words, helps to pinpoint the cause-effect relations within an operation.
Given that continuously improving the process, through waste reduction and an increase in value delivery to the customer, is one of the main goals of Lean, it’s no wonder the DOE method has found use in this realm.
Design of Experiments can be used to:
- define the optimal conditions for running a process
- identify current wastes and potential savings in material and energy use
- compare possible solutions
- foresee possible effects of the changes that are being considered - for example, during the Analyze phase of DMAIC
- reduce process variability.
Design of Experiments characteristics
The Design of Experiments tool is prevalent in the scientific domain, where statistics play a key role. But the technique applies to a much broader range of repeatable activities and is often used in Lean Six Sigma practice of process management and optimization.
In the use of Experimental Design, you may come across terms such as:
- Input factors
These are both the controllable inputs - e.g. the materials - that can be modified with each new experiment, as well as uncontrollable inputs to the process, for example, workers’ potentially unpredictable behavior.
They represent the varied duration, or level of force applied to an action within the process, for example, how long a part spends in a given machine, or under which machine setting level. It’s the extent to which the factor has been altered.
This is the measurable and, importantly, the replicable output of the experimental iteration, so its result.
How to apply the Design of Experiments approach?
Step 1: Define the goal
Make it clear what it is that the experiment is supposed to measure or analyze. Depending on the kind of question it’s meant to answer, you will need to reach the answer in different ways. For example, screening design will help you identify the factors with the highest impact, when you’re on the lookout for the cause of inefficiency, while comparative design will pinpoint the most important factor from among the few that are suspected to be causing an undesired result.
Step 2: Choose the factors
List the inputs to the process that you trust to be of the highest impact, and please note that less will be more, in this case. You don’t want to be trying to look at a multitude of factors at once - a couple at a time will be sufficient. As part of this step, it should also be made clear which process response (result) you will be looking at.
It’s also recommended to keep in mind the interactions of the factors at play. Take advantage of DOE making this possible, as opposed to the traditional “let’s try and see” approach. Pinning down an interaction of factors that causes the most damage to process efficiency will be of great value.
Step 3: Adjust the levels and run the experiment
Take note of the level to which the factors are to be modified. If multiple levels are being considered worth testing, which often is the case, you will need to do a separate run for each level to observe the various results.
It’s good to double-check the data before giving the experiment a go-ahead, to avoid mistakes. Commonly, even when not testing several different levels, it will still be required to run the test at least a few times, to replicate the results.
Step 4: Evaluate the results
Measure the outcome and note down the achieved results - be objective and do this in a quantitive, statistical manner. To do this right, teams typically turn to flowcharts, histograms, and scatter plots. It’s advisable for the person running the experiment to have sufficient statistics knowledge, to facilitate a correct setup and understanding of the responses.
How can your process benefit from the application of Experimental Design?
- The process can be tweaked and optimized based on measurable data, not guesswork, and multiple factors can be experimented with at the same time.
- It will make it easier to estimate the true weight of each process’ element and its impact on the rest of the system.
Thanks to Desing of Experiments, your process can be improved at the exact points where it matters the most, in turn making your product better, waste and costs lower, and revenue higher.
What’s more, through the process of factor and outputs analysis, DOE won’t only solve critical issues and improve process efficiency, it will also clarify to the team which factors are the most important within the production and need paying the highest attention to.