Quality Management

A Guide to Design of Experiments in Six Sigma

Written byContributing Writer
|
Updated onJanuary 15, 2024

First designed in the early 1980s, Six Sigma has been the hot trend in quality management for several decades. Businesses have strived for a competitive edge by implementing Six Sigma for individual processes and overall operations in their organizations.

Expertise in Six Sigma has thus become the hallmark of a good quality professional. In this article, we will go through one of the fundamental components of Six Sigma—Design of Experiments.

In this guide, we will discuss DOE, what it is, its benefits, its importance, and its implementation. We will also talk about a Lean Six Sigma program that can prepare you for a lucrative career in quality management.

What is the Design of Experiments or DOE?

DOE stands for Design of Experiments. DOE is defined as a systematic approach to evaluating the effect of various factors on a process. It is a statistical tool for planning, executing, analyzing, and interpreting controlled tests for efficient data collection and analysis.

It works by manipulating multiple inputs to identify and examine their effect on the output. It was developed by Ronald A. Fisher in the 1920s and is also referred to as experiment design or experimental design.

Also Read: A Guide to Six Sigma Projects

The Importance and Benefits of DOE

Although experimentation is decades old, design of experiments brings structure and quantification to this process. Naturally, it has become massively popular in the Six Sigma approach to ensure continual quality improvement. Here are a few reasons why it is important:

It helps to examine the cause-and-effect relationship between factors.
It enables you to input several factors that can be manipulated to evaluate their response.
You can examine the factors’ individual and combined effects.
It is useful in detecting a critical response that may be observed only during the simultaneous application of multiple factors.
You can conduct experiments for either the full or a partial set of factors.
It provides a systemized approach to problem-solving rather than randomly dabbling in unnecessary factors.
You can run trials for the complete range of experiments as DOE helps identify the critical factors and levels.
You can establish the most optimal settings for the factors.
You can use DOE to determine the statistical significance of your experiments and responses.

Further, this statistical method has found universal application for certain key benefits that it provides. Let’s delve into what they are:

Time savings: DOE results in significant time savings as it helps to focus on the most decisive factors and the levels rather than going by intuition or blind choice.
Cost-efficient: DOE helps save money on useless experiments by helping you determine exactly which factors must be manipulated and what responses to look for.
Process enhancement: Processes are improved by obtaining pertinent data with the help of suitable experimental runs.
Quality improvement: DOE facilitates the identification of the factors that can enhance product quality. It guides you on how to implement the experimental learnings.
Better understanding of complex systems: DOE helps to lay any complex system threadbare by systematically separating each critical factor, assessing its response, combining multiple factors, and then assessing the synergistic response. This helps to evaluate each element of a complex system without any confusion.
Superior decision-making: Concentration on relevant factors, a structured procedure, and detailed analysis help to make better and more informed operational and financial decisions.
Greater innovation: DOE assists in identifying the gaps in the current market or research and planning experiments accordingly.

Online lean Six Sigma training will equip you with a solid foundation in the Design of Experiments and set you up to take advantage of these benefits.

DOE Design of Experiments Terminologies to Know

DOE comprises certain terminologies to indicate various aspects of the process. Let’s look at some of the common ones.

Factor – A factor is an independent variable or parameter of the process. You can independently modify or set it to the required values without affecting other process variables.
Treatment – A factor occurring at a specified level is called a treatment. A set of such factors with their specified levels is a treatment combination.
Level – Level is the value or setting of the factor. They can be attributes or variable measurements, depending on the process.
Response – The output of a process is called a response. It is the outcome of all the actions of the factors. It may depend on various aspects of the process, such as productivity, performance, quality, and safety.
Effect – Effect is the quantification of the modification in the response due to a change in the factors.
Experimental run – An experimental run is the sequence of setting a process at a particular treatment combination, running the process, and gathering the response data for analysis.
Randomization – Randomization is the method of randomly running an experiment’s different steps. This is typically done to avoid errors that may be missed in the order given in the template and to minimize their effect on the variables outside of the experiment.
Interaction – Interaction deals with the variation in how the factors behave when they work together. It occurs when you change the levels of the factors, leading to a modified response that is different than what they would produce individually.
Blocking – When all the treatment combinations are run level-wise, the process is called blocking. After the treatment combinations at one level are completed, those at another level are then processed. However, randomization may be conducted within a level.
Replication – Replication is the method of conducting several experimental runs for every treatment combination. This is useful for determining the effects’ statistical significance by ascertaining the common cause.
Reflection – The set of treatment combinations run at levels that are opposite to those of the original set is called reflection.
Experiment – It’s the procedure that applies a treatment to the subject group in order to assess the response.
Experimental unit – The subject unit on which the experiment is conducted is called the experimental unit. It experiences the effect of the chosen factors and gives a response for further analysis.

Also Read: Why Choose Six Sigma Methodology for Project Management?

Example of Design of Experiments

To understand DOE better, let’s look at an example of a cosmetic company that wants to develop an organic nail polish that is on par with synthetic ones.

A typical DOE commences by deciding the factors, namely the constituents, their chemistry, non-hazardous nature, and longevity. This step determines the number of constituents in the nail polish.

The next step is to set levels for each factor. Here, the composition of each constituent in a nail polish sample, the properties to be evaluated, the experiments required to investigate the properties, and the benchmarks for each property form the core of this step. The chief properties of a good nail polish include adhesion to the nail, good gloss retention, easy application, quick drying, and long life.

During the experimental runs, the factors are manipulated to assess which constituent gives better adhesion, longer life, or better gloss. Here’s where DOE is the most useful. DOE helps design the experimental runs so that these properties will be investigated at a single level first. Once you know which factor fails to reach the benchmark, the factor is manipulated, and the experimental runs are conducted at that level.

All the experiments go through replication to confirm their favorable or unfavorable results. The data collected from the host of experiments is finally assessed for the statistical significance and comparison of the properties to the benchmarks.

How to Apply Design of Experiments?

Now that we have covered what DOE entails, here’s how to apply it.

Problem Statement

First, it’s crucial to identify the knowledge gaps, market demand, quality issues, and process bottlenecks. They will help to define exactly what problem needs to be addressed. You can also determine the objective, whether arriving at a solution, increasing the quality to a benchmark, or reducing the process time.

Planning

Once the problem is defined, you can begin planning for the experiment. This entails determining the factors to be varied, the expected responses, experimental procedures, and equipment to be used. If the number of factors is large, you may need to perform screening experiments to choose the most crucial ones. There are different design types that you can use to achieve this, such as fractional factorial design, Plackett-Burman design, and definitive screening design.

Experimentation

You can now conduct the experimental runs level-wise. You can also use randomization if the procedure allows for it. Once the data for a particular factor or level is collected, proceed to the next level of treatment combination. Replication of the experiments is important to confirm the statistical significance of the data.

Data Analysis

This step requires using the most suitable tools and processes for data analysis. Using a baseline result, you must ensure that any noise in the data is accounted for. The analysis should give values for the properties that can be compared to the benchmark.

Recommendation

Based on the results, you can recommend further steps of applying the results for process improvement, commercial release of the product, or quality enhancement.

Also Read: Quality Management Process: A Beginner’s Guide

DOE Best Practices

Design of Experiments is most useful for simplifying the experimental process while maintaining (or improving) the process and product quality. There are a few best practices that serve the purpose exceedingly well. Here are some of them:

Establish the objective of the DOE and the problem statement.
Determine the factors carefully based on the research question and existing data.
Ensure that the unimportant factors do not interfere with the results of the critical ones.
Take measures to capture data of the responses of the chosen factors only.
Select the levels in a realistic and practically applicable manner. Choosing higher or lower levels only because your equipment can measure them will waste time.
Establish logical reasoning for each step in DOE so that you are not left high and dry when questioned about the necessity of any factor, level, or experimental run.
Always conduct screening experiments to check for feasibility and potential issues.

Learn DOE and Other Six Sigma Basics

Six Sigma has several other vital components like DOE Design of Experiments. Any aspiring project or quality manager must be conversant with the fundamentals, essential tools, and their practical applications.

This comprehensive Lean Six Sigma Bootcamp enables you to master agile management, Lean Six Sigma Black Belt, Lean Six Sigma Green Belt, and tools such as Jira and Minitab. Besides offering classes led by industry experts, the course gives you hands-on experience with real-life projects.