Locked lesson.
About this lesson
Design of Experiments (DOE) is an experimental technique for identifying the primary factors within a system that determine system performance. DOE is particularly useful in complex systems where there are interactions between factors and relationships are not obvious.
Exercise files
Download this lesson’s related exercise files.
Intro to Design of Experiments - Exercise.docx61.3 KB Intro to Design of Experiments - Solution.docx
61.2 KB
Quick reference
Intro to Design of Experiments
Design of Experiments (DOE) is an experimental technique for identifying the primary factors within a system that determine system performance. DOE is particularly useful in complex systems where there are interactions between factors and relationships are not obvious.
When to use
DOE is normally used in the Improve stage of a Lean Six Sigma project if the solution requires a common cause process change. The DOE identifies which factors must be controlled and at what level they must be controlled to have a product or process that meets desired performance targets.
Instructions
Experimental design approaches are inquiry or discovery approaches and therefore are used to discover new knowledge. If a product or process is well-known and characterized, experimental design is not needed to find and fix problems. However, if the goal is to create a new system or to extend the existing system into a region for which it was not originally designed, experimental design approaches are applicable. The experimental design approach recommended for Lean Six Sigma projects is the Design of Experiments because of its statistical underpinnings.
DOE uses a series of controlled tests of various combinations of controllable factors, uncontrolled variable factors (such as environmental conditions), and uncontrolled discrete factors (such as organizational or system constraints and boundaries) to determine the relationship between the factors and output performance characteristics. This relationship leads to a mathematical model or formula for system performance that can be used to design and optimize system performance.
On rare occasions, DOE is used to discover relationships among factors that lead to the problem being investigated. This is usually done as a last resort since other analysis methods are much faster and easier. It is more common to use DOE during the Improve phase to determine how to modify an existing system or the characteristics to be pursued in the design of a new system or process. Once the model is developed, it can be used to set target and tolerance levels on inputs and to explain realistic expectations on the variation that will be experienced in the new improved design solution. It is much more likely to be used when the problem is of a common cause nature requiring a system change. If the problem is special cause, there is normally a clear root cause, and the time and expense of a DOE are not needed to determine what needs to be changed.
Hints & tips
- DOE is a very powerful technique, but it is not cheap. It will often require dozens of tests to be conducted under very controlled circumstances.
- Most statistical analysis software applications, such as Minitab, have DOE matrices and statistical algorithms programmed and setup. Just follow their wizard to construct the study and analyze the data.
- In my experience, a DOE would be helpful on about 25% of Lean Six Sigma projects.
- 00:05 Hello, I'm Ray Sheen.
- 00:06 Design of Experiments is a very powerful technique for
- 00:10 determining the design of a new system or process.
- 00:13 We have an entire course on this topic, but
- 00:16 I want to take a few minutes now to introduce DOE to you.
- 00:21 Lemme start with the whole concept of experimental design as compared to
- 00:24 theoretical design.
- 00:26 The design process is an inquiry process.
- 00:28 Its goal is to discover a new or better mousetrap.
- 00:31 It's about creating a positive change for the customer, either with a new or
- 00:36 improved product or service.
- 00:38 Experimental design does that inquiry through a set of experiments.
- 00:42 It tries something, and based upon the result, it tries something else.
- 00:46 This may continue for numerous tests or
- 00:48 experiments until a design is deemed acceptable.
- 00:52 The goal of experimental design is to gather enough actual data that a model or
- 00:56 formula can be created that explains how things work.
- 01:00 This is in contrast to theoretical design that starts with the formula.
- 01:04 It creates a theoretical design concept and
- 01:06 then builds it to confirm that the theoretical performance can be achieved.
- 01:10 One of the outcomes of experimental design is the Y = FX equation that
- 01:15 we've discussed earlier.
- 01:17 Now, there are a lot of different systems and methods for
- 01:20 doing experimental design work.
- 01:22 But they seem to fall into two categories, organized inquiry and
- 01:26 disorganized inquiry.
- 01:27 Most designers use disorganized inquiry.
- 01:30 They eat too much and get indigestion and that keeps them awake all night.
- 01:34 And sometime during the night as they're thinking about the most recent
- 01:38 Star Wars movie they get an idea of something that they want to try and
- 01:41 next day they set about trying it.
- 01:43 Well, DOE does not rely on midnight inspirations, it is organized inquiry.
- 01:49 So let's look at the theory behind design of experiments, or
- 01:53 as it's commonly called, DOE.
- 01:55 The DOE approach is to first quantify the relationships between factors of An output
- 02:00 performance for the product or system.
- 02:02 It analyzes the different combinations using statistical techniques to quantify
- 02:06 the relationships.
- 02:08 Once those relationships have been determined,
- 02:10 the design can then be optimized for the key performance metrics.
- 02:13 Since we have an entire course on DOE, this module will just discuss the concept.
- 02:18 To create this model, the various process or product factors and
- 02:21 the characteristics are divided into several subsets.
- 02:24 One of these is controllable factors, these will be studied to provide
- 02:28 insight into how to use them to manage process performance.
- 02:32 Another category is variable uncontrolled factors, these are often environmental
- 02:37 factors, like temperature, humidity, time of day, or common cause process variation.
- 02:43 The third category is discrete uncontrolled factors, many times
- 02:46 the analysis will be to determine if these discrete factors should be controlled or
- 02:51 uncontrolled.
- 02:52 These factors could be, which suppliers material are used, or
- 02:56 which location does the processing.
- 02:58 These are often determined by system boundaries or constraints.
- 03:02 The fourth category is the set of system performance or response parameters, for
- 03:06 DOE to work, there must be at least one measurable parameter.
- 03:10 It's even better if that response is variable, but
- 03:13 it can work with a discrete factor.
- 03:15 We can actually use DOE in two modes.
- 03:17 One is to help further isolate a problem in problem analysis mode, and
- 03:21 the other is a design mode for a new system or process.
- 03:25 But let's look at the problem solving one first.
- 03:27 If one of the output or response factors is the problem condition,
- 03:31 DOE can tell us which factors or combinations of controlled and
- 03:35 uncontrolled factors will create the problem condition.
- 03:38 In this case, we can use DOE to establish expectations about process capability.
- 03:44 Which you should recall, is the ability of the process to reliably deliver outputs
- 03:48 that are within the customer's specifications.
- 03:51 In other words, DOE will tell us if the process will always be creating some
- 03:55 results that are out of tolerance.
- 03:57 That will set us up for a discussion with stakeholders about realistic expectations.
- 04:02 The DOE can also be used to show which factors have a positive, negative or
- 04:07 nil effect on the problems condition.
- 04:10 In addition to using DOE for analysis,
- 04:12 DOE can be used to create a new design to create a new normal as part of
- 04:16 the improved stage in your Lean Six Sigma project.
- 04:19 Remember we said,
- 04:20 the DOE creates the model describing how all the factors both controllable and
- 04:24 uncontrollable work together to generate the product or system response.
- 04:29 It explains the behaviors and allows us to predict the performance for
- 04:33 different sets of input factors.
- 04:35 A great aspect of DOE is that most of the DOE analysis
- 04:38 methods will analyze both primary effects and combination effects.
- 04:42 When working with complex systems, many times the most important factors are based
- 04:47 upon the interplay of two or three factors rather than on a single one.
- 04:51 So through the use of DOE, we can design a system and
- 04:54 determine the best set of controls to drive system performance.
- 04:58 We can also determine the allowable tolerance for
- 05:00 those controls that will ensure that we get acceptable process performance.
- 05:04 And we can tell the stakeholders what are the realistic expectations on
- 05:09 performance based upon the factors that they want to control.
- 05:13 This analysis can be used to illustrate the effect of common cause variation and
- 05:17 what it would take to change that variation.
- 05:19 By understanding the factors and how they contribute,
- 05:22 the team can explain to stakeholders what is needed to resolve the problem.
- 05:26 They can create several solutions with different control factors and tolerances.
- 05:30 Stakeholders can then evaluate the costs and benefits of each.
- 05:35 DOE will help you to understand and get control of factors contributing to
- 05:40 common cause variation that is part of your problem.
- 05:43 Now, it's a powerful technique, but it's not needed on every project.
Lesson notes are only available for subscribers.
PMI, PMP, CAPM and PMBOK are registered marks of the Project Management Institute, Inc.