A screening DOE will allow you to look for the relationship between your process variables and the process response as the values of your process variables change. This article will explain what a screening DOE is and how it contrasts with a full factorial DOE. We will also provide some tips on how to successfully run this kind of experiment.

## Overview: What is a screening DOE?

A screening DOE, also commonly referred to as a fractional factorial DOE, is one of the techniques used in the design of experiments. This technique is best used when you have a number of possible independent variables, which, if you used a standard DOE design, would require you to run a large number of experiments.

The purpose of the screening DOE, as the name implies, is to reduce the number of total runs of your experiment by screening out those variables that are not statistically significant. In addition to using a fractional factorial, you can also use a technique called the Plackett Burman Design. This also allows you to screen out variables that don’t have much of an effect on the response.

When you do a full factorial DOE, you test every possible unique combination of your independent variables and how they impact on your response variable. You also get the maximum amount of information about the process by looking at the main effects and all interactions.

By using a screening DOE, you lose some information, specifically regarding the interactions of your variables. That may or may not be important to you depending on the purpose of your experiment. We refer to that as losing resolution.

## 2 benefits and 1 drawback of a screening DOE

As with most things in life, there are some benefits and drawbacks of using a screening DOE. Let’s take a look at them.

### 1. Saves money

In most DOEs, the object produced by the experiment will need to be scrapped or discarded after the experiment. Since a screening DOE will utilize less runs, you won’t need to scrap as much product.

### 2. Fewer runs

By design, the screening DOE will require fewer runs to explore your process. Fewer runs means less scrap — and less time and cost to gather your needed information. The screening DOE is an efficient way to capture the information you need to make necessary decisions about the running of your process.

### 3. However, lost resolution

As we said earlier, if you run a full factorial, you get the main effects of each variable on the response, plus all the effects of the interactions. This provides a complete set of information about the process variables and process response.

Unfortunately, depending on the type of screening DOE you run, you will lose information about the interactions. This may be important, but since your main objective is to screen out and reduce the number of variables, you can make up for this reduced information by running a full factorial with the remaining reduced number of variables.

Screening DOEs are usually a resolution III, which only allows identification of main effects that are confounded with two-way interactions.

## Why is a screening DOE important to understand?

By itself, DOE is a somewhat complicated technique. It requires a lot of experience and planning to properly execute. A screening DOE will help you manage your experiment more efficiently.

### 1. Reduced effort to execute

If you must do a DOE, understanding how a screening DOE works will help you better manage the process and reduce the efforts needed to gather relevant information.

### 2. Importance of identifying variables

Doing some statistical work, such as regression analysis, prior to your screening DOE may allow you to reduce your selection of variables. Only those variables with a good chance of being significant should be explored first.

### 3. Reduced information

All screening DOEs, by their design, will provide less information about the process than a full factorial experiment. If you intend to do a screening DOE, be sure you have some insight into the possible existence and importance of interactions and resolution.

## An industry example of a screening DOE

While DOEs are usually thought of as only applying to a manufacturing environment, let’s look at how one can be used in marketing. A company was planning to launch a new product that they thought would become a market leader. Since they were still in the early stages of development, they wanted to be sure that their initial design and pricing assumptions were valid. One of the company’s MBBs was enlisted to help marketing decide on the final product design. The MBB thought that a screening DOE might be the way to go.

The marketing team was asked to develop a list of the variables that they thought might impact sales of the product. These were:

• Shape
• Graphics
• Price
• Name
• Shelf placement
• Media exposure

While the company could produce sample products for each combination, it would be a time-consuming and expensive process. For just a two-level experiment, the number of runs would be a 2^6 experiment, or 64 runs. But that would be for only one box of the product. To adequately gauge sales, the company would have to produce hundreds of units and test them in multiple locations. The MBB quickly concluded that wasn’t going to work.

Instead, the MBB suggested a screening DOE using a fractional factorial design of one-eighth, which is a resolution III design. They ran the screening DOE, and the company was able to come up with the best combination of the six variables. The product was a success.

If marketing had just used intuition to select the best combination, it’s not likely they would have come to the optimal solution that was achieved using the screening DOE.

## 3 best practices when thinking about doing a screening DOE

Doing any type of DOE can be complicated and fraught with danger. Keeping some of these tips in mind will help you avoid most of the common pitfalls.

### 1. Eliminate any noise

Any DOE is designed to explore the relationship between process variables and the response or output of the process. If other extraneous variables are allowed to interfere with the experiment, the results may become contaminated and meaningless.

### 2. What’s the purpose for doing a screening DOE?

Given the possible complexity and consumption of resources, you must define why you’re even doing a DOE. Be clear on your purpose and desired outcome before contemplating doing a DOE.

### 3. Understand interactions

You will be losing information (or resolution) when you do a screening DOE. You should have some understanding as to the importance of any higher-order interactions to determine if the loss of that information is critical or not to your final decisions as to which variables are statistically significant.

### What is resolution?

Resolution is a measure of the confounding that occurs when you do a screening DOE. In other words, you’re not able to distinguish whether the effect you see on a response variable is due to a single variable or the interaction of variables. Screening DOEs are usually resolution III designs intended to identify only main effects, not necessarily any interactions.

### What is the most common type of screening DOE?

A fractional factorial DOE is the most common type. In this type of design, you’re taking a fraction of the total runs you would do in a full factorial design. The intent is to screen out variables that don’t appear to have any significant effect on the response variable.

### When should I use a screening DOE?

Screening DOEs are used when the number of variables makes the experiment costly and time-consuming, making a full factorial design impractical.

## Revisiting a screening DOE

Design of experiments is a statistical technique that allows you to gain insight into your process by identifying which of your process variables have the most effect on your process output. If you’ve identified a large number of variables, the cost and resources needed to fully explore your process relationships may make a full DOE impractical.

The screening DOE allows you to reduce the number of viable process variables by running a subset or fraction of the number of runs that a full design would require. In doing so, you save time and resources.