How to Perform a Design of Experiments

Jay Arthur
888-468-1537
303-756-9144
KnowWare International, Inc.
DBA LifeStar
2253 S. Oneida
Ste 3D
Denver, CO 80224

We work with companies
that want to fire up their profits using
Lean Six Sigma

Copyright © 2011

Design of Experiments Example

For simplicity, let's assume you are writing a cookbook and want to find the best directions for baking a cake (which is similar to baking paint on a car finish). You can save time by performing a design of experiments test.

First, determine the "factors" you want to test and establish the high-low settings for each factor in your study. Let's suppose you have four factors (a four factor experiment):

1. Pan shape: Round (low) vs square (high) pan
2. Ingredients: 2 vs 3 cups of flour
3. Oven temperature: 325 vs 375 degrees
4. Cooking Time: 30 vs 45 minutes
   

Let's say that you'll rank each resulting cake on a 1-10 scale for overall quality.

You then use the +/- values in the orthogonal array to perform a test of every combination (16 total):

• High: all high values (+ + + + = square pan, 3 cups, 375 degrees, 45 minutes)
• Low: all low values (- - - - = round pan, 2 cups, 325 degrees, 30 minutes)
• In Between: every other combination ("+ + + -", "+ + - -", and so on).

To optimize your results, you might want to run more than one test of each combination. Then you just plug your data into a DOE template (Taguchi or Plackett-Burman format) like the one in the QI Macros and observe the interactions.

Here is a sample of the QI Macros L8 Taguchi Template. Input areas are shaded yellow for easy identification. The four factors are input in rows 3 to 6. The red outline shows the combination of factors that are used for the 16 trials. The results of the three separate tests of each trial (rating cake quality on a scale of 1 to 10) are input in columns J, K and L in the responses section.

Design of Experiments Service Example:
You can even perform a design of experiments test in the service industries.

People who send direct mail rigorously tally their results from each mailing. They will test one headline against another headline, one sales proposition against another, or one list of prospects against another list, but they usually only do one test at a time. What if you can't wait? By performing a Design of Experiments, you could test all of these factors simultaneously. Design your experiment as follows:

1. Headline: Headline #1 (high), Headline #2 (low)
2. Sales proposition: Benefit #1 (high), Benefit #2 (low)
3. List: List #1 (high), List #2 (low)
4. Guarantee: Unconditional (high), 90 days (low)

This way you might find that headline #1 works best for list #2 and vice versa. You might find that one headline works best with one benefit.

Performing a Design of Experiments can help you shorten the time and effort required to discover the optimal conditions to produce Six Sigma quality in your delivered product or service. Don't let the +/- arrays baffle you. Just pick 2, 3, or 4 factors, pick sensible high/low values, and design a set of experiments to determine which factors and settings give the best results.

Start with a 2-factor and work your way up. Have fun! It's just not that hard, especially with the right software.

Design of Experiments

DOE Software

QI Macros Design of Experiments DOE Video

Design of Experiments templates for Taguchi 4, 8 and 16 factors and Plackett-Burman are included in the QI Macros for Excel SPC Software.

Download the FREE 30-day Evaluation copy of the QI Macros Excel SPC Software for Six Sigma

Try QI Macros
FREE
For 30-Days

Or Buy It Now!