Asking the Right Question 2

Asking the right question is essential regardless of how you perform experiments. If you don't ask the right question, your answer will have little or no value. Especially in these economic times, companies cannot afford to waste time and money answering the wrong question.

So why aren't researchers careful to ask the right question? It's very hard work. It is the hardest part of any project and it can be very time consuming. When you are pressured for time, it is very easy to skip this step.

Here are some ideas to make the process more efficient and less demanding:

1. Know the voice of the customer. (I discussed this in the last blog.) Satisfying your customer is the only real measure of success.

2. Know what you will measure and how you will measure it. You frequently need a substitute quality characteristic because the customer's desires are not easy to measure. You will need to use all your engineering and scientific knowledge to specify these substitute quality characteristics. For example, your customer wants high quality photographs. You have to determine what this means before you can even begin to specify properties to measure. If it means crisp edges with no blurring you will measure one property. If it means bright colors, you will measure another property. If it means both, you need to measure properties relating to both.

3. Be sure your measurement system is adequate for the job. It is a huge waste of time to collect data so noisy that they tell you nothing.

4. Pick your factors carefully. It is frustrating to pick "factors" which don't influence your responses. Every "trend" you find is coincidental and not repeatable. Keep an open mind when choosing factors -- let data drive your decisions, not beliefs.

5. Anticipate anything you are doing that colleagues or management may disagree with. It is best to reach agreement before data are collected.

6. Check every assumption you are making. Assumptions can very easily lead you to answer the wrong question.

7. Get input from everyone you can. This includes buyers, production workers, marketing -- anyone who might have some insight. I once had a buyer warn me that the ingredient I wanted to use in an ink was single sourced and put the company at risk. Once I knew this, it was easy to find an alternative that was multi-sourced, but without that input my solution wouldn't have satisfied my company's needs.

If you are looking for a good New Year's Resolution, consider resolving to ask the right question every time. The job you save may be your own.

Happy New Year!

Bill Kappele.

Asking the Right Question

Failing to ask the right question leads to waste. Answering a wrong question wastes your time and your company's resources. No matter how you perform experiments, it is imperative that you ask (and answer) the right question.

The key to asking the right question is understanding what your customer wants and needs. Your customer is whoever is willing to pay you for your work. It may be a consumer, an employer, or a funding agency. Whenever anyone pays you, he has certain expectations of you. If you fail to meet those expectations, you will eventually lose your job.

If a co-worker asks you to do some work for him you may be able to treat him as a customer. If your employer agrees that this work needs to be done, you can treat him as a customer. If not, you should spend your time on other work.

The book, "Creating a Customer-Centered Culture," by Robin L. Lawton, is an excellent resource for learning to ask the right question. It covers everything from identifying customers (and understanding the type of customer) to measuring how well you have satisfied your customer. If you want the details behind Step 1: Asking the Right Question, then this is the book for you.

Enjoy reading it!

Bill Kappele.

Seven Steps in Design of Experiments

Objective Design of Experiments teaches a seven step approach. These steps are:

1. Ask the right question
2. Choose a model
3. Choose an experiment design
4. Collect data
5. Analyze your data
6. Test your model and your Sweet Spot
7. Take pride in a job well done!

In our workshops you learn to apply these steps in a series of exercises. With onlya few days to learn this, there isn't time to delve nto the details behind the steps.

In the next several blogs I hope to discuss these details. If you have specific questions about any of the steps, please let me know so I can address them.

Next time I will start with Step1.

Bill.

Gosset for Windows

Yes, you can run Gosset on a Windows computer.

Thanks to the free software, Cygwin, Gosset can run on a Windows computer.

You can get everything you need to run Gosset on your computer with Windows XP or Vista by calling (866) 683-6173 and requesting the CD for Gosset for Windows. When you have your CD, watch our video to learn how to install it. Beware -- installation is not for the feint-hearted! You may want help from your IT department.

Good luck!

Bill Kappele.

A Design of Experiments case study using the Objective Design of Experiments 7 Steps is featured in Quality Digest this month.

US Synthetic, a manufacturer of synthetic diamond compacts for the oil drilling industry, used the 7 Step process to improve the quality of one of their excellent products.

You can read their story here.

One of the most common questions I have been asked is, "What sample size do I need for ... ." The "for ..." can mean "confidence limits of a certain width,"" or 99% reliability with 95% confidence," or "to run a designed experiment with 5 factors and a full quadratic model."

To answer the most common of these "for ..."s I wrote the Sample Size Calculator. You can download a 14 day trial copy from
http://www.objectivedoe.com/software/SampleSize.exe

Here's a complete list of what it includes:

1. Confidence limits


2. t-Tests


3. Cpk


4. Percentages


5. Minimum experiment trials (DoE)


6. Reliability


7. Tolerance limits

If you like it and want to keep using it, you can make a donation of $20 or more to St.Jude Children's Research Hospital at
https://shop.stjude.org/GiftCatalog/express-donation.do?fnl=don_sin

Forward your e-mail receipt to Hero@ObjectiveDOE.com and you will receive
a registration code in a few days.

This is a great way to help children suffering from cancer and calculate sample sizes for your work. You may even be able to claim a tax deduction!

You can learn more at
http://www.objectivedoe.com/software/software.html

Take care,

Bill Kappele.

Taguchi, Factorial, RSM, ... -- Which DOE is Best?

Design of Experiments (DOE) is an elegant technique for efficient, thorough, objective experimentation. It has been around since the 1920's and has been used successfully in industry since the 1950's. This technique is so fundamental that it has been used in numerous experimental strategies -- often with the goals of the particular strategy being mistaken for DOE itself. Design of Experiments includes these features wherever it is used:

1. A mathematical model is chosen to describe the situation under
study.
2. An experiment design is created to collect data in the best way
to fit these data to this model.
3. The data are collected and analyzed using the model initially
chosen.

As an analogy, think of the "model" as a "recipe." You use this to create a "design" or "shopping list." Without the recipe, you can only guess what ingredients you may need.

Without a model, you can only guess which experiments to run. "Analyzing" the data is like putting away the groceries in an organized manner.

So why does DOE have so many names?

Actually, it doesn't -- it just appears to. The names come from various strategies that use DOE. Here is a list of some of the most popular experimental strategies and their goals:
1. Taguchi is interested in finding a "robust" answer to the
experimental question. It seeks an answer that is insensitive
to factor variations and noise. It doesn't predict the best
combination of factors to achieve your goals.
2. "Factorial", or "Classical DOE," was the first technique used
with designed experiments. It allows you to see which factors
are most important and helps you to identify important
interactions among the factors. It doesn't predict the best
factor levels to meet your goals.
3. Response Surface Methodology (RSM) uses your model to
make contour plots of predicted behavior. Using these plots
you can actually predict the best combination of factors to
meet your goals.

All of these strategies use DOE. The way DOE is applied will differ depending on the goals of the strategy, but the DOE technique does not change.

What do you think?

Bill Kappele
Technical Director
Objective Design of Experiments
Bill@ObjectiveDOE.com