In today’s complicated world, the quality professional is faced with an increasing array of differing methods to solve problems and improve processes. Advertisements shout to the prospective buyer that this software will make their dreams come true – guaranteed! Best selling books are written that extol the virtues of special techniques and then proceed to take you down such a complicated path that it’s difficult to perceive if you’re going in the right direction. Many improvement programs within organizations fail just because they are too complicated and no one quite knows how to proceed or what the results mean at the end.
Life doesn’t have to be full of complications. If the full blown version of an improvement program is introduced and not understood, the result may well be that no improvements happen, yet employee’s time is taken up because they still think something is supposed to happen. The improvement teams stay together and have “sometimes meetings” because they don’t have the permission of management to quit, but are able to throw attention off of their non-results by identifying progress as “ongoing”. This is not a good situation for anyone. The management team is fooling itself by not monitoring the team progress (or lack of) properly and not identifying that a possible time and money pit has been created. The employees feel inadequate because they can’t come up with a success and are keeping out of trouble by standing still. Don’t underestimate the power of these complete programs. If implemented properly with appropriate support and training, they can offer a tremendous amount of payback over time. The trouble is that many improvement techniques are introduced without the high-end support structure that is needed, and just end up sucking resources from the company and frustrating everyone involved.
Make Life Easier
Whenever you feel yourself deluged by the complications of introducing a particular improvement subject, sit back for a minute and ask yourself several questions.
What is my objective?
Write down what your problem is and the desired result you would like to see.
What do I think will fix it?
Write down the technique(s) you feel will be appropriate to cause the desired result.
What are the basic concepts of that technique?
Break up the activities within the technique that you will be using into critical steps. Recognize if you really need software, massive calculations, weeks of training, and 10 or more planning sessions to get you started. Or can you pick something simpler that at least will give you a partial success, with minimum work, in a short period of time. Remember that a small success is better that no success!
Design of Experiment
A good example of a complicated improvement technique that can be downsized, broken into critical components, and made available even to the line operator, is Design of Experiments (DOE). DOE’s come in many formats and gained a great height of awareness with the introduction of the Taguchi method in the latter part of the 20th century. DOE is used when other methods to improve processes such as Statistical Process Control (SPC) have been exhausted. DOE’s allow a team to see how the different parameters of a process affect a given result that they would like to improve (ex: defects, inefficiencies). These experiments also allow the team to see if interactions exist between the different critical parameters of the process such as temperature, speed, pressure, etc.
Genichi Taguchi was an engineer and statistician who developed a methodology for applying statistics to improve the quality of manufactured goods. Although experimentation to establish how different process parameters affected each other and come up with a holistic combination was not new, creating special production runs to involve all possible combinations was time consuming and expensive. Taguchi made DOE’s simpler by introducing a new methodology and philosophy, including the use of a Loss Function to decide when a process needed to be optimized. He advised engineers to look only at the critical process parameters rather than all of them, and to not look at interactions at all unless they felt compelled to. This enabled organizations to reduce the amount of experiment runs they had to make from thousands down to less than twenty with a comparable result. Even Taguchi’s simplified methods, however, were interspersed with complicated mathematical formulas and functions such that would scare off the layman technician and floor operators. Many DOE’s were created within organizations that had erroneous results that couldn’t be explained because the coordinator didn’t understand the critical concepts behind the DOE techniques.
DOE Made Easy
Go back to the basics. With Design of Experiments we can further the philosophy of Taguchi by making it even easier to understand how to make a successful designed experiment that everyone can understand – without complicated mathematics and expensive software. We can do this by defining a DOE as a series of critical steps in a process, with the end result being an improvement in the “Effect” being studied..
Step 1 – Decide Your “Effect”
The effect is the item that you want to improve. In most quality environments it will probably be a defect rate or capability of a certain characteristic such as electrical test fallout for a circuit board. You may have a 7% defect rate that you want to reduce by optimizing the process. In this case, the effect is merely “electrical test”.
Step 2 – Gather Your Team
Teams do better than individuals. This is a known and researched fact. You will need representatives from different areas that know about the process in question. Ideal teams are from 3 to 6. Example positions of team members would be engineers, managers, technicians, inspectors, supervisors, operators, and maintenance personnel.
Step 3 – Train Your Team
If you want your team to help you to their fullest ability, then at least give them some training on the subject that you want them involved in. This will enable the team members to identify obstacles and solutions each step of the way during the DOE process.
Step 4 – Brainstorm Factors Using Cause and Effect Analysis
Using your identified “Effect” as the focal point, have the team brainstorm all of the possible “Factors” that may cause variation in the effect. Remember that you are not yet interested to hear if it will improve or degrade the effect, only if it may vary the end results. ALL ideas from the team go up on a Cause & Effect diagram (sometimes called an Ishikawa diagram or Fish Bone diagram), even if they are made in jest. No one should be allowed to dominate the meeting, or debate another’s idea as they are put up on the board. After all of the possible factors are identified, the board is left up for 48 hrs. in an area where the team can re-visit and add ideas previously missed. After 48 hrs. re-assemble the team and agree on which of these factors they feel are critical to the final result of the effect. At this time, the team should also decide whether they feel strongly about potential interactions between certain factors.
Step 5 – Pick an Orthogonal Array
After the brainstorming session, the team should have come up with a certain number of factors and/or interactions that they feel are critical, and whether they would like to test each of the factors at 2 different settings, or 3. Remember that the more settings that you want to test the factors on, the higher the amount of experiments you will have to conduct. They would then pick a structured mathematical model called an orthogonal array that would best fit the number of factors and/or interactions that they have decided upon. The orthogonal array tells you how many experiments that you have to perform and at what levels each parameter has to be for each experiment. It is your recipe.
Step 6 – Picking a Linear Graph
The linear graph is also a model and you can pick one that fits with your orthogonal array. It helps to decide the placement of factors and/or interactions within your orthogonal array so that the final experiment results will make sense to you. It also helps to make your experiments run as smooth and efficiently as possible. Both the orthogonal array and linear graph are very visual tools and easy to use.
Step 7 – Run the Experiments
Now you’re all set. Create the sample runs of product according to the orthogonal array and record your results. Make sure that the samples don’t get mixed up, record any possible process changes during the sample runs (change in materials, operator, etc.), and record your results. The hard part is over.
Step 8 – Graph Your Results
Compare the average results between levels of the factors and/or interactions to see which level for each factor makes a more desirable outcome. Plot the comparisons on a hand made graph and pick the optimum level for each factor and/or interaction. The potential optimum level for the process is the collection of optimum levels for each of the factors. The interactions are only studied for further information about how the process works.
Step 9 – Verify Your Optimum Level
Remember that the optimum level that you picked is still only a potential one. Make sure that you verify by further sample runs until you feel 200% confident. Check to make sure that your changes don’t cause any other undesirable side effects to happen. Then count the amount of dollars you’ve saved your organization this year.
Congratulations. You’ve just had a success. Don’t forget to celebrate.
Want to optimize YOUR process?
Contact us to arrange for a HUMAN onsite interactive DOE Made Easy workshop. Train your process improvement team together as they have fun completing a short designed experiment together in the classroom, and end the workshop by planning their actual DOE. Click here for more details.