Tuesday, November 18, 2008
Lengthy development processes not only reduce companies’ profitability, they deprive us access to life-saving cures and energy independence! Most R&D and Product Development processes do not utilize the many optimization methods that exist – methods that not only speed up the process drastically, but also result in optimal solutions (safe, reliable and robust products).
Typical Product Development efforts often include:
• Haphazard and wasteful “trial and error” approaches to evaluating design alternatives
• Inappropriate comparison methods (not statistically based) when evaluating alternatives—inadequate analysis
• A significant amount of guessing or “judgment” is used for decisions regarding materials, dimensions, processing steps, and other important product details
• Inadequate reliability testing is done to ensure products will perform over time in the conditions that products will encounter
• In some cases, products are over-designed (at a higher cost) due to the uncertainty from lack of knowledge
• Testing or experimentation which does NOT generate predictive models to identify optimal solutions
All of the above points contribute to lengthy and often unsuccessful development efforts. Additional outcomes often include products that are not reliable, competitive, desired, effective, or safe.
Optimal solutions (meeting all product requirements at the lowest overall cost) are rarely stumbled upon by chance—that is, trial and error. The fact that products must simultaneously satisfy multiple requirements (what good is a rechargeable battery that goes 500 miles between charges but occasionally catches fire?) makes finding optimal solutions even more difficult.
What Can Be Done
Fortunately, help is available. Academics have produced a plethora of mathematical and statistical methods to quickly evaluate an infinite set of combinations and find optimums. While you may be thinking of computer simulations that may or may not capture reality, we are not. Instead, we refer to acquiring actual data (physical data) from the judicious and strategic testing of products.
Most industry professionals are unaware of these methods. This is extremely disturbing—especially given challenges such as cancer, the quest for alternative energy sources, and the survival of some of our economically important industries.
Researchers may be highly trained in subjects such as chemistry, physics, biology, medicine, and engineering, but they are typically not armed with strong quantitative and optimization backgrounds. As a result, they do not apply advanced mathematical methods to aid them in their research. However, by pairing scientists and engineers with talented applied mathematicians, critical discoveries and solutions could be achieved quickly.
Design of Experiments (DOE)
Applied mathematicians with specialties in optimization can significantly reduce product development time via numerous quantitative techniques. One of the many useful techniques is Design of Experiments (often referred to as “DOE”).
Unfortunately, most DOE approaches (and efforts) are unsuccessful, and most do not produce effective predictive models. However, when applied correctly, DOE results in powerful predictive models that predict product performance as a function of relevant design choices. The predictive models also empower researchers to determine how changes in inputs (e.g. specifications) impact product performance.
A critical feature of these models is that they describe the effect of interactions between various inputs (where the effect of one factor depends on yet another factor). The functional forms can be non-linear—which is important since much physical behavior is not linear. Ultimately, mathematical models can be used to identify optimal solutions, and these may be validated with strategic follow-up testing.
Future blogs will delve into the use (and misuse) of DOE in practice as well as other techniques for accelerating product development to produce superior products and new technologies. Pharmaceutical R&D (and other formulation problems) have unique challenges—and while traditional DOE does not apply (due to many constraints on formulation problems), Mixture Experiments definitely apply—and offer tremendous benefits.
The question is whether scientists and engineers will welcome mathematicians in their quest for crucial innovations. Einstein did…but will today’s researchers?
By: Steven Wachs & Allise Wachs, http://www.integral-concepts.com/
Friday, October 31, 2008
Let’s begin with a question. “How did a small, bombed-out nation with few natural resources become one of the greatest industrial powerhouses?” In answering this question, fundamental reasons for our industrial demise will become clear—but so will solutions.
There was a time when the United States produced the most highly demanded automobiles, electronics, appliances, furniture, textiles, and consumer products. However, over the past decade, the United States has become notorious for exporting jobs and factories.
Consider the automotive industry. The steady downward trend began in the 1970s. In that era, the top 3 US automakers held 90% of the U.S. market share. Today, the top 3 US automakers have only 40% of the U.S. market for passenger cars. They have 50% of the entire U.S. market which includes trucks.
During the same time period, the Japanese automotive market (in the U.S.) share grew steadily from roughly 1% to 38%. So now, nearly 4 in 10 vehicles sold in the U.S. are Japanese.
Unemployment rates in Michigan (the state most affiliated with automotive jobs) have more than doubled from 3.3% in January 2000 to almost 9% now. Since 2000, the state has lost over 400,000 jobs—mostly from the auto industry.
Similar trends exist in electronics. Japanese brands like Sony, Sanyo, Sharp, Mitsubishi, Nintendo, Panasonic, Toshiba, Canon, Epson, and many more now dominate in electronics manufacturing.
Consider a statistic from last year’s import and export data collected by the U.S. Census Bureau. In 2007, the U.S. exported 136.4 Billion dollars worth of consumer goods category. However, that same year, the U.S. imported 465.7 Billion dollars worth of consumer goods. Despite the weak U.S. dollar, we are importing nearly 3.5 times more than we are exporting in the consumer goods category.
Once we acknowledge that we have lost our position as the greatest producer of highly demanded products in the marketplace, we can look for ways to regain our preeminence.
After World War II, Japan’s major cities, industries, and transportation networks had been destroyed and it faced a serious food shortage.
Despite Japan’s size and deficiencies, it became the world’s second largest economy when measured by gross domestic product. So, how did Japan rise from a state of devastation and poverty to become a respected manufacturer of automobiles, electronics and high-tech equipment?
The answer is deceptively simple. Some of the Japanese companies learned four of the most important keys to manufacturing excellence:
1. Manufacture products with as little variation as possible—that is, make components as similar to each other as possible (Minimize Variation)
2. Design/Engineer parts so that they will work in all realistic environments and for a long period of time (High Reliability)
3. Prevent and Predict issues—rather than Detect and React
4. Educate Employees in statistical methods to accomplish the first 3 items
One of the most famous consultants responsible for the success of many Japanese manufacturers was William Edwards Deming. He was an American with a Ph.D. in mathematical physics, but his passion was industrial statistical methods.
He helped manufacturers predict system performance and minimize variation and inefficiencies via the use of industrial statistical methods. From Deming, several Japanese companies quickly learned how to quantify variation, understand its grave ramifications, and to ultimately minimize it. Most American companies were uninterested in the topic of variation.
Let’s digress shortly to discuss variation—and why it destroys a manufacturer’s ability to compete. Suppose we are making caps that need to fit on a particular type of soft drink bottle. The first thing to understand is that no two items are identical. So, no two caps are identical. Because they are not identical, we say there is “variation.”
Suppose we have a bottle, and we try to fit one of our caps on it. It will fit a certain way….it will take a certain torque to remove the cap, and the contents may or may not leak. However, if we put a different cap on the bottle, it will not possess the exact same fit. It may take a slightly different amount of torque to remove it, and it may leak at a different rate. Whether or not the two caps behave differently in a practical way, we do not know, but they do fit differently.
The variation in product components (and how they fit together) largely explains why products fail at different times. For example, two of the same model washing machines may fail at drastically different times in service. The impact of variation is often noticed immediately in assembly operations, but when it’s not observed during assembly, it will be noticed by customers. Variation is measured by a statistic called a “standard deviation.” It should be one of the most important measures to manufacturers.
Products fail at different times. You’ll hear someone rave about their model X vehicle—while someone else curses it. The data that supports largely different failure times resides in warranty databases, customer satisfaction datasets, recall databases, and vehicle registration databases.
While some manufacturers have historically tried to reduce variation, others have not. Consider some automotive data. It is estimated that GM spent approximately $4.5 Billion in warranty costs last year, and Ford spent roughly $3.8 Billion in warranty costs last year. The reason for vehicle failing during warranty periods is product variation—since many vehicles do not fail during that time. If all products were identical, they would fail at the same time (approximately).
From 2003 to 2006, DaimlerChrysler spent an average of approximately 4.5% of revenue on warranty. GM spent roughly 3% of revenue on warranty. Toyota spent only 1.25% of its revenue on warranty.
Then, there is safety recall data (see www.nhtsa.gov). Although it’s not a completely fair comparison because production units and severity of the recall have not been adjusted, it is still interesting to note that from 2000 – 2006, GM had 1,014 safety-related recalls, Ford had 558, DaimlerChrysler had 374, Honda had 165, and Toyota had 131—despite the fact that Toyota surpasses both Ford and Chrysler in U.S. sales volumes.
Perhaps the biggest reason for the U.S.’s decline of the automotive and electronics industry is what happens after warranty periods. When consumers find themselves forced to pay for costly repairs after the warranty ends, many lose loyalty to the brand. It is difficult to quantify the actual lifetimes of automotive and electronics products, but if used sales are any indication of the long-term performance of products, an interesting picture emerges.
According to Edmunds & KBB, all 10 of the Top 10 Resale Value Vehicles are foreign. 8 of those 10 are Japanese.
Profit Losses from Product Variation
Because of excessive variation, huge investments are made in inspection processes and equipment. Since most manufacturers neither predict nor prevent the production of bad product, they purchase expensive (and not always reliable) systems to detect bad product. Again, the impact on profitability can be staggering. Deming’s goal was to eliminate the need for inspection by understanding and predicting manufacturing process behavior.
Huge costs that we don’t hear about in the news are losses from launching new products. The problems and expenses associated with introducing new products into the market are difficult to quantify, because the information is not publicly available, but consumers hear of numerous examples daily. They include potentially unsafe drugs, unsafe consumer goods, unsafe vehicles, and so on. They also include delays in promised arrival dates of new products.
It’s not clear whether our industrial executives disregard the statistics or simply fail to comprehend them. It is disconcerting to continually hear U.S. executives (and the media) claim that quality and reliability of U.S. products has essentially caught up with the Japanese. They argue that there is only a “perception problem.” However, the data strongly suggests otherwise. These executives only need to look at statistics kept by organizations like Consumer Reports, the Consumer Product Safety Commission, the National Highway Traffic Safety Administration, the Federal Trade Commission, the National Transportation Safety Board, their own warranty data, and customer complaint data. They should also consider the lawsuits against their products, and the scrap and rework within their plants that obliterate their profits.
Quality & Productivity Improvement Programs
While some of these corporate programs have returned some improvements, they clearly have not made the vital difference we need. The economic statistics indicate a continuation in the decline of our manufacturing base.
The dire consequences of a diminished manufacturing base include dwindling consumer purchasing power, limited opportunities for our children, less tax dollars, higher poverty levels, and many other dismal results.
What Can We Do?
So, what can the United States do? We can:
1. Use proper industrial statistical methods to understand and minimize variation in components and products.
2. Develop products with high long-term reliability—not just acceptable initial quality; this can only happen with the use of PROPER statistical and reliability methods.
3. Prevent manufacturing problems through the use of PROPER statistical process control (SPC).
4. Educate employees in necessary quantitative methods for superior engineering and statistical prediction of process and product behavior.
This may be a difficult pill for our society to swallow. There are several reasons:
1. There are very few highly educated/credentialed statisticians working in manufacturing.
2. Most engineering programs require little or no training in statistical methods.
3. Most corporate trainers and consultants of statistical methods possess no formal education in statistics.
4. Many professionals believe that 4 weeks of statistical training makes one an “expert.” Yet, one might be hesitant to go to a physician, attorney, or any professional who has only received 4 weeks of training.
5. Methods like Designed Experimentation, Reliability Analysis, and Statistical Process Control are either misapplied or not used at all. Hence, manufacturers have gained nothing in terms of predictive ability and prevention of problems.
6. Our best-educated statisticians are working in areas that appreciate statistical reasoning: Insurance, Medical Research, Marketing, Academia, and “Think Tanks.” The manufacturing sector does not appreciate the value of industrial statistics.
Finally, we have an education crisis—especially in mathematics, statistics, engineering, and the physical sciences. Consider the following data from http://www.nces.ed.gov/:
1. Our 15 year-olds rank 25th worst in standardized mathematics testing among 30 participating countries.
2. Our 15 year-olds ranked 21st worst in standardized science testing among 30 participating countries.
3. Of the masters degrees we issue in science and engineering, 40% go to foreign nationals.
4. Of the Ph.D. degrees we issue in science and engineering, over 60% go to foreign nationals.
5. The number of American 18 – 24 year-olds who receive scientific degrees has fallen to 17th in the world. We were 3rd in the 1970s.
Most U.S. companies do not teach proper industrial statistical methods to their employees. Nor do they attract highly competent industrial statisticians. Instead, they offer “easy-to-digest” quality programs taught by non-statisticians, like the currently popular “Six Sigma” programs and Taguchi programs.
If these programs were so effective, why hasn’t our manufacturing base been able to improve its warranty situation? Why haven’t they improved profitability, market share, and their ability to compete? Why do our manufacturing imports continue to rise in spite of a weak dollar? Why haven’t product safety-recalls declined? Why do we continue to send our jobs and plants overseas?
Early statistics seem to indicate that Japanese products and performance is now declining due to more variation in their products. The Chinese manufacturers are the new threat. The time is ripe for U.S. manufacturers to heed the advice of Dr. W. Edwards Deming. Our economic future depends on it.
By: Allise Wachs, Integral Concepts, http://www.integral-concepts.com/