Tenfold Improvement

Performance Improvement H. James Harrington
“Six Sigma began with upper management’s demand for radical change”

In 1979, recalled John A. Young, former president of Hewlett-Packard Corp., “I launched a new quality campaign by announcing what I call a ‘stretch objective.’ I asked our product failure rates to be cut to one-tenth their current levels by the end of the 1980s. Why ask for a factor-of-10 improvement number? If I called for an improvement of only two-to-one, our people wouldn’t have done anything until 1988. They wouldn’t have been forced to radically rethink their operating procedures.”

Hewlett-Packard’s stretch objective seemed difficult, but around the same time, Motorola had committed to goals that were even more aggressive. “In 1981, we developed one of the top 10 goals of the company, the Five-Year, Tenfold Improvement Program,” explained William J. Weisz, former chief operating officer of Motorola. “This means that no matter what operation you’re in, no matter what present level of quality performance, whether you’re a service organization or a manufacturing area, it’s our goal to have you improve that level by an order of magnitude in five years.”By 1986 Weisz was talking about another tenfold improvement and wanted his company to accomplish it in just three years. This aggressive objective was necessary because Motorola had lost more than 30 percent of its market share. The company’s efforts to reach this goal resulted in major performance improvement and led to Motorola winning one of the first Malcolm Baldrige National Quality Awards in 1988.

When Weisz was told by one of his vice presidents that there was no way the company could improve tenfold in just five years “doing business as they are,” Weisz answered, “You’re right. You can’t improve that much doing business as you are. You’ve got the message!”

The question that Motorola managers wrestled with between 1981 and 1986 was how they were going to measure a tenfold improvement. If your process has a first-time yield of 98 percent and you improve it to 99 percent, that’s little more than a 1-percent improvement (1/98 = 1.02%). If the first-time yield is 50 percent and you improve it to 95 percent, that’s a 90-percent improvement (45/50 = 90%). Even if you improve your first-time yield to 100 percent, that’s just a 100-percent improvement (50/50 = 100%), not the 1,000 percent that tenfold improvement would require.

Motorola’s answer was to focus on reducing error rates, not on improving the percentage of good items. If the error rate is two per 100 items processed (98% first-time yield), you get a tenfold improvement if you reduce the error rate to two per 1,000 items. However, if you’re measuring yield improvement, it’s only a 1.84-percent (1.8/98 = 1.84%) improvement.

Here’s another way of looking at error reduction. If the mean time-to-failure in the field is three defects per 1,000 hours of operation, you get a tenfold improvement if you reduce the failure rate to three per 10,000 hours of operation. Hewlett-Packard had limited its tenfold improvement effort to products for which accept/reject criteria were readily available. When Motorola required a tenfold improvement in all areas, that presented a major problem. Most of the support areas had never established an accept/reject criterion for their people’s output. Errors in support areas were considered only human and part of the way the company did business.

To address this measurement problem, Motorola invented the concept of “error opportunities.” An error opportunity is every characteristic of any output that’s evaluated by the individual who receives the output.

There’s the famous example of a manufacturing engineer who’s performing a design review on a drawing with 100 dimensions, and each dimension has a plus-or-minus tolerance. In reviewing only a single print for dimensioning, the engineer would have 300 error opportunities (100 dimensions + 200 tolerances). That doesn’t count the error opportunities in analyzing the part’s function.

There are many examples for calculating error opportunities. Personally, I like to look at the total, not the parts, when counting errors. I believe that’s what the person who receives any output does, too. Either the phone works or it doesn’t. My brakes stop me or they don’t.

I agree with the concept of error opportunities, but let’s not play games with numbers just to look good. Let’s focus on improving the organization’s performance levels. Is your Six Sigma program doing the following?
  • Increasing customer satisfaction
  • Increasing profits
  • Increasing value per employee
  • Increasing return on investment
  • Increasing market share

If not, then maybe you need to redirect your quality system, including your Six Sigma methodology.

Comments are closed.