Vol. 3, No. 8

Six sigma and more:  Performance evaluations may doom Six Sigma efforts? by David R. Schwinn A cover story in Business Week last month announced, 'Ford: It's Worse Than You Think. ' As I read the article, I became deeply disappointed. I saw a series of problems, but what caught my eye was the discussion of Ford's Six Sigma effort and its employee evaluation system, introduced last year.

The article described the Six Sigma effort as saving $52 million last year, but not improving Ford's quality. Ford officials said there were not enough project results yet to have an impact on quality. That may be true. I'd like to discuss another factor besides 'not enough time '--one related to the introduction of the performance evaluation system.

The new policy, according to the Business Week article, requires supervisors to rank employees from worst to best along a curve:10% get A's, 80% get B's, 10% get C's. Getting a C means no bonus. Two C's in a row represents grounds for demotion or dismissal. Already, class action suits have been filed protesting the system and its impact. As a result, Ford now requires that only the bottom 5% get C's. Without knowing the details of the system or the approach to its implementation, it is clear that creating a potentially artificial ranking and rewarding system and punishing people according to it makes very little sense. Dr. W. Edwards Deming, Alfie Kohn, and others have pointed out to managers all over the world since the early 1980s, including Ford, that systems that quantify human performance undermine commitment to improving an organization and create competition rather than cooperation among employees.

Ford's new employee evaluation system looks like many others. Rather than pick on Ford, let me generalize to the other systems I have seen that assume that an employee's performance can be judged quantitatively. People are entirely too complex for that. Even if we could find relevant numbers to capture, what about the ones we've missed? Let's say we could capture all the numbers, and then deal with the 'outliers. ' Statistical theory says the outliers are not 10%. They are not even 5%. They are generally less than 1%. What would be the logical response to these outliers? Well, we might reward, recognize, and learn from those outliers who have somehow beat the system. We could use that knowledge to change the system.

What about those outliers--employees whose performance is particularly below that of the system? Managers might work with them to find out why they are performing poorly. They may not have the tools, equipment, resources, management or staff support, desire, or skills necessary to perform. These things can generally be remedied with a cooperative effort. People used to ask Deming, 'What about the bad apples?' Deming would respond, 'Retrain them, maybe for another job--maybe for another job in another company.'That half-joking response had some wisdom in it. The point is if you find people who are really performing poorly, you work with them to improve their performance, looking at the whole system in which they labor. That exercise usually uncovers problems that, if corrected, will improve the performance of many others as well. If nothing can be done, then help the employee into a new role, maybe at another organization.

Now you may respond, that's all very nice but, bottom line, what's really wrong with identifying the bottom 5 percent and cutting their bonuses and putting them on notice that they may be demoted or dismissed? Let's talk about basic psychology. When the bottom 5 percent get punished, people do not want to be identified as the poorest performers. In order to avoid that, they will try to improve their performance. But improved performance will do them no good if everyone else's performance also improves. So they must also take steps to hold down other people's performance. People will attempt to grab and hold onto common resources. They will share no knowledge among their peers. They will attempt to make their peers look bad. They will focus on the numbers on which they are ranked and ignore the others. They will operate as individual agents and avoid teamwork at all costs. In general, they will try to simultaneously improve their individual performance while pulling down the performance of everyone else. Maybe this possibility explains why Ford's Six Sigma performance has been a little lackluster.

So, what's the alternative that will make a Six Sigma effort really cook? Start by being clear about the purpose of the performance evaluation system. Is it to decide who gets bonuses? Is it to decide who gets promotions? Is it to decide who gets raises? Is it to develop a legal record in case someone is to be dismissed? Or is it to improve performance? Most performance evaluation systems embrace all these purposes and even more, simultaneously. Different purposes frequently require different systems. A single performance evaluation system that attempts to serve all these purposes may end up failing at least some of them.

Let's assume we want a system that encourages excellent performance on the part of everyone, not one that encourages people to avoid being identified as poor performers. I believe the best way to do that is to involve everyone in the design of the system. Identify the numbers that are critical to organizational performance, while recognizing the downside of ignoring the numbers and the 'unmeasurables, ' such as success and integrity. Take this step carefully. Once the numbers have been identified, use good analytic statistical theory to analyze and react to and use the numbers to drive the Six Sigma effort. The suggestions I made earlier may be useful. Assume that performance problems lie with the system, not with the individuals, and you'll be way ahead of the game. Don't rank, reward, and punish according to some arbitrary numerical guideline.

If you are trying to encourage performance in an organization or an initiative, such as Six Sigma, design a system that encourages the performance of everyone.

As always, let me know what you think. You can reach me at support@pqsystems.com.