We are all prone to making judging based on past performance. Barry Staw, then at the University of Illinois and later at the University of California, conducted an experiment in which groups of participants were asked to estimate a company's future sales and earnings per share based on a set of financial data. Afterward, he told some of the groups they had performed well, making accurate estimates of sales and earnings per share, and told other groups they had performed poorly—but Staw did so completely at random. In fact, the "high-performing groups" and the "low-performing groups" had done equally well in their financial calculations; the only difference was what Staw told them about their performance.

Then he asked the participants to rate how well their groups had done on a range of issues. The results? When told they had performed well, people described their groups as having been highly cohesive, with better communication, more openness to change, and superior motivation. When told they had performed poorly, they recalled a lack of cohesion, poor communication, and low motivation. Staw concluded that people attribute one set of characteristics to groups they believe are effective, and a very different set of characteristics to groups they believe are ineffective. That's retrospective attribution (the Halo Effect) in action.

Of course, these findings do not mean that group cohesiveness and effective communication are unimportant in group performance. It only means that you can't hope to measure cohesiveness or communication or motivation by asking people to rate themselves when they already know something about the outcome. Once people—whether outside observers or participants—believe the outcome is good, they tend to make positive attributions about the decision process; and when they believe the outcome is poor, they tend to make negative attributions. Why? Because it's hard to know in objective terms exactly what constitutes good communication or optimal cohesion or appropriate role clarity, so people tend to make attributions based on other data that they believe are reliable. Performance is a cue by which people attribute characteristics to groups and to organizations.

Some people questioned Staw's findings. They doubted whether an experiment that put strangers together for just thirty minutes could accurately capture the perceptions of work groups. A team led by H. Kirk Downey at the University of Oklahoma therefore replicated Staw's study, using the exact same set of financial problems, but with groups of people who had a prior history of working together, and giving them considerably more time to make their calculations. Again, groups were told—at random—that they had performed well or poorly. The results were virtually the same as in Staw's experiment. Once again, "high-performing teams" reported that their groups had been more cohesive, that teammates were of high ability and had enjoyed working together, that communication had been of a high quality, that they had been open to new ideas, and that overall they had been satisfied with the group process. All because of the randomly assigned description of performance—nothing more. Like Staw, Downey and his colleagues found a strong tendency to make attributions on the basis of performance.

Surprising? It probably shouldn't be. Picture a group where people express their views vigorously and passionately, even arguing with one another. If the group performs well, participants might reasonably look back and say that open and forthright expressions of opinion were a key reason for success. They'll say: We were honest, we didn't hold back—and that's why we did so well! We had a good process! 1Biut what if the group's performance turned out to be poor? Now people might recall things differently. We argued and fought. We were dysfunctional. Next time we should follow a respectful and disciplined process.

But now imagine a group where people are calm, polite, and respectful of one another. They speak quietly and in turn. If the group does well, participants might look back and credit their courteous and cooperative nature. We respected one another. We didn't fight. We had a good process! But if the same group's performance was poor, people might say: We were too polite. We censored ourselves. Next time, we should be more direct and open, not so concerned about one another's feelings. The fact is, a wide variety of behaviors can lead to good decisions. There's no precise way to engineer an "optimal" discussion process. We may try to avoid extremes, sure, but between those extremes is a wide range of behavior that might be conducive to success. And because we really don't know what makes an optimal decision process, we tend to make attributions based on other things that are relevant and seemingly objective – namely, what we’re told about performance outcomes.

A serious scholar of leadership, the late James Meindl at SUNY Buffalo concluded after a series of insightful studies that we have no satisfactory theory of effective leadership that is independent of performance. We think we know what good leadership is all about—clarity of vision, communication skills, good judgment, and more—but in fact a wide range of behaviors can be said to fit these criteria. Show me a company that delivers high performance, and I can always find something positive to say about the person in charge—about the clarity of his or her vision, about good communication skills, sound judgment, and integrity. Show me a company that has fallen on hard times, and I can always find some reason to explain why the leader failed.

All of which brings to mind a 1964 Supreme Court case about free speech and pornography, in which Justice Potter Stewart memorably wrote that while he could not provide a good definition of hard-core pornography, "I know it when I see it." Since good leadership is usu­ally difficult to identify in the absence of data about performance, it seems that leadership is even more difficult to recognize than is hard-core pornography—which at least Justice Stewart knew when he saw it. For all the books written about leadership, most people don't recognize good leadership when they see it unless they also have clues about company performance from other things that can be assessed more clearly—namely, financial performance. And once they have evidence that a company is performing well, they confidently make attributions about a company's leadership, as well as its culture, its customer focus, and the quality of its people.

Retrospective attribution is how individuals think about decision processes, an organization's people, and leadership—and it doesn't go away when we conduct large-scale surveys, either. Quite the contrary. If we're not careful, surveys might be little more than large collections of retrospective attributions. Consider “Fortune” magazine's annual ranking of the “World's Most Admired Companies”, the one that named IBM as Most Admired in 1983 and 1984. Every year, Fortune asks thousands of business executives and industry analysts to evaluate hundreds of companies in eight categories: quality of management, quality of products and services, value as a long-term investment, innovativeness, soundness of financial position, ability to attract, develop, and retain talented people, responsibility to the community and environment, and wise use of corporate assets. Mix the answers together and you get the World's Most Admired Companies in each of these categories—as well as the overall winner. It's an impressive effort, and it produces an eye-catching cover story every year. Over the years, Fortune has named not just IBM, but luminaries like General Electric, Wal-Mart, and Dell—a very impressive bunch.

But when some researchers took a closer look, they found that Fortune’s Most Admired” ratings were heavily influenced by retrospective attribution. The scores on the eight different factors for a given company turn out to be highly correlated—much more than should be the case given variance within each category. Furthermore, many of the scores were very much driven by the company's financial performance, just what we would expect given the salient and tangible nature of financial results. Two different studies showed that a company's financial performance explained between 42 percent and 53 percent of the variance of the overall rating. In other words, when a company posts high profits and its stock price is moving upwards, the people who fill out Fortune's survey tend to infer that its products and services are of a high quality, that it is innovative and well managed, that it is good at retaining people, and so forth.

Cisco offers a case in point. In 1997, the same year Cisco leapt onto the cover of leading business magazines, it made its first appearance on Fortune's Most Admired list, entering the charts at number fourteen. Then it rocketed upward, reaching number four in 1999 before topping out at number three in 2000. It's no surprise that Cisco rated high for investment value—its stock value was, after all, going stratospheric. But Cisco was rated high for lots of other things, too: quality of management, innovativeness, quality of people, and more.

When the tech bubble burst and Cisco's stock fell, in 2001, Cisco's rating as an investment value quite naturally fell. But with retrospective attribution, its ratings fell across the boards. Cisco was now less admired for innovativeness, for people, the whole works. Its overall rating dropped to number fifteen in 2001, then twenty-two in 2002 and twenty-eight in 2003.

Fortune's survey isn't the only one to be undermined by retrospective attribution. Remember the “Financial Times’” survey of Most Respected Companies?  In 1996, when ABB was at its peak, it was rated high across the boards, for business performance, corporate strategy, and maximizing employee potential, and its leader was applauded for his strategic vision and focus. Again, the pattern is entirely consistent with retrospective attribution.

And there’s more. In 1984, an organization called the Great Places to Work Institute made a big splash with a book called The 100 Best Companies to Work for in America. Every year since then, it has compiled the Best Companies to Work For index. Based on these findings, the “International Herald Tribune” claimed that being a Great Place to Work leads to high performance, noting that the companies on the 1998 list had a total market return (share price plus reinvested dividends) over the next five years of 9.56 percent, compared with a return of 3.81 percent for all the companies on the S&P 500.

The inference was clear: Companies that care about creating a great place to work will attract good people and help them be more productive, leading to superior performance. It all makes good sense. But how did the institute determine what's a great place to work? Simple, they asked employees. Employees were asked to rate their companies on two attributes: trust and culture. The trust index had five elements: credibility, respect, fairness, pride, and camaraderie. Credibility, in turn, was measured by responses to statements like this: Management keeps me informed about important issues and changes. People around here are given a lot of responsibility.

High agreement meant high credibility, which meant a Great Place to Work. Respect was measured by asking for responses to questions like this: Management involves people in decisions that affect their jobs or work environment. I am offered training and development to further myself professionally. Again, high agreement meant respect, which was associated with being a Great Place to Work. The website also gathered comments like this one, said to be from an employee in a sample company: "There is a high level of trust & empowerment here. We are not bound by any rules & we can do whatever we want at work. We receive encouragement & motivation from our team leaders. We have company events & wellness pro­grams which allow us to balance our personal & professional lives."

At first glance, this all looks plausible, but it's undermined by retrospective attribution. Companies that are profitable, prosperous, and growing fast will often be perceived as desirable places to work. Again, look at Cisco. It debuted on the charts in 1998 at number twenty-five, then climbed to twenty-third place in 1999. In 2000, when Cisco was briefly the most valuable company in the world, it shot up to third place, where it stayed for two years. Once the layoffs hit and the stock price tanked, how was Cisco rated as a Great Place to Work? It fell to thirteen in 2002, then to twenty-four, and finally twenty-eight in 2004 – not exactly tracking performance, but pretty close.

Did Cisco become a worse place to work after 2000? Yes, if we think in terms of employee moral and the chance to get rich. But that’s a reflection of performance, not a cause of it. If you don’t believe the Fortune and Best Place lists are shaded by retrospective attribution, you have to believe that the people who filled out the surveys are not affected by the same tendency found in participants of Barry Staw’s experiments, which would seem doubtful.

 Business is shaped by a number of delusions: in particular remember:

·                        If independent variables aren't measured independently, we may find ourselves standing hip-deep in Halos.

·                        If the data are full of Halos, it doesn't matter how much we've gathered or how sophisticated our analysis appears to be.

·                        Success rarely lasts as long as we'd like—for the most part, long-term success is a delusion based on selection after - the fact.

·                        Company performance is relative, not absolute. A com­pany can get better and fall further behind at the same time.

·                        It may be true that many successful companies bet on long shots, but betting on long shots does not often lead to success.

·                        Anyone who claims to have found laws of business physics either understands little about business, or little about physics, or both.

·                        Searching for the secrets of success reveals little about the world of business but speaks volumes about the searchers—their aspirations and their desire for certainty.

Once we've swept away these delusions, what then? When it comes to managing a company for high performance, a wise manager knows:

·                                           Any good strategy involves risk. If you think your strategy is foolproof, the fool may well be you.

·                                           Execution, too, is uncertain—what works in one company with one workforce may have different results elsewhere.

·                                           Chance often plays a greater role than we think, or than successful managers usually like to admit.

·                                           The link between inputs and outcomes is tenuous. Bad outcomes don't always mean that managers made mistakes; and good outcomes don't always mean they acted brilliantly.

·                                           But when the die is cast, the best managers act as if chance is irrelevant—persistence and tenacity are every­thing.

Will all of this guarantee success? Of course not. But I suspect it will improve your chances of success, which is a more sensible goal to pursue.

 With thanks to "The Halo Effect" by Phil Rosenzweig