212
Collins University of Colorado, looked at our dilemma and concluded that we do not have a statistics problem, pointing out that the concept of "statistical significance" applies only when sampling of data is involved. "Look, you didn't sample companies" he said. "You did a very purposeful selection and found the eleven companies from the Fortune
500 that met your criteria. When you put these eleven against the
seventeen comparison companies, the probabilities that the concepts in your framework appear by random chance are essentially zero" When we asked University of Colorado applied mathematics professor William P. Briggs to examine our research method, he framed the question thus What is the probability of finding by chance a group of eleven companies, all of whose members display the primary traits you discovered while the direct comparisons do not possess those traits He concluded that
the probability is less than 1 in
17 million. There is virtually no chance that we simply found eleven random events that just happened to show the good-to-great pattern we were looking for. We can conclude with confidence that the traits we found are strongly associated with transformations from good to great. Q Why did you limit your research to publicly traded corporations Publicly traded corporations have two advantages for research a widely agreed upon definition of results (so we can rigorously select a study set) and a plethora of easily accessible data. Privately held corporations have limited information available, which would be particularly problematic with comparison companies. The beauty of publicly traded companies is that we don't need their cooperation to obtain data. Whether they like it or not, vast amounts of information about them area matter of public record. Q Why did you limit your research to US. corporations We concluded that rigor in selection outweighed the benefits of an international study set. The absence of apples-to-apples stock return data from non-U.S. exchanges would undermine the consistency of our selection process. The comparative research process eliminates contextual "noise"
similar companies, industries, sizes, ages, and so forth) and gives us much greater confidence in the fundamental nature of our findings than having a geographically diverse study set. Nonetheless, I suspect that our findings will prove useful across geographies. A number of the companies in our study are global enterprises and the same concepts applied wherever they did business. Also, I believe that
much of what we found-Level 5 leadership and the flywheel, for instance-will be harder to swallow for Americans for from other cultures.