The fourth and final challenge with talent analytics that I want to emphasize today is reverse causality. And this is the tendency, when we see two correlated factors, to believe that one causes the other. Especially if there's an intuitive direction. We look for this kind of causality, it's natural for us to seek it. But the trouble is often the causality, there may be no causality at all but worse, the causality might actually go in the other direction. So for example, are charismatic leaders more successful? This is a widely held opinion. And it makes a lot of sense. One can make the argument that well, there are all these benefits to being charismatic, it's naturally associated with successful leaders. And so sure, let's just go with that. Charismatic leaders are more successful. Fortunately for us, a few people have studied this. Agle et al a few years ago published a paper in the Academy of Management where they looked at charisma and success among 128 CEOs of major US firms. They used a longitudinal design so they could actually look at the perceptions at one time and the outcomes at another. So what they found was that charismatic CEOs did not actually have more organizational success down the road. But interestingly, successful CEOs were perceived as more charismatic. So it is not the case that charisma leads leaders to be more successful but rather just the opposite. That the success itself breeds in us the perception that they're charismatic. We find similar dynamics around power. And the warning is that we have to be very careful about what causal stories we tell ourselves. So we're driven to make sense of the world. We're driven to make sense of the world we live in, to build causal stories from what we observe. The trouble is that leads us to see things that don't exist. It leads us to give people credit and people blame in situations they may not deserve it. An interesting study that showed this in kind of profound ways was an early study, a 1977 study by James March the very famous organizational scholar James March. And his son, also James March, they looked at the careers of school superintendents in Wisconsin. So they got this great data set. They could look over a long stretch of time, 32 years, everybody who'd been a superintendent in the state and their trajectory of their career. So you can imagine, how do you think this goes? What would you expect to be the story? You would expect that there's great range in abilities among superintendents. And there are a range of differences in the jobs. You might want the jobs in the bigger cities and not be as interested, or maybe a few people want them in the bucolic small towns. But you would expect great variation in both places and there's probably this formal evaluation system that makes sure that the best people get placed in the best jobs. This is what we'd expect, this is kind of what the system is designed to do. What they found, however, is that when you look at what happens, it is very difficult to distinguish the observed career paths from what you'd expect from a completely random process. That even though we might look and say, well, this superintendent achieved this success, or this superintendent achieved this failure. Or this guy is in this place, this woman is in that place, for certain stories we can tell the stories. It turns out there's very little difference. It's hard to find any difference from what we observe in the data and what we would have expected if it was just completely randomly distributed, matching people to jobs by chance. So these guys conclude their study with a profound statement that's a good warning to all of us about reverse causality. The normative lesson is the stories we tell each other about success and failure in top management, like the stories we tell about success and failure in gambling, are in large part fictions intended to reassure us about justice and encourage the young. So a little bit of a cynical note from Jim March and son there, but it's well intended. It's well intended because he wants to urge us, don't draw these inferences. Be careful about causal reasoning. Be aware that often the causality runs the other way. And you've gotta run the data, basically. The principal means by which you do this with talent analytics is make sure you have some causal structure in your data. That you're using longitudinal panels. That you're very careful about the inferences you have to draw when you're stuck with just correlational data.