Simon London: Hello, and welcome to this episode of the McKinsey Podcast, with me, Simon London. A certain breed of executive has always looked down on the people side of management. They see it as soft, squishy, and lacking in hard data. But while that may have been somewhat true 20 or 30 years ago, it certainly isn’t true today. There really is a revolution in progress as companies start to apply big data and advanced analytics to the human side of the enterprise. To talk about the promise and pitfalls of people analytics, I sat down in Philadelphia with McKinsey partners Bryan Hancock and Bill Schaninger. As we’ll hear, Bryan and Bill are optimistic, with the caveat that getting real value from people analytics requires not only technical smarts but also a solid understanding of organizational behavior and a pretty good grasp on how the business actually makes money.

So Bryan and Bill, welcome back to the podcast.

Bryan Hancock: Thank you.

Bill Schaninger: Thanks for having us.

Simon London: So an obvious first question for a nonspecialist is: When we’re talking about people analytics, what are we talking about?

Bryan Hancock: What we’re talking about is bringing data on people to specific business decisions. It can be a decision to hire. It can be a decision on how to configure a team. It can be a decision on where to source people. But it is bringing a data set to people decisions. It’s as simple as that.

People analytics has existed as a concept for a long time. It’s not that there’s anything radical about the idea of people analytics. What’s cool is that now there are new sources of data and advanced computing powers that allow you to do more with the data. But the underlying idea of using data to inform people-related business decisions is not necessarily a new thing.

Bill Schaninger: We make decisions every day about who we hire, how we deploy them, what teams we put them in, what we have them working on. Then we sit in judgment of their performances. Every one of those decisions can be made better with data. Not all those decisions are equally important, so you don’t have to bring it to bear in all of them, but you should probably bring it to bear more than we are doing today.

Simon London: As you say, it’s not an entirely new idea. But what’s the opportunity today? What makes this a particularly important and interesting topic?

Bryan Hancock: I think what makes it a particularly interesting and exciting topic today is a combination of a few factors.

One, there really are new advanced computing capabilities that allow you to factor in more variables and determine more of what really matters. Of course, you can’t just do that independently. It has to be linked to good research in what matters, but the advanced computing power does matter.

There are new and different sources of the data—super exciting and interesting. Those matter.

Also, there is the acceptance, more broadly, of advanced analytics—be it from marketing to sports. That is making people think, “Oh, if I can do this in understanding my consumer or understanding the quality of my first baseman, why can’t I do this for understanding what makes a good salesperson tick?”

Simon London: Talk a little bit about the sources and the categories of data. What kind of data are we talking about here?

Bill Schaninger: There are some interesting pools that I think, in many cases, we hadn’t really connected. Maybe it’s just because we were isolating around the individual employee in a way that wasn’t helpful. So there’s the obvious. There’s the information about the employee: where they went to school, and where they’ve worked. That’s basic.

But then you can also get into things about their attributes, their traits, their personalities. And then you think, “OK, well, what other data do we have?” We can collect data about performance. We can also collect data about the environments they’re in: perceptions of the boss, the boss’s effectiveness, what they’re working on, who they’re working on it with, how long they’re working, the company overall, competitive positioning, location.

Historically we wouldn’t have always brought those together. But as soon as we say, “We really want to understand the person, the environment the person’s working in, who they’re working with and how that’s going,” it’s way more natural to bring that together. And you can start showing the longitudinal effects.

Simon London: And then, presumably, perceptions as well: That’s another bucket of data?

Bill Schaninger: For sure. Most data around the organization—its effectiveness, its climate—is perception based. A good rule of thumb about this is if you’re about to make a choice about a person, you can probably do it better with some data. It is really that simple. Basic things like, Who should we hire? Where should we go to recruit? Who should we put together on a team? What should that team work on? How well are they performing? Why are there differences between units in performance?

We have a couple of things coming together at once. Of your two capitals, we are long on financial capital, and we are short on human capital. So now every decision you make about people matters more. We’re at an era where computing power and analytic techniques have allowed us to do more than we could in the past.

In particular—in “stat speak”—you would swap the models really quickly. You could only test so many interaction terms. When they talk about machine learning and the advances in nonparametric statistics, basically, what they’ve done is said, “We’re going to ignore the rules of parametric statistics. We won’t assume everything’s normal. And we’re going to run many, many more combinations than previously possible to find unique segments of people.”

Simon London: From what we see today, how many companies are actually doing this at scale? How many companies have really adopted people analytics in a way that is really having an impact?

Bryan Hancock: There’s a spectrum of what people analytics can be and do. The majority of companies, the vast majority, are doing basic reporting-type analytics: Who’s turning over? Where are they turning over? They’re doing some initial root-cause problem solving as to what’s behind it. Most companies are doing assessments of folks who are coming in to see if they’re a good fit for the role.

The question is, how many companies are going beyond that basic reporting, basic analytics, to using some of the bigger data sets, to using advanced computing power, and combining those data sets with well-proven academic theory on what really drives performance in an organization? How many organizations are using that at the next frontier? Very, very few.

Bill Schaninger: Three years ago, when we first started talking about this and started investing in it heavily, the basic question was, Who’s about to leave? It was about retention. That was the earliest use case across the board. That was pretty straightforward, because you were predicting a one or a zero.

As the people who’ve bought into this have migrated toward answering business problems and considering the impact of people and the combination of people on business problems, the use cases have gotten better and more impactful. But that list is probably still pretty small. We’re still running into some similar roadblocks, if you will. There is a belief set, convenient or not, that “we don’t have enough data.”

This is not true. Whatever data you have, you can start with. And, particularly now that you don’t need a data lake, you need an ecosystem of data, that’s a real “unlock.” You can get at it. Acknowledge that the outside world probably already has better data on your employees than you do. Just go use it. It’s there. They’ve given it to other sites. Go get at it.