Will Data Science Save HR—or Kill It?
After listening to what data-science consultants are doing in human resource departments, it seems to me that data science may well save and elevate the function of human resources—yet it will also eliminate much about how HR, as we know it, functions.
In the 1920s, sophisticated management had become a branch of engineering: scientific management, time-and-motion studies and the related practices that grew out of industrial engineering programs. They saw the goal of management as getting employees to conform to the logic, the pace and the rationality of modern machinery.
The human relations movement of the 1930s—which explored the link between employee satisfaction and workplace productivity—was a reaction to that engineering approach. In 1960, Douglas McGregor described the debate, or more accurately, the continuing battle between the engineering orientation and the newer behavioral orientation as Theory X versus Theory Y, respectively. The decades since saw the gradual victory of Y over X. At least as traditionally practiced in industrial engineering, Theory X was essentially gone. About the only workplace topic where real engineering principles remained relevant was in setting up work schedules to accommodate demand fluctuations. Teams now are ubiquitous, local leaders are empowered to make their own decisions, CEOs worry about intangibles like culture and so forth.
Now comes data science. A good way to orient yourself is to note that data science came from and resides in engineering schools. At its core is the same notion that drove scientific management: the possibility of optimizing performance and outcomes with rules derived from objective principles.
When data scientists descend on a workplace, the first place they typically land is on those practices that have been delegated to line managers, starting with hiring. The question they ask is, “What criteria are you using to hire?” The answer they typically get is that, in practice, hiring managers do more or less what they want, using whatever criteria they want. The data scientists then ask, “What measures do you use to see if your hiring practices are working?” The answer is usually, “Well, none.”
At this point, the heads of the data scientists are ready to explode—truly, the data scientists I know find such a response hard to believe—and they set to work gathering measures of good performance and then building models to predict who will be a good performer. Then, they go a step further by calculating how much value that adds. Now, they have the attention of the business leaders.
In the process, the data scientists start rebuilding the capabilities that HR had for almost a century. So far, so good for HR as a function, right? They are making the case that HR could not or would not make for itself: that taking these practices seriously and handling them systematically has a real payoff for the business.
But they don’t stop there. They also ask questions about the practices that still exist, including those anchored in the paradigm of psychology and the theories behind it. Take, for example, a practice like the use of personality tests for assessing candidates, something that has almost a century of science behind it. Data scientists ask, “How much does this predict job performance?” The psychologists answer, “It is a valid predictor with a solid theoretical justification.” The data scientists reply, “That’s not what we are asking. What we care about is predicting as much as possible about who will be a good performer—optimization—and that relationship with personality is pretty tiny. So we’re going to build our own model.”
In workforce planning, in succession planning, in practice after practice, the data scientists are starting to ask, “Does what you are doing actually predict anything? We don’t care whether everyone else does it or whether you’ve always done it. We are going to check to see if it works, and when it doesn’t, out it goes.”