Human Resources: Key skills for analytics pros
By Ryan Hammond
When one thinks of big data and analytics, human resource management is not the first application that comes to mind. However, over the past decade, HR has significantly increased its sophistication in making data-driven decisions. Today, chief human resource officers (CHROs) are beginning to recognize the potential power of big data and are embracing the analytics revolution. Consequently, “people analytics” teams are beginning to appear in leading firms. Though these teams are new, they come from highly trained backgrounds and are looking to employ the same analytical tools that are now commonplace in marketing, logistics, finance and sales.
So what will this new era of people analytics look like? At hiQ Labs, we seek to help human resource teams find new and innovative ways to bring insights and efficiency to their organizations. What follows are some predictions about the future of this field, gleaned from our interactions with some of the most
forward-thinking people analysts across the country.
First, people analytics will go beyond the traditional data pools owned by HR (namely HRIS and employee engagement surveys) to mine insights from a much wider array of sources. Second, people analytics teams will have an increased diversity of skill sets that have not conventionally been found inside the HR function. Finally, the use of predictive analytics, machine learning and related techniques will require HR to think more creatively about the intersection of data and practice.
HR Meets Big Data
HR data are rarely located in a single, easily integrated system. The data tend to be “owned” by different stakeholders across the organization. They are often of highly varying quality and completeness, making them difficult to use. This is a huge roadblock for people analytics teams looking to do sophisticated analysis, even in well-integrated HRIS systems.
While working to improve this situation, people analytics teams have realized how much valuable data exist outside the confines of the HRIS and its related systems. For example, no internal data source can provide insight into employees’ professional Web presence, an increasingly important element in understanding a firm’s exposure to aggressive recruiters. hiQ Labs started to examine the data on the open Web and found that the most up-to-date, easiest-to-access data on an employee’s work history, accomplishments and competencies could be found not on internal company systems, but on their professional profiles in places like GitHub, Indeed.com and LinkedIn!
hiQ Labs predicts that in the near future people analytics will not be satisfied with rehashing the data in a company’s HRIS system. Instead, these newly sophisticated teams will capitalize on big and often unstructured data not accessible for analysis in the past. Often, the richest sources of data are those that are hardest to wrangle. These new data sources will put HR directly in the realm of big data and push it beyond traditional HR data sources. Data from the open Web, free-form resumes and narrative employee performance reviews are just some examples of what are available. To make full use of these data sources, HR practitioners will need new skills and tool sets.
Meet the New HR Analysts
At hiQ Labs, we meet people analytics teams from across the country at our corporate councils where they come to share insights, struggles and triumphs in building people analytic practices. We have been struck by how many of these teams are led and staffed by individuals who come from outside of HR. In the past, sophisticated data analysis within HR was typically done by Ph.D.s trained in fields such as industrial and organizational psychology, organizational behavior and economics. These will remain core pipelines of deep analytics talent.
Now, however, we are also seeing data scientists from engineering, human computer interaction, computer programming or, in the case of hiQ Lab’s head data scientist, astrophysics. Big data mining requires tool sets that are simply different from even the most quantitative HR professionals, and this is driving HR leaders to reach outside their usual talent pools to bring in that expertise. Tools such as Python, R and Hadoop are commonly used by these new analysts. This cross-fertilization also extends to the statistical techniques they bring to the table. Machine learning, semantic analysis and Bayesian methods are beginning to take a place beside the traditional workhorses of structural equation modeling, ordinary least squares and logistic regression.
These new tools crack open new data sources for HR, allowing teams to generate novel insights. Comb through mounds of unstructured text to create quantitatively analyzable data? Sure. Gather huge amounts of data about the variation in local labor market supply and demand, organized by skill set? Why not? Build and validate predictive models using testing and training data sets? Yes, please. If one linear model is not sufficient to cover the complexity of a phenomenon, why not neural networks? All these and more are now being explored within the HR function.
Analytics Meets HR Practice
The influx of new perspectives is just as important as the new tools. Challenging old assumptions, self-imposed limits and institutionalized practices will help spur innovative new ways of thinking about HR and its connection to the business. Put an astrophysicist in a room with HR people and interesting things begin to happen.
Here is a concrete example from our experience at hiQ Labs. Our data scientists were asked to consider the problem of analyzing attrition. HR has a long tradition of studying and analyzing turnover. In fact, it is among the most studied problems in the field. So what could a fresh set of eyes and tools bring to the table?
Our data scientists noticed that almost all the focus and analysis studied past attrition to answer the question of why people are leaving. Answering “why” requires careful analysis with clean estimates of causal effects. It also requires hard-to-get data on employee motivations and intentions, which evolve in real time.
These challenges left analysts in a position where they were chasing causality while only being able to catch correlation. Yet, these analyses drove the creation of expensive retention programs that often treated not just those at risk but a much broader set of employees. One of our scientists pointed out that many machine-learning techniques would have a tough time answering why people left in the past, but would do very well at predicting who might leave in the future. This insight led to our first product as a company – forward-looking, individual risk predictions that trade off a hypothesized chain of causation in favor of predictive accuracy.
HR can and should be running to catch up in the world of big data analytics, as the big data movement gains momentum. By building on existing analytical tool sets and the rapidly expanding base of open-Web data, HR will be given the capacity to actively manage HR processes in real time, while discovering new links between management practices, employee productivity and key business results.
Ryan Hammond is the head of people analytics for hiQ Labs. He has worked in people analytics since the early 2000s. He holds a Ph.D. from MIT’s Sloan School of Management and has worked in the United States, Europe and Asia. He can be reached at firstname.lastname@example.org