Tools that crunch numbers and provide fast, accurate, and correct decisions are now omnipresent. Amazon’s algorithm recommends products to users, Netflix makes it easier for us to decide what to watch. Data analyst has become a sexy career. Algorithms do not only simplify our free time and buying decision, but they are also helping companies to make better business decisions. This trend of using data to make decisions also applies to human resources (HR).
HR has long been considered a pure cost center. It imposes processes that complicate hiring, firing, and promotion decisions to name just a few. However, over the past years, HR has started to reinvent itself. The Harvard Business Review dedicated an issue to the changes in HR in 2015. In light of its move to reposition itself, HR can make use of the information it gathers about employees to transform itself to a strategic unit that provides expert insights into how to manage the workforce. Data analytics will be part of this new HR.
During the Wharton People Analytics Conference a number of trends were presented. These related to the topics of teams, methods, diversity, performance evaluation, and limits of analytics.
On the topic of teams, the focus on numbers has been pushed to the background in favor of discussing the importance of establishing a climate in which teams can operate and processes that support high team performance. The focus was on teaming, individuals being able to deal with coordinating each others input, being present for the team, and putting the team before the individual. Achieving high performance teams requires ‘soft processes’ at times hard to quantify. However, companies such as Humanyze might provide the necessary tool to collect more data about teams. The question remains if employees will cooperate.
The Behavioral Insight Team, during the method session presented presented a new hiring processes. Hiring is a difficult processes due to it subjectivity. The CV is full of information that creates biases for or against the candidate. However, an algorithm doesn’t care about ethnicity, gender, age or the reputation of your university (unless you program it, of course). They tested the effectiveness of their hiring tool by selecting candidates the traditional ‘human’ way and through the algorithm. All selected candidates then proceeded to the company’s assessment center. The outcome was that several people who got hired, would have never hired, if only humans would have been involved in the recruitment processes.
The next session discussed the topic of diversity. An Important point raised is that diversity is in the eye of the beholder. It is always compared to a reference category (e.g., women vs men or African-american women vs Caucasian women). This reference category will determine the direction of the discussion in your company.
The next discussion turned towards the topic of performance management. This touched upon the disliked practice of appraising the performance of others through forms. The current goal of performance management is often focused on determining the top and bottom performers, and take actions based on the resulting ranking. However, recent trends moved towards ‘ongoing discussion’ as a way to manage the performance of employees. At Adobe this turned into the abolishing of forms and a system of check-ins which created maximum flexibility. Other companies, for example Deloitte, also moved away from the annual performance management to more regular discussion but retained forms. Other companies, such as General Electrics are also developing new ways to assess and manage performance.
The conference ended with a discussion on limits of analytics. This discussion tied well into the opening keynote of Daniel Kahneman. Every tool is only as good as its programmer. Of course algorithms can be programed to learn, but the point is that most analytical solutions work well for predictable scenarios. According to Daniel Kahneman this provides a wonderful opportunity. Machines will not take away all of our jobs. Unpredictable manual and cognitive work will be done humans. Tools should be used for making decisions they can do better, faster, and more accurate than we can do. Once things get fuzzy we jump in. The discussion also highlighted that decisions derived from analytics will only be implemented if the client trusts the algorithm. Given clients even a small chance to modify the algorithm can greatly increase their trust in the tool.
As someone with one foot in the educational field, a discussion that was missing at the conference was how to spot learning needs and development opportunities. What is the optimal sequence of job function for a specific senior position. Is there one? How to develop talent? Shall employees be stimulated to job hop within the company? What about the benefits of sending employees to learning programs, letting them do MOOCs or paying for their MBA? These questions all point towards the need to be able to prepare the workforce for tomorrow’s unknown and unpredictable challenges.
If you are interested in recent advances in computational social sciences, have a look at recent posts on The Observatory.