HR Analytics is Primarily a First Step in Anticipative Policy

By: Timothy Vermeir
Interview with Prof. dr. Luc Sels and Prof. dr. Bart Baesens
Translated from

In HR circles, HR analytics has been talked about for many years. Today, there are reasons to believe that this might be the time for businesses to take the plunge. Thus, HR would – finally – base decisions and policies on hard data, like other management domains already do. “That is one way to make clear that you can be a lever for better results, for more informed decisions and for strategic value.”

HR analytics is where the disciplines of professors Bart Baesens and Luc Sels intersect. Luc Sels with his HR expertise is able to make the knowledge that his colleague Bart Baesens, a member of the Department of Management Information Systems from the University of Leuven, gained in other business domains applicable to human resources. Bart Baesens started his research in analytics and (big) data during his PhD on credit scoring. He then shifted his focus to marketing analytics, fraud analytics and now HR analytics. His career runs parallel with the applications of analytics in business, with human resources as the new kid on the block.

Luc Sels: “It strikes me that HR has a huge amount of data and has actually digitized quickly – the integration of HR processes into ERP packages and the development of the social secretariats have played an important role – ,but that in no other domain so much time has elapsed between the data being available and it actually being used for analytics. Companies have huge amounts of data about their employees: background data on people you recruited or have not recruited, data on which employees collaborated with which customers, on billable hours, on absenteeism, on career moves, relational data … You can name it and it is somewhere in the company. The problem is not the availability of data, but the strategic use thereof.”


It might be a good idea to properly define (HR) analytics first…

Bart Baesens: “Analytics is a very broad term. People who put data into a spreadsheet to create graphs, actually already do analytics. It is typical that you start with very descriptive analytics , where you look at a number of key metrics, extract a number of graphs from the data, explore the information that you already have available. ”

“In addition to purely describing, you can also go one step further and move towards predictive analytics. You can then make predictions about e.g. absenteeism, burn-out, engagement … ”

Luc Sels: “HR analytics is about analytically dealing with available data in terms of a particular goal – I would therefore rather stay away from the term ‘Big Data’ in this domain, because when is data ‘big’? The goal for me is to improve business processes and results by better understanding the relationships between, for example, employee behaviour and customer behaviour, or behaviour of employees and results of employees. Analytically dealing with that large amount of data, has to help you make better decisions and help you better support decisions. They call it ‘data-driven decision making’.”


In the article “Is Your Company Ready for HR Analytics? (published in MIT Sloan Management Review, together with Sophie De Winne) you write that this is the time to invest in HR analytics. Why now?

Bart Baesens: “I think we can answer that question from two points of view – Luc from an HR perspective and myself from my analytics perspective.”

“In the field of analytics, we have learned a lot of lessons by trial and error in all other domains. We learned how to link data together, how to clean up data, how to build analytical models and which steps you have to go through in the process. We learned which analytical techniques to use and how to evaluate your models, how to validate them and how to build a bridge to the business. Today, we are already so far along that we feel that we can easily apply everything in HR. ”

Luc Sels: “My feeling is also that for 20 years, we have kept on talking about the importance of HR moving in the direction of the core of the organization. HR should not just organize their own processes optimally and help realize engagement and commitment, but also be a lever for a more efficient and effective deployment of staff in the operations of an organization. Through the sensible use of that gold mine of HR data, the latter is now more easily possible, especially when linked to other data. ”

“Fortunately, we have learned that people should not only be considered as a cost. But that does not mean that the total payroll does not considerably prevail in many organizations. That should have to encourage the use of analytics even more, in order to help ensure adequate deployment. Today, the tools are available and provide results that are understandable to a layman, the data are analysed much more easily than roughly ten years ago … HR needs to meet the challenge themselves, because it is a way to make clear that you can be a lever for better results, for more informed decisions and for strategic value. You should seize that opportunity.”


How disruptive is this for HR? How much can data analytics change the domain?

 Bart Baesens: “I think the disruptive insights will mainly come by linking data together. Not only within the HR functions, but also if you are going to make the link with various other departments of the company. Then some valuable insights will be established that impact the dynamics of the company, among employees, but also between employees and customers, and other stakeholders. ”

Luc Sels: “There are also some risks for the HR profession, I think. If you connect HR to data and results in other domains, you can also bring HR in the area of interest of those other domains. You win legitimacy, but the risk is that the HR data can suddenly be incorporated into the overall analytics exercise of the company. That engineers and IT specialists get to work with the data, and HR is “imposed” what they are supposed to do on the basis of these analyses. It’s a double-edged sword: on the one hand you gain legitimacy, on the other hand your life is lived by others. I do not know where this is going to end… ”


Then the HR manager might merely become an executor of what the data prescribes…

 Luc Sels: “The art of a strong HR manager is to filter out for which questions you fall back on analytics, and for which ones you do not. Not all decisions can or need to be supported by analytics, and you should also never let yourself be driven by the data only: you also have a certain vision of the culture that you want to have in the organization and that perhaps contradicts certain links that you get from the data. ”

“Analytics can, for example, make it clear that the gap between good and bad sales people is so large that you would better lay off the bad ones, while your HR gut feeling tells you that this step can eventually have a very negative impact on the good ones. Well, then you should not do that. But you can perhaps learn from the analysis of the good sales people what you can do with the bad ones to reduce the gap. As HR, you cannot lapse into a tendency to play scientist in your company and to be guided entirely by evidence: there is also a normative side, a values side, a culture side that are all equally important. The big challenge is to filter out when it is and when it is not useful, and to learn to convince the board NOT to follow the conclusions of analytics when it is inconsistent with the culture or the vision of the organization. That is something that only HR can do.”


How do you decide whether a problem can best be solved by looking at the data, or whether HR-expertise and humanity are needed?

Luc Sels: “It is very difficult to say for which topics analytics can and cannot be used, because it very much depends on whether there is data on the subject …”

“The trick is to dare to make a distinction between long and short term. If you have a problem between a manager and an employee, then you cannot wait for the analysis of the data to determine whether this is a symptom of a broader pattern within the company. No, you have to tackle the problem. But if there is a second complaint, and a third, a fourth … then it becomes an issue and that should ring a bell that this is a risk factor in the company and that you should look for a pattern or explanatory factors in the data, parallel to the HR solutions in the short term. That can give more structure to future decisions. ”

“HR analytics is primarily a first step in anticipative policy: you obviously respond to problems, but if problems continue to repeat itself, then you should move to another mode and see if you can anticipate the problems in the future. What are the levers that we should use? This could come out of the data, although the possibilities depend on the type of data that you have. Do you have more ‘soft’ data, collected via surveys or climate measurements? This data is common, but is rarely linked to background variables, or individual or team performance. And yet, it is only by these links that it really gets interesting. Are the more satisfied teams also more productive? ”


Where do you start in HR analytics?

Bart Baesens: “You can start from the data and see what you can do with it, but you can also start from a problem and see which data can help you to solve it.”

“You have descriptive analytics and predictive analytics. With descriptive analysis, you dive into a dataset and you find certain things in it. That can be clusters: segments of employees who are homogeneous with respect to a particular type of behaviour, for example, in terms of job performance or engagement. You can also discover anomalies, such as an employee who performs exceptionally well or exceptionally weak compared to his or her colleagues. ”

“With predictive analytics, you set out from a crystal clear question and you are going to predict something. You can start, for example, with data on all employees and employee turnover as your target variable. Then, using analytical tools, you can see which explanatory variables there are. This leads to an analytical model, which is actually a mathematical formula, but can, for example, also take the form of a decision tree. That model tells you what the probability is that something happens given what we know about that employee. An important assumption you make here is that the historical data is representative of the future business setting. That is sometimes a fairly heavy assumption, but it is the best we can do.”


What is low-hanging fruit for HR analytics?

Luc Sels: “That varies from organization to organization and to a large extent depends on the problems that are identified. Today, the technology is so advanced that when the data is available, in principle, many cases are low-hanging fruit. Today, you can identify what the prominent problems are, and then search for an answer with analytics. In a company that has a 15 percent voluntary turnover, this will probably be churn prediction, while a company that finds that its clients are much more loyal to one store or the other, might have to figure out the differences in leadership, management and team cohesion that help explain the discrepancies. ”


HR has a lot of data available, as we have already mentioned several times. But what are the requirements to ensure that this data is useful for HR analytics?

Bart Baesens: “That is a very important issue. One often talks about the GIGO Principle: Garbage In, Garbage Out. If you have bad data, you can have the most advanced analytical model and put the smartest data scientists to work, you will get nothing. Data quality is therefore important, though it is a very multidimensional concept. You can look at the accuracy of the data, but also at the completeness of the data – you should have no missing values – or the recency of the data. ”

“In those three dimensions, you will never achieve a perfect score: you are never going to have one hundred percent accurate data, never one hundred percent complete data nor one hundred percent recent data. If you wait until the data is perfect, you will never produce anything in terms of analytics. Each dataset contains noise, but you must learn to deal with the data and the noise. The analytical techniques should also have a sort of built-in robustness to deal with this issue.”

Luc Sels: “There are two additional aspects that are typical for HR. Firstly, data often has a limited lifetime. With individual data, this can be dealt with, but once you aggregate to team data, for example, this is a problem. Any change in a team, with a new employee or supervisor, or a new division of the customer portfolio, creates a new reality. So, you have to be very careful not to use the model too long without testing if it is still correct.”

“Secondly – and this is very difficult in many companies -, you need to store data in relevant categories in terms of content, instead of the categories that are laid out purely because of legal or administrative reasons. For example, I conducted research on the databases of social secretariats, and found that the period of notice was just registered as months worked, because they only required the information in the context of payment. It is, however, impossible to calculate the total cost of dismissal, which is an analysis that might be relevant from a policy point of view.”


Does HR analytics require a large investment from companies?

Bart Baesens: “That really depends on whether you are going for the low-hanging fruit or not. If you have a dataset of 200 employees, neatly ordered in a spreadsheet, and you primarily want to extract a number of graphs, then the investment is not so big. But if you go to the advanced predictive analytics for large datasets, the investment can be quite substantial.”

“If you are going to do analytics, you should take the total cost of ownership into account, with the software and hardware, but also – and this is the lion share in analytics – the costs of the data scientists that you need to recruit.”

“A major concern in analytics, for HR but also in other areas, is the return on investment – what are the benefits? There are some tangible things, like when you can save FTEs by automating certain tasks. But it is not always so clear. There are applications of analytics that will improve  and complement employees’ decisions, and it is much more difficult to identify what the economic value and impact of better decisions is. ”

“There are many studies that report significant returns, though I have – frankly – not seen any yet for HR analytics.”

Luc Sels: “It is also about how we can increase the likelihood that CV’s are screened in a good way. It is not just about the interests of the company, but also about the interests of candidates.”


How optional is HR analytics? Is this something that companies can ignore today without it having an impact on their success? Or is it really necessary?

Luc Sels: “Above all, HR managers need – even more than is the case today – to become more evidence-based in their attitude. Does that mean that they should immediately go to more complex HR analytics models? That is something else. But there is certainly room to act less on the basis of impressions and intuition alone. And looking more at the evidence that can be found in data, starting with simple, descriptive statistics. ”

Bart Baesens: “You have two important assets as a business: customers and employees. What you want to do with analytics, is get to know your customers and your employees better in order to service them better. If you say you do not want to invest in analytics, then you are saying that you are not looking at all the possibilities that are available to get to know either the customers or the employees.”

Luc Sels: “But again, there is another side to that story. After all, you can also use the data to implement a very hard policy. And just because of this, is it so important to continue to emphasize that HR must not lose its identity here. More than that: HR managers are required to ensure that companies use the data in a correct, socially acceptable manner, in consultation with the employees or their representatives. In the competency models for HR managers that Dave Ulrich developed 15 years ago, it was already stressed that a minimal expertise in information management – what should you (not) do with data and information, and how to reach agreements on information usage – is a skill that is crucial to HR. An understanding of what is present in the databases of the company, and how HR data can be linked to operational data, to performance data, and so on, would help us well on our way.”