Nowadays, many firms are already using big data and analytics to manage and optimize their customer relationships. Both technologies can also prove beneficial to leverage a firm’s other key assets: its employees! Various HR analytics (also called workforce analytics) examples can be thought of.
A first application is analyzing employee churn, or turnover, which is a major problem for many companies these days. Great talent is scarce, hard to retain, and highly sought by headhunters. Hence, given the well-known direct relationship between happy employees and happy customers, it is of utmost importance to keep employees and understand the drivers of employee dissatisfaction. Similar to customer churn, analytics can also be used for this purpose. A first step to doing so is to collect historical data on employee churn. The more data, the better, as this gives more opportunities to find previously unknown, interesting insights into employee behavior. Popular examples are staffing data, performance and productivity data, engagement data (e.g., collected through surveys), payment data, and task-specific data, among other types. Following the analytics process model, in a next step, the data will be consolidated, aggregated, and cleaned such that it becomes ready for analysis. An analytical model can then be built to predict employee churn. From a pure analytical perspective, the cross-fertilization between customer churn and employee churn is immense, since it essentially also boils down to a classification exercise of which the results need to be evaluated in a business-relevant way. Recommended analytical techniques are logistic regression and decision trees since both are white-box techniques which provide clear insights into why employees leave the firm. HR directors can then use these insights to work out employee retention strategies.
Another interesting application of HR analytics is analyzing employee absenteeism. Employees may be absent due to illness, accidents, or burnout. The latter has recently received wide attention since various studies have shown that your most highly motivated employees are particularly sensitive to it. Hence, by using analytics, it now becomes possible to adequately understand the drivers of employee absenteeism and act upon this understanding before the problems start to occur. Employee absenteeism can be tackled using both classification (employee is absent or not) or regression (number of days absent) techniques. Both can also be combined to determine the expected number of days absent (EDA). More specifically, a classification model can predict the probability that an employee will be absent (PA), e.g., during the next 12 months. A regression model can then be used to predict the number of days absent for those employees that are absent, or in other words, days absent given absence (DAGA). The expected number of days absent (EDA) can then be calculated as follows:
EDA=PA·DAGA+(1-PA)·0=PA·DAGA
Besides predictive analytics, social network analytics can also be used for HR purposes. Understanding, modeling, and measuring your employee network should be a key ingredient to your strategic HR decisions. As already noted by other researchers, a well-designed employee network essentially comprises the social capital of a firm referring to all the assets or resources than can be mobilized through it.
The key question to answer first is how to build an employee network and leverage it when making HR decisions. Although the nodes are obviously the employees, the links are far less intuitive to define. These should be established based upon two sources of information: communication patterns (e.g., emails, Skype calls) and joint project allocations. Obviously, there is a strong correlation between both, but state-of-the-art analytical techniques are nowadays perfectly capable of filtering this out and determining the optimal blend of information. One way to quantify the links is by using the RFM framework: Recency (when was the most recent email exchange/joint project allocation?), Frequency (frequency of emails or joint project allocations), and Monetary (average size of emails, or average man-months of joint project allocations). It is important to commit to anonymized analysis and respect privacy at all times. In other words, emails can only be analyzed in terms of sender and receiver (but not in terms of content), and the necessary disclosure agreements should be properly agreed upon between the stakeholders involved. The constructed network can then be used for various HR activities.
As with any analytical exercise or disruptive technology, it is important to start small but think big. Hence, you can start by analyzing your employee network in terms of descriptive social network analytics. An example of this could be identifying the key nodes (e.g., hubs) in the network. A next obvious step would then be to leverage it for predictive purposes whereby network information is used to predict employee churn or engagement, or even for recruitment purposes. An even more ambitious goal would be to directly link the employee network to the customer network. This could reveal new serendipitous insights into the relationship between employee and customer drain, which would allow a firm to create an unprecedented competitive advantage.
These are just a few applications of HR analytics, and many others can be thought of. Examples are understanding workforce collaboration patterns using social network techniques, job recruitment based upon intelligent recommender systems, career path analysis using sequence rules, and talent forecasting. Despite its potential, not many successful applications of HR analytics have been reported yet in the industry. This can be attributed to a variety of reasons.
First of all, HR is still struggling with the perception of being less strategically important than—for instance—risk management, marketing, and logistics. Hence, it is usually the last in line to benefit from new technologies such as big data and analytics. This is unfortunate, given the tremendous potential of both for improving the HR function.
A next barrier, which also applies to the other analytics applications, is related to the skills needed to work with both technologies. The job profile aimed for is the one of a data scientist. In the industry, there are strong misconceptions and disagreements about what constitutes a good data scientist. A data scientist should possess a multidisciplinary mix of skills: quantitative modeling (e.g., statistics), ICT (e.g., programming), communication and visualization, business understanding, and creativity. It is quite obvious that this is a unique skill set, and not many universities worldwide are offering educational programs for data scientists yet. This explains why there is currently a huge international shortage for this job profile. Some organizations are setting up in-house training and coaching initiatives to turn some of their employees into data scientists. Others are considering outsourcing as a possible solution. Despite the short-term benefits of the latter, it should be approached with a clear strategic vision and critical reflection with awareness of all risks involved.
Another notable issue is the presence of employee unions. In strongly unionized countries, companies have to provide clear justifications about how they intend to utilize big data and analytics to study their workforce behavior. It is in our human nature to approach any new technology with a certain degree of anxiety; hence, it can be anticipated that unions might also be skeptical. To remedy this, it is of key importance to introduce both technologies as facilitators or opportunities rather than a looming danger or threat, and clearly illustrate how working conditions and/or employee satisfaction can be improved as a result of their usage.