Modern mathematical models are producing powerful insight into the causes of employee turnover, and the most effective ways to combat it.
People used to believe that AI and other computer models would never be able to offer true insight into human behavior and, by extension, the human mind. Yet today, the quest to better understand the causes of employee turnover and other business-relevant aspects of a worker's personal experience is more rapidly being pursued through the lens of big data analysis.
Modern statistical and computer models have fundamentally changed how the HR department approaches oversight and self-improvement, always in the direction of accuracy. In that spirit, ADP Research Institute® (ADP RI) details the results of a new analytical model in the Revelations from Workforce Turnover. A Closer Look Through Predictive Analytics report. The report is built solidly on a model that pulls information from the anonymized payroll data of 41,000 companies and a total of 12.5 million employees.
It turns out that the most insightful points about the highly personal, deeply human question of why people voluntarily leave their job can now come from an inhuman mathematical model. For HR, these models offer not just all-new tools to be used, but all-new lessons to be learned.
What Models Reveal About the Causes of Employee Turnover
The analysis reveals an array of factors that influence the decision to leave, but not one of them should be particularly surprising. The most important issues influencing a worker's wish to leave is their level of compensation and/or seniority. This is followed by equally logical concerns like overtime work and pay, commute time and the level of perceived tenure within the organization relative to experience.
Employee demographics can also play a huge role in determining their behavior. Millennial employees, for instance, are not only more amenable to switching jobs but much more concerned with issues like work-life balance. Thus, commute time and burdensome overtime work may influence the behavior of millennial employees more powerfully than that of their coworkers.
Note that these factors all influence the chance of someone leaving an employer, rather than increasing that chance; if addressed properly, these are the very same attributes that can most powerfully create loyalty within a workforce and reduce turnover rate.
While ADP RI's analytical model wasn't the first to discover any one of these factors, its unbiased approach to analyzing the data available has produced new insights into the complex patterns of these concerns that ultimately determine whether a worker leaves or not. The report explains that "because the model is always 'learning,' current data can be applied to predicting future probabilities" in ways that traditional HR analytical techniques cannot.
Understanding Turnover Allows Control of Turnover
Ultimately, a model can only point out an issue; it's up to management to actually take action. Just because you know that commute times are an issue, doesn't mean you'll know quite what to do about it. Perhaps the solution is to allow increased telecommuting, offer a shuttle service, increase pay — or perhaps the solution is to do nothing.
Even if a particular driver of voluntary turnover must be endured, it's still good to know what that driver is and how it's affecting your business. Indeed, as ADP RI points out, understanding something allows you to do a lot more than just destroy it: "If attrition is the goal ... now you know how to make that happen faster."
Once a model has identified the most important factors affecting voluntary turnover at your business, it's possible to take action. For instance, the ADP RI report found that for the model example "an employee's tenure relative to the overall experience levels" accounted for 29.4 percent of the forces contributing to turnover. Knowing this, the organization can then take the most appropriate actions — in this case, a review of promotion practices might be in order.
In the same vein, Forbes recently noted that "high voluntary turnover of employees within the first 30–90 days can signify that there may be interpersonal conflict with new employees and their management or coworkers, or they simply could not align to the company culture."
The goal, then, is first to diagnose, then to act. What ADP RI and others are showing is that analytical models are the new, premiere diagnostic tools.
Models Can Do Amazing Things If You Feed Them Enough Data
Step one in building this sort of model for yourself, or in just making use of one that already exists, is to generate the data it needs to work. This means both investing in payroll and employee management systems that track more nuanced forms of employee data over time, as well as one-off measures like business-wide surveys. Once your data exists in a curated form, then a machine learning model can compare it against masses of historical data — and that's where the fun begins.
Stay up-to-date on the latest workforce trends and insights for HR leaders: subscribe to our monthly e-newsletter.