Beyond Forecasting: Use Predictive Analytics to Enhance Organizational Performance

When employees quit, it’s costly.

Predictive analytics can help HR with employee retention and improve an organization’s performance. Through statistical modeling, it is possible to determine which of the many possible factors most influence the workforce’s decisions to stay or quit, and then to develop a strategy and act on it.

Human resources at FinCo Management, a large financial services firm, did exactly this for its call center employees. The Fortune 500 Company believed retention was crucial to its business. FinCo built predictive turnover models for its call centers, drawing on historical personnel data from its HR systems to understand what factors were causally connected to higher or lower probabilities of quitting. The results revealed clear links and actionable insights around compensation, career development, experience levels, mobility and work/life balance as its key drivers of retention. The results helped to reinforce and refine current efforts in these areas, ultimately improving call center performance.

Exploring the Past to Predict the Future

By looking across five years of historical data, predictive statistical models were developed for call center employees—modeling their likelihoods of quitting within the following month and within the following year. The objective was not to forecast turnover rates or head counts. Instead, these turnover models were developed to study and understand which predictors led to the greatest increases and decreases in the probability of call center employees quitting, and by how much.

Human resource practitioners have long estimated that costs associated with employee turnover can range up to a substantial portion of an employee’s salary. FinCo estimated that for every employee retained, it could realize a savings of approximately 25 percent of the employee’s first-year salary through productivity gains, as well as from avoiding costs associated with the recruiting, onboarding and training of replacement hires. FinCo set out to gain insights from large amounts of archived data to help build a better place to work for its employees, while also improving its own organizational performance.

Three broad categories of data were collected and examined as predictors of turnover in its model: individual attributes, organizational factors and external influences. These categories were selected based on research from global consulting leader Mercer and from experience that linked them as the common sources for causal factors leading to employment outcomes. Individual attributes included factors about the employees’ work history such as age, tenure, career level, promotion and so on. Organizational factors consisted of the internal work environment such as organizational segment and the recent turnover rate at the call center site. External influences included characteristics about the outside market and the geography in which the employees resided, such as local area unemployment rates.

Except for external influences, all data were sourced from FinCo’s human resource information systems, where data for all employees who had worked at its call centers had been accumulating for numerous years. All of these factors were examined simultaneously. This means that inferences were isolated to each individual predictor when explaining its impact on an outcome. For example, this method allowed for inferences such as “If an employee is promoted within the past year (the predictor, X), he or she has a 30 percent lower probability of quitting within the next year (the outcome, Y)—holding all other factors constant.” Put another way, it allowed for causal inferences. In this example, if two individuals had similar qualities and experiences, and the only material difference observed was that one was promoted and the other was not, the employee who was promoted had a lower chance of quitting (30 percent less). In other words, promotions reduced an individual’s likelihood of choosing to quit.

Know Data Inside Out

To drive change with data, it is critical to know the audience and know your data. Analyses can always examine more factors and data will never be perfect, so understanding where there are potential gaps and pitfalls is crucial. Study their implications and be ready to talk about them with senior leaders. Insightful analytics and good storytelling are often the first steps in driving change, especially as organizations grow larger and become more complex. Discussing the findings, developing buy-in from leaders and decision-makers, and gaining credibility and building positive relational capital are important to successfully achieving data-driven actions and results.

Key Takeaways

• Use predictive analytics to drive organizational intelligence and performance. Uncover workforce dynamics and drivers of employment outcomes; use this to start the dialogue around truly evidence-based policy actions and practices.

• Findings are unique to your organization. Don’t cut and paste results from other organizations. What’s important to Company ABC or Competitor XYZ may be less relevant to your business and organizational culture.

• Be patient, and stick with it. It can take years to build the right metrics and analytics into a workforce management process.

• Leverage what you have and combine multiple sources of information. Start with the data you and your organization are already measuring, and look for evidence across multiple sources of information.

• Know your data, your audience and the internal landscape. Understand and navigate the internal dynamics within your organization to successfully use data for driving change. Seek out well-connected leaders within your organization to promote, facilitate and lend credibility to your findings.

  Min Park is a senior associate with Mercer’s workforce analytics and planning team. She can be reached at min.park@mercer.com

©2015. International Association for Human Resource Information Management. Used with permission.