The New Retention Model: Talent Science and Predictive Analytics

Jill Strange, Director of Human Capital Management & Behavioral Science, Infor

Jill Strange, Director of Human Capital Management & Behavioral Science, Infor

The terms “talent science” and “predictive analytics” conjure up different meanings for different people. If there existed a Business Buzzword Dictionary (not a real book - yet), talent science would generally be defined as the practice of using mathematic principles, algorithms, and Big Data to more objectively manage the identification, hiring, and longterm retention of the workforce. Predictive analytics goes a bit deeper by using those same non-subjective principles to see into the future, based on events in the past. These prognostications may include inventory management, workforce scheduling, or they may be more people-centric, like recommendations on job candidates who will stay employed longer or excel in one job role versus another, and which training needs should be addressed to move an individual up the ladder of success. There has been no shortage of business discussion around these topics over the last few years, especially heating up over the last five. The phenomenon of people analytics has been plastered across TIME magazine recently, right up there on the cover like Russia’s Vladimir Putin and Facebook’s Mark Zuckerberg, which certainly indicates a mainstream presence for these analytical concepts. Could talent science be the new retention model for the 21st century?

Improving Retention, Reducing Attrition

Employee retention falls under the broad umbrella of Human Capital Management (HCM), and the past several years have revealed an increasing focus on talent science and predictive analytics in the HCM market. Large HCM juggernauts are claiming the ability to predict the likelihood of an employee leaving the company by leveraging post-hire T data collected through HR systems. Others use pre-hire data combined with existing incumbents and performance measures to predict who has the right stuff to raise the bar. Let’s first define attrition prediction systems and then shed light on some of the inherent issues associated with utilizing a system such as this within an organization.

Attrition Models: What are They?

The prediction of employee attrition within organizations is rooted within the study of customer turnover or churn–the loss of clients. Many organizations use customer churn as a key business metric and have a keen interest in predicting when and how it will occur. Stemming from this need, organizations have developed predictive models around the factors leading to customer churn along with strategies to address these factors as, or before, they occur. Much like with employee turnover, organizations make distinctions between voluntary and involuntary churn, with the distinction being churn related to controllable factors (i.e., customer leaving to Human Capital Management JULY 2015 38 HR TECH OUTLOOK With high turnover costs and increasing demands on performance, organizations have a vested interest in ensuring employees stay with the organization for the long term go with a competitor) versus uncontrollable factors (i.e., relocation, death).

Customers don’t just leave–they usually leave for a reason or combination of reasons. And much like customers, employees don’t just leave. They leave for a better organization, better job, and better pay. With high turnover costs and increasing demands on performance, organizations have a vested interest in ensuring employees stay with the organization for the long term. Given these similarities, HCM leaders can draw upon research and practice related to customer churn and begin examining the factors related to employee turnover to develop similar predictive models.

What Model is Correct?

Many predictive attrition models in the marketplace are primarily derived from data collected after the point of hire. Once an employee comes into the organization, various data are collected on the person. With enough people in the organization, predictive models can be developed and refined over time with the addition of more data, then reapplied to the incumbent workforce to determine turnover likelihood. The end result is a predictive model describing the factors most likely to result in employee turnover. These factors include variables such as time since last promotion, pay rate compared to co-workers and market. These factors are primarily influenced by data collected once the person has already been hired.

Most models neglect to take into account initial fit to the job at the point of hire, as well as other characteristics available for collection during the application process. You can choose to leverage the data generated by predictive analytics based upon behavioral, cognitive, and cultural preference data to determine a person’s fit to a role.

With an accurate and objective behavioral measure of the candidate, the next question then becomes, “How does an organization truly know who they are looking for?” This question leads us to the key Talent Science ingredient, the custom Performance Profile. Infor Talent Science creates this unique benchmark for each role within each client company, before assessing new candidates. Much like a pollster’s ability to accurately predict the outcome of a general election by sampling a relatively small pool of voters, Infor’s team of Industrial Psychologists create these benchmarks by leveraging performance indicators and behavioral data from a sample of company employees coupled with big data workforce analytics. The combination of these data points allow for the creation of an objective benchmark identifying the behavioral patterns driving performance in the role.

It is important to note that the creation of these benchmarks relies on employees from across the performance spectrum, not just top performers. To create a truly predictive model, the behaviors and performance of every incumbent in the sample must also be considered so that evaluations can be made for the behaviors that drive both high performance and low performance. As an example, knowing that high performing accountants are detail oriented is not enough to say that a company will hire better accountants if they can just pick out the detail oriented applicants. If low performing accountants also turn out to be detail oriented, then you have done nothing more than identify a common attribute of all accountants.

Talent Science Drives Predictive Analytics

A successful retention model must be able to identify what separates high and low performing employees and then apply that information to the pre-hire process. When performancebased models identifying best-fit candidates are combined with predictive analytics using post-hire variables, companies can implement processes that drive drastic reductions in turnover, but also result in long-term positive increases in performance across the organization.