Finance Teams Can Apply Algorithms to Aid Employee Retention

Applying machine learning to standard HR data helps spot employees poised to quit their jobs.
Tim Glowa and Eric GonzagaMay 19, 2022
Finance Teams Can Apply Algorithms to Aid Employee Retention
Photo: Getty Images

Working in tandem with fellow executives, CFOs have every reason to find the most effective employee retention measures. Advances in data analytics can provide large employers with more sophisticated insights into which workers may be poised to quit. And here’s the good news: many companies already have the data they need for this analysis. 

Machine learning that taps into the immense data set maintained by a human resources team can build a model that will accurately predict and identify employees close to leaving. Then it becomes possible to anticipate their behavior and boost management’s understanding of why they might leave or what might get them to stay. 

  Tim Glowa

Such efforts are needed. Almost one-third (29%) of workers are actively looking for a new job at a different company, according to Grant Thornton State of Work in America research conducted in January and February 2022.

Machine Learning Opportunity

HR data sets usually contain salary history, performance ratings, and disciplinary notes for every employee as far back as their hiring date. Machine learning algorithms can also work with information like whether an employee has applied for an internal job posting, whether they manage people, whether they are critical talent, and whether they have been flagged as a high-potential employee. 

Drive Business Strategy and Growth

Drive Business Strategy and Growth

Learn how NetSuite Financial Management allows you to quickly and easily model what-if scenarios and generate reports.

  Eric Gonzaga

The model can be tailored to fit a specific company’s data. But the breadth of the data that most organizations can marshal is at the heart of why this method works. 

Overall, the machine learning required is not particularly difficult compared with other artificial intelligence applications. Marketers have been doing something similar for years. Subscription-based businesses such as streaming services or mobile phone providers are good examples. These businesses embrace the idea that it’s possible to take customer data and deploy machine learning to: identify customers they may lose, discover which actions will prevent them from leaving, and persuade other customers to buy more. 

Conjoint analysis, which marketers use to understand the relative value customers place on product or service features, can help measure an individual’s sensitivity to changes in benefits and rewards. 

In working with Grant Thornton clients, we have found a machine learning model can effectively predict which individual employees plan to leave within six months or one year. In an engagement for a large federal government agency, machine learning helped accurately identify individuals at risk of quitting or taking early retirement. This was accomplished by examining multiple years of HR information system (HRIS) data. From there, we suggested proactive measures to extend those employees’ tenures. 

Employee Preference Optimization

Once a company identifies a high-performing employee who may be about to leave and decides it wants to keep this person, the next question is what to do — the effort shifts to deploying relevant retention tools. 

The goal is to understand how sensitive workers are to specific changes in the value proposition represented by compensation, benefits, and any other rewards package. A well-designed survey will not simply ask employees what they want directly. Employees typically say more of everything, which is not actionable. 

Instead, ask employees to compare the value they perceive in different packages. In one instance, we found that if a company boosted the monthly car allowance for top employees, they would perceive the value of that perk to be almost double its cost. 

More broadly, listening to employees through surveys makes it possible to analyze an array of benefits, ranging from health insurance coverage to retirement savings to vacation policy. The analysis is based on the benefit’s perceived worth as a fraction or a multiple of the actual expense. Conjoint analysis, which marketers use to understand the relative value customers place on product or service features, can help measure an individual’s sensitivity to changes in benefits and rewards. 

Employee surveys make it easier to perform effective analyses of the best benefits options for employers and employees. We can often identify a total rewards package that 70% or 80% of employees think is better than their previous package but costs the employer thousands of dollars less per person per year. Using such tools, an employer planning on boosting spending on rewards and benefits — escalating the war for talent — can be confident it is doing so effectively. 

Tim Glowa is a principal and leader of employee listening and human capital services offerings at Grant Thornton LLP. Eric Gonzaga is the national managing principal of human capital services at Grant Thornton LLP.