It always comes as a surprise, doesn’t it?
One day a vital employee at your company is doing their job, seemingly happy as a clam. The next they’re turning in their two weeks notice because they found something else. Now your recruiting team has to scramble to fill the role as quickly as possible.
It’s a situation no small or midsize business (SMB) wants to find themselves in, which is why doing an employee flight risk assessment is beneficial to weathering the storms of worker turnover.
In this article, we’ll explain why assessing employee flight risk is more important than ever, and how even the smallest companies can begin to do it.
To help, we’ll look at best practices cited in Gartner’s report, “Use Data Science to Address Employee Flight Risk” (content available to Gartner clients).
Here’s what we’ll cover:
Why Predicting Who’s Leaving Matters More Than Ever
The 3 Steps of Assessing Employee Flight Risk
#1: Historical Data Collection
#2: Model Design
#3: Model Testing
Why Predicting Who’s Leaving Matters More Than Ever
In 2017, the U.S. labor economy vastly favors the employee over the employer, according to data from the Bureau of Labor Statistics. The number of nonfarm job openings is over 5.5 million—one of the highest numbers ever recorded since job openings started being tracking in 2001—while the unemployment rate sits at a decade-low 4.7 percent.
Sure enough, employees are quitting their jobs now more than ever, and employers have to cover the high cost to replace them. This can range from 16 percent of annual salary for high-turnover, entry-level jobs, all the way up to 213 percent of annual salary for specialized executive positions, according to a study by think tank Center for American Progress.
According to Gartner Analysts Helen Poitevin and Alexander Linden, doing a proper employee flight risk assessment can help avoid some of that cost by giving businesses better indicators of which workers are at high risk of leaving, which in turn yields many benefits:
“Managing employee flight risk can be linked to significant cost savings, in terms of avoiding the cost of replacing employees who have left the organization, as well as loss of productivity. By appropriately managing attrition with the right management practices to encourage retention, employee engagement should increase. High engagement levels correlate strongly with increased business performance.”
Source: “Use Data Science to Address Employee Flight Risk”
With workers more likely to leave than ever, it becomes important to not only understand which workers are in danger of exiting, but to also address organizational factors that are pushing people out the door.
The Three Steps of Assessing Employee Flight Risk
Historically, leveraging flight risk indicators to better manage employee churn has been the domain of massive companies like Hewlett-Packard (HP), but that shouldn’t stop SMBs from doing their own analysis.
Discovering the factors that predict which workers are about to leave involves three steps: data collection, model design and model testing.
#1: Data Collection
This is (hopefully) a no-brainer: To analyze a data set for trends, you need a large enough data set.
Whether you’re using spreadsheets or a formal core human resources (HR) system, you should be tracking basic employee information (name, job title, department), termination status and a bunch of other attributes related to all of your individual workers, past and present, in order to identify commonalities among those who leave.
Gartner offers some of the following attributes as examples to track:
- Time in position
- Time since last promotion
- Commute time
- Compensation level
- Performance review rating scores
- Peer feedback
HR data is the obvious place to start, but in some instances, communication data or data from employee monitoring tools (such as hours logged in a time and attendance system) may provide better insights.
You may have your own unique attributes you want to track as well. Regardless, give it time.
The more attributes you track, the more data you accumulate and the more employee attrition cases you can observe, the better your employee flight risk predictors will become.
Just make sure that when you start to analyze the data in step two, you focus on a fairly homogenous subset of employees (e.g., just salespeople or just factory workers).
#2: Model Design
Once you have a large enough data set to work with, it’s time to leverage statistical modeling techniques to uncover which attributes historically occur most often leading up to an employee leaving the company.
There are three types of tools you can use on your data to do this, each with pros and cons:
Tool #1: Excel
Though it’s the most manually intensive tool on this list, it is possible to apply a logistic regression model to your data to uncover employee flight risk indicators using Excel. If you’re not a data scientist, there are tutorials such as this one to guide you.
Pros: Excel is free, and your team is likely already familiar with it.
Cons: It’s a manually intensive and therefore more error prone method.
Tool #2: Predictive Analytics Tools
Predictive analytics tools like SAP BusinessObjects can take a data set and apply regression models to it automatically, adjusting for best fit. These platforms can even use survival analysis techniques to further predict when an employee is likely to leave.
Pros: Can do a more thorough and automated analysis using multiple sources of data.
Cons: Requires purchasing a standalone tool and more statistical know-how.
Tool #3: Embedded Flight Risk Predictors
Some talent management suites such as UltiPro have embedded flight risk predictors that use proprietary algorithms with your HR data to predict an employee’s likelihood to stay or leave.
Pros: Dedicated to employee flight risk management and already found in talent management suites.
Cons: Talent management suites represent a significant investment. Employees may still leave even with a high retention score.
Whatever tool you choose, the result should be a list of employees predicted to leave along with their likelihood to do so based on affecting factors.
#3: Model Testing
This last step is perhaps the most important. No data analysis is perfect, so before you take your employee flight risk predictors and apply them to your current workforce, it’s always a good idea to test the model you created against a historical data set to determine how accurate it is.
Software Advice’s Daniel Harris explains why predictive analytics can be risky when mishandled here.
Applying the model to your historical data set, you should see: 1) how many actual employee attrition cases it was able to predict, 2) how many were missed and 3) how many false positives were flagged.
Gartner recommends aiming for a success rate of 60 to 70 percent. So, if you have a sample set of 500 employees where 80 of them left the company, for example, your model should be able to correctly identify at least 48 of them.
If your success rate comes in below 60 percent, you may need to adjust the weights of your flight risk predictors or grow your data set further.
Once you’ve achieved a desirable success rate with your model, it’s time to implement it with live data to show users which employees might leave based on past behavior and data patterns.
When properly implemented, the ability to assess employee flight risk can become a powerful tool to identify at-risk workers so management can intervene and make necessary retention adjustments. But, as Uncle Ben once said to Spider-Man: “With great power comes great responsibility.”
Gartner recommends some final considerations for using employee flight risk assessments in an accountable way:
- Determine a purpose. Why are you measuring flight risk? Be specific. It could be a cost-saving measure or a way to influence employee engagement, but it shouldn’t be just because you want to analyze what your quitters look like.
- Communicate your plan to all… According to Gartner, “clear and consistent communication is critical to the successful deployment of a flight risk indicator.” Explain what data is being collected and how it’s being used to employees, managers and other HR professionals.
- …But limit access to a few. This information can be easily exploited. Managers can use flight risk predictors to push workers out, while employees could use it to negotiate their own conditions. Be judicious in who can use this data and how.
- Be ethical with any follow-up. Ultimately, this is a data point and should not replace the emotional intelligence and interpersonal skills of proper people management. Be specific with management about what they should do with flight risk information and how they should speak with high flight risk employees.
- Update your model as needed. The labor market is constantly changing, as are the roles at your company and the determinants of their success and satisfaction. If your model’s reliability starts to slip, update it.
If you have questions about any tools we’ve mentioned or where to go next, email me at firstname.lastname@example.org.