You may be surprised by the signals of employee attrition in your company. You’ve boosted pay to competitive levels, but turnover continues. You have worker mobility within the company, but employees leave their new roles. You replace a famously bad manager, but the team still bleeds staff.

Identifying the signs of impending turnover at your company requires thorough (but privacy-aware) data collection, careful consolidation of the information, and review by a combination of teams. Plus, for each company, “the signals may be different,” said Sudha Solayappan, director of digital HR at Intuitive Surgical.

Signals of attrition can run deeper than the company level, said Andrea Haan, senior VP of global talent at market-research company IRI. “You see different drivers of retention and engagement based on the talent segment that you’re studying and researching.” What may be a sign in your business-development department may be totally different than a sign in your finance department.

Solayappan and Haan took part in a From Day One webinar, titled “The Role of Analytics in Boosting Your Retention Strategy,” in which five leaders in people operations and analytics discussed the ways employers can predict and mitigate employee turnover using data unique to their organizations and teams.

Recognizing the Unique Signals of Attrition

The panelists have identified the signs of coming turnover in their companies–and all of them are unique. Vahed Qazvinian, co-founder of Praisidio, a company that makes talent-retention software, said one of his clients figured out that tardiness can be a signal that turnover is coming. Not necessarily for those who are late, but for the new hires around them. When new employees see their colleagues disengaging, they’re more likely to feel disengaged themselves.

Another of his clients found that too much movement is not a good sign. “We would ask [managers] if they knew how many times their employees have changed managers in the past, and they would have no idea. That’s a really important signal for attrition. If you’re not connected to your manager, that creates a lot of risks.”

At Intuitive Surgical, employees tend to follow peers with similar skills, often because of the ways competitors are recruiting, Solayappan said. “When you see people leaving from engineering roles to go to your competitors, that means you need to start focusing on the employees who have similar skills that are present in those teams.” She called this a “pull factor,” which is a reason a worker might be attracted to another company. Compare that to a “push factor,” or an internal factor that drives workers out of your company. “The pull factor is sometimes stronger,” she said.

Yet there are also signals that your employees are likely to stay, and those are just as important to diagnose. “An employee expressing interest in growing with the company, and learning new things, is a really big indicator of someone’s tendency to stay,” said Kathleen Brenk, who is the senior director of national HR and talent acquisition at Marvin, which manufactures windows and doors. “We can track that through analytics as people are moving about through the organization, whether that’s laterally or through growth opportunities.”

Pooling Data for Maximum Knowledge

A clear understanding of employee turnover requires the combination of many data sources, said Laurie Shumake, VP of global talent management and HR at IT-systems company TD SYNNEX. However, she acknowledged that pooling data is easier said than done.

“There is a tension that comes pretty naturally with any reporting. That’s an area a lot of organizations grapple with, particularly as we’re centralizing our data into one system or into as few systems and platforms as possible,” she said. “What we’re talking about is something that’s differently layered and may take disparate data sets to be put into something like a data lake in order to reconcile. Then it’s about making interesting hypotheses.”

Pooled data from across the company is precisely what describes employee experience, said Haan. Combining demographic data with productivity with absenteeism with work load with PTO taken, for example, can tell you a great deal about working conditions.

She believes the entire organization should be thinking about retention, “whether you’re an HR business partner or you’re leading compensation or the benefits team. Regardless of the service, that’s something I expect to see disciplines maturing around, finding ways of understanding how their service impacts stickiness within the employee population.”

The hard work of combining employee data and working across teams is worth it. Qazvinian said this is how companies will find richer information about employee retention. “The CTO or chief people officer might be interested in measuring culture with respect to connection, growth, workload, burnout, and so on, whereas the DEI executive may look at it from a different angle, measuring the health of the organization with respect to diversity.”

The more you can demonstrate the relevance of this information across the company, the more likely you are to get the attention of the C-suite. “Pair that with turnover data and find pockets of spots where the turnover is not ideal, and that is connected to the business strategy,” Haan said. “The pairing of that data and information to pain points that business leaders are having is a great way to get resources and support for the work.”

Respecting Employee Privacy

If you’re using data to predict the likelihood an employee will leave the company, privacy matters.

“[We have] a responsibility to understand the stressors that employees are having,” said Solayappan. “There is an important distinction between employee listening and employee surveillance. You cannot one cannot go into the surveillance area.”

To cultivate trust, companies can start by giving employees access to their own data–like productivity, PTO usage, or time spent in meetings–to help them understand how they’re working, then letting the employees elect to share it with management. “Now that you’re comfortable with this data, you can opt in as an employee to say, ‘Yes, this is something I brought my management to understand, so that they are able to save the macro trends that are happening here,’” said Solayappan.

Employees also need to know that once the data reaches managers, it won’t be used against them. Managers should be trained on how they can and cannot use this information, “making sure there’s not a personal conversation, of ‘Hey, I don’t see you working these hours, but your team is working all of these hours.’” That’s when it becomes surveillance, said Solayappan.

Similarly, managers deserve reassurance that data collection intenditied to identify attrition won’t come back to bite them. If employers want managers to self-report data that reflects attrition risk on their teams, “you have to have a culture that makes that okay, where supervisors and managers understand attrition is natural, a portion of attrition is really healthy for your organization to have an influx of new brainpower and skill and thought diversity,” Brenk said.

Shumake added, “you’ve got to surround managers with resources to distill all of this information into a couple of things they can focus on, then make sure they have the right resources to drive a team discussion.”

This can happen only in a “high-trust environment,” said Brenk. “How have you as an organization teed up the data? How are you going to use it? Build that trust with supervisors and managers and employees to say, ‘We want the best for you, and this is how we intend to use this data.’”

Emily McCrary-Ruiz-Esparza is a freelance writer based in Richmond, Va. She writes about the workplace, DEI, hiring, and issues faced by women. Her work has appeared in the Washington Post, Fast Company, and Food Technology, among others.