Voluntary turnover is the most expensive HR problem that is most predictable in advance—yet most organizations track it retrospectively, after resignations are submitted. These nine metrics turn turnover from a lagging indicator into a leading signal, giving HR teams and managers the advance warning needed to intervene before departure decisions are made.

1. Manager 1:1 Meeting Frequency Decline

Employees who are considering leaving often reduce engagement with their managers before resigning. Calendar data analysis that tracks the frequency of manager-direct report 1:1 meetings shows a consistent pattern: a 30%+ decline in meeting frequency over a 4-week period correlates with elevated turnover risk. The signal is most significant when the decline is unilateral—initiated by the employee, not the manager. Identify employees with declining 1:1 frequency and flag for proactive manager outreach within 5 business days of the signal triggering.

2. Performance Review Submission Latency

Employees who are disengaging from their organization become less compliant with administrative requirements. Performance review submission latency—the time between the submission deadline and when the employee actually submits—shows a measurable increase among employees in pre-departure stages. A 3-day or greater increase in submission latency compared to the employee’s historical baseline is a risk signal, particularly when combined with other indicators on this list.

3. PTO Utilization Pattern Anomalies

Two PTO patterns correlate with elevated turnover risk: (1) a sudden spike in PTO requests after a period of low utilization—often associated with job interview scheduling, (2) a prolonged period of unused PTO accumulation in an employee who historically takes regular PTO—potentially indicating disengagement or a known departure date being worked around. Neither pattern is conclusive alone; both warrant a check-in conversation when combined with other risk signals.

4. Compensation Percentile Position at Tenure Milestones

Employees whose compensation falls below the 40th percentile of market at tenure milestones (18 months, 3 years, 5 years) show materially higher voluntary turnover rates than employees at or above market. The financial signal: TalentEdge reduced agency fees by $312K and achieved 207% ROI partly through reducing preventable turnover driven by compensation compression—the employees who left were below market, not dissatisfied with the work. Monitor compensation percentile position at each tenure milestone and flag employees below the 40th percentile for a compensation review conversation before they reach the point of active job searching.

5. Internal Mobility Application Activity (or Absence)

Employees who are engaged with their organization apply for internal mobility opportunities when they are ready to grow. Employees who stop applying internally—particularly high performers who previously expressed interest in growth—are signaling either that they have accepted that internal mobility will not happen or that they are focused on external options. Track internal application activity relative to historical patterns and flag employees with above-average performance ratings who have not applied for any internal opportunities in 18+ months.

6. Project Workload and Deadline Pattern Changes

Employees preparing to leave often begin winding down project commitments: declining new project assignments, completing existing projects without taking on successors, and reducing involvement in long-horizon planning discussions. Project management data (task assignment rates, planning meeting attendance, successor designation behavior) shows these patterns 8–12 weeks before resignation in many cases. This signal is strongest for senior individual contributors and managers whose departure requires extended transition planning.

7. Engagement Survey Response and Score Decline

Engagement survey scores that decline 15+ points from an employee’s individual baseline—not just the team or company average—are a retention warning. The individual baseline comparison matters: an employee who consistently scores in the 70s who drops to the 50s is at higher risk than an employee who has always scored in the 50s. Survey response abstention (not responding to a survey the employee has historically completed) is also a risk signal, as disengaged employees disengage from feedback mechanisms before they disengage from the role formally.

8. Benefits Utilization Decline

Employees who plan to leave in the near term reduce investment in employer-provided benefits—particularly benefits with enrollment costs, long-term commitments, or use-it-or-lose-it dynamics. A measurable decline in healthcare utilization, 401(k) contribution changes that reduce employer match capture, and declining use of professional development benefits (tuition reimbursement, conference attendance) can indicate a shorter expected tenure horizon. Benefits utilization data requires careful handling under GDPR and HIPAA—access must be strictly controlled via RBAC and use must be disclosed in employment data processing notices.

9. LinkedIn Profile Update Frequency

Significant LinkedIn profile updates—adding skills, updating job description language, enabling “open to work” signals (even privately)—correlate with active job searching. While this data is not directly accessible to employers, third-party HR analytics platforms that monitor public LinkedIn signals (frequency of profile changes, new skill endorsements, connection growth patterns) can surface this as a risk indicator within compliant data use frameworks. This signal has the shortest lead time—LinkedIn activity spikes typically occur 2–4 weeks before resignation, making it a confirmation signal rather than an early warning indicator.

Frequently Asked Questions

How early can HR analytics predict employee turnover?

Well-constructed predictive models using the metrics described here identify turnover risk 30–90 days before resignation with 70–80% accuracy in validated implementations. The leading indicators—manager 1:1 frequency decline, performance submission latency, PTO utilization spike—typically show anomaly patterns 6–8 weeks before a resignation. This window is sufficient for intervention: a proactive career conversation with a high-risk employee at week 6 before their likely resignation date produces materially better retention outcomes than the same conversation at week 1.

What data sources are required to build HR turnover prediction analytics?

The minimum viable data set requires: HRIS data (tenure, role, compensation relative to market, performance ratings, promotion history), ATS data (time-to-fill for the employee’s role, sourcing channel), and engagement survey results. Supplementary data that improves model accuracy: calendar data (meeting frequency, manager 1:1 cadence), task management data (project load, deadline patterns), and exit interview data from historical voluntary departures to validate which factors most strongly predicted attrition in your specific organization.

How do you act on turnover prediction signals without creating bias or privacy issues?

Three principles: (1) use prediction signals to trigger conversations, not decisions—a high-risk score should prompt a manager check-in, not a performance action, (2) ensure prediction models are tested for demographic bias—if the model flags one demographic group at disproportionately higher rates, investigate the model inputs before acting, (3) limit access to individual prediction scores through RBAC—HR business partners see their population’s scores; managers see their direct reports’ scores; no one sees scores outside their purview. Aggregate trend data (department-level risk) is appropriate for broader leadership visibility; individual scores require strict access controls.