The average cost of replacing an employee is 50 to 200 percent of their annual salary — and in most organizations, the vast majority of voluntary departures were preventable. The challenge isn't that HR teams don't care about retention; it's that by the time the resignation letter arrives, the decision to leave was made six to twelve weeks earlier. Employee attrition prediction shifts the timeline by identifying flight risk before the employee starts looking — while there's still time to act.
What Employee Attrition Prediction Actually Is
Attrition prediction uses machine learning models to analyze patterns in your existing HR data — compensation, tenure, engagement scores, performance ratings, leave patterns, manager quality data — and produce a risk score for each employee. The model isn't predicting the future from nothing; it's identifying which combinations of current conditions have historically preceded voluntary departure in your organization. The more HR data your model has access to, the more accurate and specific it becomes.
This is different from a gut feeling or a manager's instinct. A manager might notice their best engineer seems disengaged. An attrition model notices that the same engineer's compensation ratio dropped to 0.82 of market midpoint six months ago, their last performance rating was delayed, their engagement survey score on "growth opportunities" dropped two points, and they used two weeks of accumulated leave last month — a pattern that in your historical data preceded departure 73% of the time within 90 days. That specificity is what makes prediction actionable.
The Seven Risk Signals AI Monitors
- Compensation ratio decay: When an employee's salary falls below 85% of the market midpoint for their role and location, attrition probability increases significantly. This is the single highest-predictive signal for most organizations.
- Tenure in role without progression: Employees who have been in the same role for more than 24 months without a promotion, title change, or significant scope increase are more likely to seek growth externally.
- Engagement survey trajectory: A consistent downward trend over two consecutive quarters — especially on dimensions like "growth opportunities" and "manager effectiveness" — is a leading indicator of departure planning.
- Manager quality score: Employees who rate their manager below 3.0 out of 5.0 on engagement surveys are statistically twice as likely to leave within 12 months as those who rate their manager 4.0 or above.
- Performance rating change: A drop from a previous high rating to average — or a sustained average rating with no development conversation on record — correlates with flight risk in high-performing employees.
- Leave utilization shift: Both sudden heavy leave usage (spending down accumulated balance) and abrupt leave avoidance (employees who normally take regular leave going weeks without any) can signal that someone is preparing to exit.
- Internal mobility absence: Employees who have applied for internal transfers and been declined, or who work in organizations with low internal mobility rates, show elevated attrition risk compared to those who see viable growth paths inside the company.
Building a Retention Intervention Playbook
Prediction without a playbook produces anxiety, not action. When an employee surfaces as high-risk, the response needs to be immediate and assigned — not a conversation that happens "when there's time." Define standard interventions for each risk category before the first risk flag appears:
- Compensation risk: Trigger an off-cycle compensation review within 10 business days. Fast-track finance approval. The goal is to close the gap before the employee has had another offer in hand for long enough to get emotionally committed to leaving.
- Growth risk: The manager schedules a development conversation within 5 business days. HR identifies one concrete internal opportunity — a project lead, a stretch assignment, a mentoring relationship — that can be offered with a realistic timeline.
- Manager risk: HR initiates a private manager coaching conversation. If the risk is severe and the relationship has significantly broken down, quietly evaluate whether a lateral transfer to a different manager makes sense before the employee self-selects out.
- Engagement risk (broad): The manager has a direct, non-evaluative check-in — "I want to make sure you have what you need here" — within 2 business days of the flag. The conversation should feel organic, not like a retention intervention script.
What to Measure: Is Your Retention Program Working?
Three metrics tell you whether attrition prediction is producing real retention impact:
- Intervention success rate: Of employees flagged as high-risk who received an intervention, what percentage are still employed 6 months later? Benchmark: a well-executed retention program should retain 40-60% of employees who would otherwise have departed.
- Regretted attrition rate trend: Track voluntary departures of employees rated "meeting expectations" or above — these are the departures that cost the most. Is this number declining quarter over quarter?
- Model accuracy: Track what percentage of actual voluntary departures in the past 90 days were flagged by the model as high-risk in the preceding 60 days. Continuously retrain your model on new departure data to improve accuracy over time.
Space HR's attrition prediction module monitors all seven risk signals continuously and surfaces intervention recommendations directly in your HR dashboard — no separate analytics tool required. Book a demo to see how it works with your data.