By the time an employee submits their resignation, the decision to leave was made six to twelve weeks earlier. The signals were there — a drop in engagement survey scores, a slowdown in goal progress updates, a pattern of using up accumulated leave, a reduction in project volunteering. Without a system that aggregates and interprets these signals continuously, HR teams are always reacting to departures they could have prevented. AI attrition prediction changes the timeline by surfacing risk before the resignation letter arrives.
The Signals AI Monitors for Attrition Risk
- Compensation ratio drift — when an employee's pay falls below 85% of market midpoint, attrition risk increases sharply
- Time since last promotion or lateral move — tenure in role beyond 24 months without growth correlates with flight risk
- Engagement survey score trajectory — a downward trend over two consecutive quarters is a leading indicator
- Goal completion rate changes — a drop from consistently high to inconsistent signals disengagement
- Manager quality score — employees rated below 3.0 in engagement surveys for their manager are 2x more likely to leave
- Leave utilization pattern — sudden heavy leave usage or the opposite (leave avoidance) can both signal plans to exit
Building the Retention Intervention Playbook
Knowing someone is at risk is only useful if you have a clear intervention path. For each risk category, define the standard intervention. Compensation risk: comp review trigger with finance approval fast-track. Growth risk: development conversation + visible opportunity within 30 days. Manager risk: private manager coaching or reassignment evaluation. The intervention should happen within two weeks of the risk flag — the longer the delay, the lower the success rate. Space HR's attrition prediction module surfaces risk by employee and by team, with recommended next steps for each flag. See how attrition prediction works in your data.
Read the full guide: HR Data Analytics Guide.