Detecting hidden employee flight risk before resignations happen
Most employees do not resign suddenly.
They disengage quietly, through overtime creep, declining performance, stalled progression, management instability, and burnout, often 6–12 months before leaving.
The Silent Attrition Engine (SAE) is a privacy-preserving, explainable workforce analytics system designed to detect these early warning signs and support proactive retention decisions.
SAE identifies employees at risk of voluntary attrition using internal behavioural and organisational signals, then simulates how targeted interventions could reduce attrition and save costs.
At its core is EARS™ (Employee Attrition Risk Score), a transparent risk score that categorises employees into actionable risk tiers.
- Score range: 0–100
- Risk tiers: Low, Medium, High, Critical
- Built from explainable signals (not black-box predictions)
Named signals include:
- Overtime Creep
- Manager Shock
- Promotion Stall
- Training Deficit
- Performance Slide
- Absence Spike
- Commute Strain
Each signal reflects known precursors to disengagement and exit.
Simulates realistic retention actions:
- Reducing overtime pressure
- Increasing training access
- Improving manager stability
Result (example run):
- High/Critical risk employees reduced from 154 → 9
- Estimated avoided attrition: 145 employees
- Estimated savings: ~£1.74 million
Silent-Attrition-Engine/ ├─ data/ │ ├─ raw/ │ └─ processed/ ├─ src/ │ ├─ data_generation.py │ ├─ scoring.py │ └─ interventions.py ├─ reports/ │ ├─ Risk_Framework.md │ └─ Executive_Summary.md ├─ run_intervention_simulation.py └─ README.md
Most organisations rely on exit interviews and lagging metrics.
SAE demonstrates how workforce data can be used earlier, ethically, and strategically to:
- reduce avoidable attrition
- protect institutional knowledge
- improve employee wellbeing
- optimise retention investment
- No protected characteristics used
- Signals are operational, not personal
- Outputs are probabilistic
- Human oversight is required
Built as a portfolio project to demonstrate applied workforce analytics, business impact thinking, and decision-support design.
If you work in HR, People Analytics, Consulting, or Data, I’d love feedback.