Case 01
Reliable Workforce Scheduling Under Uncertainty
How can staffing rules become a decision-support system that remains honest when capacity is insufficient?
- Coverage
- Qualifications
- Leave
- Rest rules
- Soft constraints
- Capacity limits
Analytical build
Developed nurse and patient-services scheduling models using MILP/Python. Converted coverage, qualification, leave and rest rules into optimisation constraints, calibrated objective weights using operational data, and supported benchmarking and validation. Soft constraints and stress scenarios allowed the models to distinguish a workable schedule from a genuine capacity shortfall.
Delivery
Coordinated task ownership across a seven-person team, consolidated fragmented contributions and delivered the final report and presentation. Key lesson: unclear ownership becomes an integration and deadline risk.
Against the historical baseline, matched leave conditions produced 42.31% lower night-shift deviation, 51.92% lower weekend-shift deviation and 57.69% lower monthly average overtime. High-severity infeasible cases were retained as evidence of genuine capacity limits. These are model-comparison results, not measured improvements from a live rollout.
Records + rules
Six months of shift, qualification and scheduling data.
Two systems
Nurse MILP/Python and patient-services scheduling.
Stress + audit
Constraint checks, exhaustive small-instance comparison and high-severity scenarios.
Capacity made visible
A tested prototype that reports when available resources cannot meet demand.
