Background
Your team has been hired as an optimization group at HorizonLogistics, one of the largest last-mile delivery companies in the region. Each team is assigned a specific county and must optimize its daily delivery routes. HorizonLogistics moves everything from medical supplies to retail goods, serving dozens of customers daily with a mixed fleet of vehicles.
Each county has a Regional Operations Manager — someone who has been running routes in that territory for years. They are available to answer questions and review proposed solutions. Critically, the manager knows the territory intimately but thinks in business terms, not optimization terms. They are busy, practical, and must be treated as you would a real business stakeholder.
Objective
Minimize total daily delivery cost while ensuring every delivery is completed within its specified time window.
| Component | Basis |
|---|---|
| Fixed vehicle cost | Per vehicle used, per day |
| Variable cost | Per mile driven |
| Driver wages | Based on total route time |
| Late penalties | Applied to time-window violations |
Problem Structure
This is a Vehicle Routing Problem with time windows and capacity constraints (CVRPTW). A valid solution must:
- Assign each delivery stop to exactly one vehicle
- Define a route — an ordered sequence of stops — for each vehicle used
- Respect every vehicle's capacity limit
- Deliver within each customer's time window
- Start and end every route at the depot
- Satisfy all constraints discovered through stakeholder interaction
Any optimization method is permitted: integer programming (PuLP, Gurobi, Excel Solver), heuristics, or a hybrid. The approach must be documented and reproducible.
Materials Provided
1. Excel Workbook (Team_XX_Data.xlsx)
All quantitative parameters for the assigned county:
- Delivery Stops — customer names, addresses, demands, time windows, service times
- Vehicle Fleet — vehicle types, capacities, costs
- Depot — location and operating hours
- Distance Matrix — distances between all locations (miles)
- Travel Time Matrix — travel times between all locations (minutes)
- Cost Parameters — fuel costs, driver wages, penalties
2. HorizonLogistics Stakeholder Portal
A custom-built web application giving each team private access to their own AI-powered Regional Operations Manager. Every county has a distinct manager with their own personality, communication style, and set of hidden operational constraints.
The portal was active for the duration of the case and is now closed.
Stakeholder Interaction
The portal supports two modes of engagement, both of which count against the team's communication budget:
- Ask questions — Chat directly with the Operations Manager about the territory, the customers, the fleet, or anything else. The manager responds in character and will not simply hand over a list of requirements.
- Submit a solution — Present a candidate route plan for operational review. The manager walks the plan against their knowledge of the territory and flags the first problem they find.
Deliverables
1. Optimization Model (code or Excel file)
- Full formulation with decision variables, objective function, and all constraints
- Must be runnable and reproducible
2. Solution Report (max 5 pages, PDF or Word)
- Executive summary of the recommended routes
- Total cost breakdown
- Route assignments — which vehicle visits which stops, in what order
- Schedule — departure times and arrival times at each stop
- Discussion of constraints discovered through stakeholder interaction
- How the model was adapted in response to new information
- Key trade-offs or sensitivity analysis
3. Stakeholder Interaction Log
Automatically recorded in the portal — no separate submission required.
Grading
| Criterion | Weight | Question |
|---|---|---|
| Feasibility | 40% | Does the solution satisfy all constraints — both in the data and discovered through interaction? |
| Optimality | 35% | How close is total cost to the best achievable solution given all constraints? |
| Methodology | 25% | Is the approach well-formulated, clearly documented, and reproducible? |
Pedagogical Design
Case 03 was built around a deliberate gap between the data students are given and the problem they actually need to solve. Textbook optimization problems arrive fully specified; real ones never do. The constraints that matter most are often the ones nobody wrote down.
Embedding those constraints in a conversational stakeholder forces students to practice the part of the job that formulation exercises skip: figuring out what question to ask, recognizing when an answer implies a constraint, and revising a model that was already "finished."
Each of the 18 counties had a distinct problem instance and a distinct manager, so no two teams could share findings. The communication budget made stakeholder time a scarce resource to be allocated, not an unlimited oracle to be queried. Iteration was the expected path — discovering a new constraint late and reworking the model was treated as evidence of good process, not failure.
Timeline
| Milestone | Date |
|---|---|
| Case released | April 11, 2026 |
| Submission deadline | April 27, 2026 |