Case Study - Route optimization and service level analysis for a cold-chain distribution fleet

Overview

FoodX Logistics runs a refrigerated distribution network from its Çayırova depot, serving 649 active branches across a wide delivery geography. The fleet consists of 10 vehicles — two small refrigerated vans (8-pallet), five medium vans (10-pallet), and three large refrigerated trucks (18-pallet) — for a total capacity of 120 pallets.

The operational problem was not fleet size. It was structure. Routes were defined as fixed clusters, and each vehicle was permanently assigned to a region. On low-demand days, vehicles ran near-empty. On peak days, the same vehicles hit capacity limits and left orders unserved. The system had no mechanism to redistribute load dynamically.

The 94-day dataset confirmed this: 307 orders could not be assigned under the current real-world constraints — not because capacity was absent, but because it was locked into the wrong configuration on the wrong day.

The Analysis Framework

The study ran four scenarios designed to isolate distinct questions:

S1 — Unconstrained Reference strips capacity and time limits to establish the theoretical ceiling of the existing route structure. It is not an operational plan; it is a benchmark.

S2 — Real Constraint Test applies FoodX's actual operational constraints to the same cluster structure. This reveals exactly where and how the current system breaks under real conditions.

S3 — Controlled Transition Model removes cluster dependency for Istanbul deliveries. Refrigerated vans handle in-city routes; out-of-city shipments are assigned based on vehicle type from the dataset. This is a lower-risk transition path.

S4 — Target Model removes cluster dependency entirely. Istanbul deliveries are vans-only; all out-of-city shipments are optimized freely between vans and trucks. This is the full dynamic optimization deployment.

Key Findings

The S1-to-S2 comparison makes the structural problem visible. Under identical route logic, applying real constraints drops deliverable pallets from 4,490.5 to 3,814.5 — 676 pallets lost, 307 orders left unassigned. The root causes split into two categories:

Capacity overflow accounts for 67% of unassigned orders. Route ASY 2.1 illustrates this: average daily demand of 8.6 pallets against a van capacity of 8. A margin of 0.6 pallets causes overflow on 13 of the route's 22 active days. This is not an edge case — it is a systematic design mismatch.

Geographic infeasibility accounts for the remaining 33%. Route DOĞ 1.1 combined Aksaray, Osmaniye, and Kilis — cities more than 400 km apart — on the same vehicle with the same time window. The vehicle cannot complete the route; it returns with empty capacity and leaves orders unserved.

Scenario Results

Moving from S2 to S3 (controlled transition) reduces unassigned orders by 70.7% — from 307 to 90 — and delivers 487.5 additional pallets using 77 fewer total routes. Fleet utilization improves: average vehicle load factor rises from 64.2% to 83.0%.

Moving from S2 to S4 (target model) reduces unassigned orders by 96.1% — from 307 to 12 — and delivers 657 additional pallets. Load factor reaches 85.5%. The 12 remaining unassigned orders represent geographic edge cases, not a structural failure.

The peak-day stress test on 19 February 2026 is the clearest demonstration of the difference. Total demand reached 93 pallets. The current model deployed only 8 vehicles, left 20 pallets undelivered, and satisfied 78.5% of demand. The S4 model deployed all 10 vehicles simultaneously, delivered all 93 pallets with zero unassigned orders.

Fuel ROI

Net fuel ROI was calculated on a 74-day equal-order subset to enable direct comparison. Using a uniform consumption model (22 L/100 km across all vehicles), S4 produces the lowest fuel cost at 976 TL per pallet — compared to 1,121 TL per pallet under S2. The total fuel advantage over the 74-day window is approximately 451,000 TL relative to the constrained baseline.

Under the vehicle-differentiated model (18.5 L/100 km for vans, 33 L/100 km for trucks), S3 becomes the most fuel-efficient scenario because it increases truck utilization while keeping van routes short. S4's higher truck deployment adds fuel cost but the corresponding service level improvement — 96.1% reduction in unassigned orders — makes the trade-off commercially clear.

The financial case for S4 is not built on fuel savings alone. It is built on recovered delivery coverage: the commercial value of 657 additional pallets delivered per quarter.

Recommended Path

The two-phase transition is the operationally sound approach for FoodX:

Phase 1 — S3: The operations team adapts to dynamic daily assignment while retaining familiar vehicle-type logic. Unassigned orders fall from 307 to 90 with lower organizational change risk.

Phase 2 — S4: Once the team is confident in dynamic dispatch, the full model removes cluster constraints entirely and achieves near-complete delivery coverage.

What we delivered