1. Market price layer
The analytical system uses agricultural market-price observations to identify potential price differentials between market pairs.
Project 1 · Methodology
Methodology
A reproducible analytical framework for graph-based combinatorial optimization of agricultural supply networks in Burundi.
Research design
From market prices to network optimization
2. Road-network layer
OpenStreetMap road data are transformed into a routable analytical network and used to estimate road distance between market nodes.
3. Opportunity screening
Ordered origin-destination pairs are evaluated using gross price gaps and transport-cost scenarios.
4. Flow optimization
Linear programming is used to maximize normalized net margins subject to product-level supply, demand and shared-corridor constraints.
5. Robustness analysis
Candidate routes are evaluated across multiple transport-cost scenarios to distinguish robust, strong, moderate and sensitive opportunities.
6. Spatial prioritization
Corridor-level indicators combine robustness, product breadth, optimization persistence and spatial efficiency to produce a research-priority hierarchy.
Optimization model
Constrained network allocation
The core allocation problem is solved with linear programming. Decision variables represent normalized flows on profitable product-specific origin-destination arcs.
Constraints include normalized origin supply limits, destination demand limits and shared corridor-capacity scenarios. The objective is to maximize total normalized net margin after transport costs.
Reproducibility
Controlled analytical workflow
The project records data preparation, network construction, optimization, sensitivity analysis, corridor prioritization, validation checks and research limitations as separate stages.
All normalized optimization quantities are analytical units and should not be interpreted as observed tonnes or transaction volumes.
Scientific status
What the methodology does — and does not — establish
Important: The analytical model is finalized and reproducible, but it has not been empirically validated with a sufficient independent set of primary freight, supply, demand, transaction-cost and holdout observations. Recalibration therefore remains blocked pending empirical evidence.