Markets linked to the BOI road-routing layer.
BOI RESEARCH · 01 · RESULTS
Results
Quantitative results from the analytical optimization framework for agricultural supply networks in Burundi.
Scientific status: the results below are analytically reproducible. They do not constitute empirical validation of the model because independent transaction-level freight, supply, demand and cost observations are still required.
Unique unordered market corridors evaluated.
Retained OSM road network length.
Transport-cost scenarios evaluated.
01 · NETWORK STRUCTURE
Road-network and corridor results
The routing layer contains 65,566 nodes and 86,205 edges after network preparation. All 72 markets were snapped to the network, producing a complete 72×72 market distance matrix with 2,556 unique unordered corridors and no unreachable market pairs in the analytical network.
02 · PRICE-GAP SCREENING
Potential arbitrage opportunities
The price-gap layer evaluates ordered origin–destination pairs using the latest available product-market prices. At the illustrative transport-cost scenarios, thousands of route-product combinations remain above the modeled break-even threshold. These are opportunity signals, not observed transactions or realized profits.
| Transport cost (BIF/t-km) | Candidate routes | Selected normalized routes | Normalized margin |
|---|---|---|---|
| 500 | 22,992 | 354 | 446,682.36 BIF |
| 750 | 22,084 | 350 | 441,679.13 BIF |
| 811.331 | 21,848 | 349 | 440,398.14 BIF |
| 1,000 | 21,226 | 345 | 436,811.97 BIF |
| 1,250 | 20,373 | 345 | 432,191.97 BIF |
| 1,500 | 19,505 | 345 | 427,724.36 BIF |
The 811.331 BIF/t-km value is a literature-derived benchmark converted using the BRB exchange rate; it is not a current observed domestic agricultural freight tariff.
03 · ROBUSTNESS
Product-level robustness across transport scenarios
Robustness measures the share of screened route-product opportunities that remain profitable across the six modeled transport-cost scenarios. Higher values indicate lower sensitivity to the assumed transport-cost range.
| Product | Robust share |
|---|---|
| Goat meat | 98.2% |
| Onions | 87.8% |
| Tomatoes | 87.1% |
| Rice | 84.6% |
| Bananas | 84.4% |
| Potatoes | 82.4% |
| Beans | 75.2% |
| Maize flour | 65.6% |
| Cassava flour | 64.2% |
| Maize | 61.1% |
04 · MULTI-PRODUCT ALLOCATION
Coupled flow optimization
The multi-product model jointly allocates normalized flows subject to product-level supply and demand bounds and shared corridor-capacity constraints. Under the 811.331 BIF/t-km benchmark with a corridor capacity of three normalized units, the model selects 349 positive flow arcs with a total normalized objective value of 440,398.14 BIF.
“Normalized unit” is a modeling unit and must not be interpreted as a tonne, truckload, or observed shipment volume.
05 · SPATIAL PRIORITIZATION
Priority corridors
Spatial prioritization combines corridor priority, persistence in the optimization scenarios, robustness and product breadth. The spatial tiering identifies 96 A-tier, 534 B-tier, 1,804 C-tier and 122 D-tier corridors.
Top spatial corridorBukirasazi — GitegaSpatial priority score: 99.0833
06 · VALIDATION STATUS
What these results establish — and what they do not
Deterministic checks and internal consistency controls passed.
Methods, assumptions and research packages are documented.
Public INSBU and World Bank evidence is being integrated under controlled crosswalks.
Independent freight, cost, supply, demand and holdout observations remain required.
BOI will only label the model empirically validated after independent observations are used for calibration and out-of-sample testing, with transparent error metrics and documented acceptance criteria.
NEXT
Explore the research evidence
Review the methodology, data sources, validation record and reproducibility materials before interpreting or reusing these results.