PROJECT 1 · DATA
BOI combines public agricultural price data, official Burundian statistics and open geospatial data to construct a reproducible analytical evidence base for agricultural supply-network optimization.
Public and secondary datasets have been integrated under explicit provenance controls. Primary empirical observations required for freight-cost calibration, supply, demand, cost, logistics capacity and independent holdout validation have not yet reached the threshold required for empirical model validation.
SOURCE REGISTER
Role: Market-level food price evidence
The current Burundi RTFP dataset covers 74 markets and approximately 17,700 observations across ten food commodities. The dataset combines direct price measurements and model-based estimates and is subject to revision.
Status: Integrated as the principal market-price reference layer
Role: Official Burundi price evidence
INSBU monthly bulletins provide commodity-price information at province level. Historical bulletins from the 18-province administrative system were reviewed, together with the five-province system used from July 2025.
Status: Controlled secondary evidence
Role: Agricultural structure and production context
RGPHAE agricultural modules, agricultural producer-price statistics and related INSBU publications provide official evidence on agricultural households, production systems and agricultural markets.
Status: Registered for controlled integration
Role: Road-network and accessibility layer
The BOI analytical network was constructed from OpenStreetMap road data. The retained network contains approximately 37,981.7 km of roads and supports market-to-market routing.
Status: Integrated into the analytical baseline
Role: Normalization, crosswalks and optimization inputs
BOI processing harmonizes commodities, dates, units and market identifiers before applying the optimization framework. Province-level evidence is not automatically assigned to individual markets.
Status: Controlled and reproducible
Role: Transport-cost calibration
Public evidence confirms the importance of domestic freight costs, but transaction-level agricultural freight tariffs remain insufficient for empirical recalibration. A literature benchmark of 811.331 BIF per tonne-kilometre is therefore retained only as an external benchmark.
Status: Empirical acquisition pending
DATA GOVERNANCE
Each integrated source is classified according to origin, observation type and scientific role. Secondary evidence is not presented as primary field observation.
Commodity names, measurement units, dates and geographic identifiers are harmonized before analytical processing.
Province-level INSBU observations are not silently converted into market-level observations. Geographic crosswalks require explicit evidence.
Data dictionaries, source registers, quality-control files and processing specifications are maintained as part of the BOI research record.
EMPIRICAL VALIDATION
A professional research platform must distinguish an analytical result from an empirically validated result. BOI therefore keeps the following acquisition gates explicit:
Once sufficient independent observations are acquired, the recalibration protocol can be executed and the model evaluated against out-of-sample evidence.
TRANSPARENCY
The current Project 1 baseline is analytically finalized and reproducible, but it should not be described as empirically validated. This distinction is part of BOI's scientific disclosure standard.
View validation status →