Data-Driven Ranking and Typology of African Agrifood Systems Using Dimensionality Reduction

Data-driven rankings and typologies increasingly mediate how public institutions read food security, yet they ship without the practices this community expects of deployed models: no uncertainty, no stability audit, no external validation, no provenance. The stakes are population-scale: roughly 309 million Africans went without enough food in 2025, and the African Union’s new Kampala CAADP Strategy (2026–2035) opens a measurement decade whose scorecards will steer policy for 1.5 billion people. From 47,965 sourced observations covering all 54 African Union member states in international statistical systems (2000–2025), we build an Agrifood Transformation Index in which every ranking carries a bootstrap confidence interval (median 90% rank interval: four positions), over and under-performance is separated from income, and validation uses outcomes withheld from construction. Three independent tests (bootstrap cluster stability, a multimodality test, and a matched unimodal-Gaussian null) show that the continent is a continuum rather than a set of discrete types, so the five-type classifications used by global dashboards impose boundaries the data do not contain. A frozen gradient panel analysis finds no narrowing of the gap between Africa’s two broad agrifood regimes in 25 years, and a reconstruction of all four CAADP Biennial Reviews (130 sourced country-round scores) explains why the official scorecard reads as unrelieved failure. We distill the pipeline into transferable trustworthiness practices for measurement systems that steer public institutions.