**Background:** The authors, associated with CGIAR, present a perspective on how data-driven approaches can transform crop diversity management—encompassing genebanks, breeding programs, and farmer engagement—for sustainable development in the Global South. They contrast data-driven approaches (integrating diverse datasets via flexible, inductive methods) with traditional model-driven approaches (starting from conceptual coherence and causal understanding). The paper identifies three historical barriers to effective crop diversity use: (1) a focus on mean response and broad adaptation that treats genotype-by-environment (GxE) interactions as nuisance factors; (2) replication and randomization requirements that widened the gap between researcher and farmer experimentation; and (3) a historical focus on productivity and stress tolerance with limited market intelligence. The authors argue that data-driven approaches can address these barriers by handling higher complexity and achieving greater representativeness of real-world crop use contexts.
**Methods:** This is a narrative perspective paper, not a systematic review or meta-analysis. The authors synthesize examples from recent literature and CGIAR initiatives to illustrate data-driven approaches across four domains: (1) better use of plant genetic resources (e.g., the Seeds for Needs initiative involving over 40,000 farmers across Africa, Asia, and Latin America; the World Vegetable Center distributing over 42,000 seed kits containing over 183,000 vegetable seed samples to smallholder farmers in Tanzania, Kenya, and Uganda from 2013 to 2017); (2) insights from systems biology and omics (e.g., genomic selection, deep learning for genomic prediction, low-cost smartphone-based phenotyping via Project Artemis); (3) adaptation to environment and management (e.g., envirotyping, crop modeling linked to genomic analysis, the tricot [triadic comparisons of technology options] approach for on-farm trials); and (4) demand for diversity related to gender and social differences (e.g., linking user trait prioritization to socioeconomic data in Nigeria, showing that gender interacts with poverty and food security to shape trait priorities).
**Key Results:** The paper does not report original experimental results but highlights illustrative findings from cited studies. For example, in Ethiopia, farmers using the Seeds for Needs approach were able to identify superior landraces of durum wheat that were directly released as varieties. A study on upland rice in Brazil showed that breeding should move away from broad adaptation and carefully place trials geographically under future climates. Ethiopian on-farm data linked to genomic data demonstrated that genomic selection can be greatly enhanced by on-farm and environmental data, representing a groundbreaking example of combining selection efficiency with fitness in target environments. The Nigerian cassava study showed that quality traits were more important for members from food-insecure households, and gender differences between men and women increased among the food insecure, where women prioritized quality traits more. The authors note that less than half of studies on adoption of climate-adapted varieties disaggregated data by sex, and that over 5.7 million accessions are stored in 831 genebanks worldwide, but the majority are not described, with backlogs in regeneration and characterization.
**Clinical Implications:** Not applicable—this is an agricultural research perspective, not a clinical study. However, the paper has significant implications for agricultural research policy and practice. The authors recommend: (i) supporting genebanks to play a more active role in linking with farmers using data-driven approaches; (ii) designing low-cost, appropriate technologies for phenotyping (e.g., smartphone-based systems); (iii) generating more and better gender and socioeconomic data, including at the individual rather than household level; (iv) designing information products to facilitate decision-making; and (v) building more capacity in data science. They emphasize that broad, well-coordinated policies and investments are needed to avoid fragmentation and achieve coherence across domains and disciplines. The paper concludes that data-driven approaches can facilitate a more open innovation process, allowing the entire crop diversity management system to carry more information and address challenges of climate change, rapid consumer trends, and gender and social equality in the Global South.