27/05/2026
WHY ON-FARM EXPERIMENTS?
Experiments conducted at the station under the researcher's
control produce impressive yield numbers. But when those same technologies reach smallholder farms, the results often disappoint. On-farm experiments exist to close that gap. They are one of the most important tools in applied agricultural research.
THE PROBLEM
The gap between on-station yield and farmer yield is large. However, it is not a mystery. Walk into any National Agricultural Research Station (NARS) in East Africa and you will find maize yields of 6 t/ha or more. Ask a smallholder farmer in the same region what they harvest, and the answer is closer to 1.5 t/ha. The gap is real, large, and well-documented.
But this is not primarily a story about farmer ignorance or poor management. It is a story about what research stations are, and what they are not. Stations offer controlled conditions, optimal inputs, and professional management. They are not designed to replicate the constraints, variability, and resource limitations of a smallholder farm. When research happens only at the station, we are testing technologies under conditions that most farmers will never experience.
Maize yield gap — East Africa
Setting Yield
Genetic potential 8.5 t/ha
On-station (researcher-managed) 6.2 t/ha
On-farm (researcher-managed) 4.8 t/ha
On-farm (farmer-managed) 3.2 t/ha
Average farmer yields 1.5 t/ha
Each step down reflects real-world management constraints — not variety failure.
“A technology that only works at a research station is not a technology for farmers.”
Five main reasons why on-farm experiments matter
1 Real soils, real variability: On-station soils are carefully managed, fertilised, and relatively uniform. Farmer fields have variable pH, texture, organic matter, and drainage, often within a single plot. Technologies tested only on-station may succeed partly because of the soil, not because of the technology itself. On-farm experiments expose interventions to the variability that farmers actually face. Farmers' fields have been described as notoriously heterogeneous.
2 Real management constraints: Farmers face competing labour demands at planting and weeding time. They cannot always apply inputs at the optimal moment. They may split fertiliser doses because cash is short. On-farm experiments, especially farmer-managed trials, expose technologies to these real constraints, revealing whether a variety or practice can survive imperfect management, which is the only kind of management most farmers can afford.
3 Farmers as evaluators, not just recipients: Breeders optimise for yield. Farmers evaluate varieties on taste, storability, cooking quality, ease of threshing, stover quality for feed, drought tolerance, and market acceptance. In on-farm experiments, especially participatory varietal selection, farmers assess these traits directly. The result is a recommendation that farmers will actually adopt, rather than one they will try once and abandon.
4 Evidence that extension and policy-makers can use: An extension officer advising a smallholder needs to know what that farmer will achieve, not what a researcher achieved at a station. Multi-environment trial data from 15–20 farmer-managed sites, spanning diverse soils and rainfall zones, is what actually underpins credible technology recommendations. Station data alone cannot do this job.
5 Bridging the adoption gap: Technologies with strong station performance but poor on-farm performance stall in extension systems for years. On-farm trials identify this failure mode early, before significant resources are invested in scaling something that will not work under real conditions. When on-farm results are strong, extension officers can speak from evidence, not from optimism.
\ Research & Analysis \AgriculturalResearch
\FarmerExperiments \On-farmResearch \Researchdesigns
