In this paper, is first described the design and development process of a modular ICT application system called GeoFarmer. Geofarmer was designed to provide a means by which farmers can communicate their experiences, both positive and negative, with each other and with experts and consequently better manage their crops and farms. We designed GeoFarmer in a collaborative, incremental and iterative process in which user needs and preferences were paramount.
The Colombian Ministry of Agriculture Colombia, an international research center and a national farmers’ organization developed a data-driven agricultural program that: (i) compiles information from multiple sources; (ii) interprets that data; and (iii) presents the knowledge to farmers through the local advisory services. Data was collected from multiple sources, including small-scale farmers. Machine learning algorithms combined with expert opinion defined how variation in weather, soils and management practices interact and affect maize yield of small-scale farmers.