Since 2016, CTA has been working with farmer organisations in Burkina Faso, Fiji, Kenya, Lesotho, Samoa, Swaziland, Trinidad, and Tobago and Uganda, to implement their Data for Agriculture (Data4Ag) project. The aim of the project is to investigate how the collection and effective management of farmer data can be used by farmer organisations to improve the livelihoods of their members.
Gathering data about farmers to profile them constitutes the first critical step in the project process. The type of data collected includes farmers’ personal details, farm details, crop type, and production information. This profile data provides farmer organisations with reliable and accurate information on their members. This information gathering does not only enhances membership management, but it is used to improve handling of credit or inputs to farmers, organising logistics, and marketing of products. Through the project, farmers’ organisation and agri-businesses have also taken advantage of geolocation data and the basic profiling of crops and farmers to provide targeted and on time services such as weather and market price information to farmers thereby ensuring a more efficient business operations.
Digital, geo-referenced data
One farmer organisation benefitting from the Data4Ag project is the Igara Tea Factory (IGTF) in Uganda. The agribusiness grows, processes and packages eight grades of tea for export and local consumption. Until early 2017, IGTF was storing shareholders’ information on one database but individual data sets were incomplete, unverified and lacked crucial geo-referencing. This led to various challenges, ranging from unreliable farmers’ records and locations of tea farms, to weak systems for tracing inputs provided to farmers on credit due to the lack of information about farmers and their farm location.
With support from CTA, IGTF established a spatial data management system and profiled farmers from mid-July 2017 to early 2018. The digital profiling involved compiling geo-referenced information about tea farmers and their land using GPS-enabled tablets. Extension officers then uploaded the data onto a dedicated online platform and subsequently onto the IGTF’s database. The established data management system led to elimination of the various inefficiencies that existed beforehand. With the help of the profile database, IGTF is now able to map their members’ locations. Information gathering has also enhanced membership management, crop collection logistics and processing, and product marketing.
Impacts for the Igara Tea farmers
The project also stimulated IGTF to launch a savings and credit cooperative. This has reduced the risk of supplying loans to small-holder farmers and farmers have easier access to credits because of the introduction of the database that monitors farmer produce and tracks delivery to the factory. To determine other specific impacts of profiling for farmers, two statistical approaches, the Statistical Package for the Social Science (SPSS) and Machine Learning (ML), were employed by the CTA research team.
Machine learning is a branch of artificial intelligence that is able to process large datasets and let them learn for themselves without been explicitly programmed. The analysis was built up using different regression algorithms and evaluation methods from the Azure Machine Learning studio.
The data for the impact evaluation was obtained from IGTF and included information regarding; farmer identification, age, gender, farm profile information, credit/input access, and yield level. The average yield of tea leaves for the various harvesting season is shown below:
From the SPSS analysis, we compared the yield difference between profiled farmers and non-profiled farmers and farmers with credit access and without credit access. The results of the SPSS analysis showed a significant difference in the mean yield of profiled and non-profiled farmers.
There was also a significant difference in the mean yield between farmers who had access to credit and those who did not. In both cases, the mean yield of farmers who were digitally profiled and had access to credit was significantly higher than farmers who were not profiled and had no financial access. The results showed a 10% increase in the yield of tea leaves of profiled farmers. Also, from the machine learning analysis, we were able to investigate the relationship between the variables of interest and the yield of tea leaves. From the results, it can be concluded that farmers with access to their digital profile are in a better position to demonstrate their credit worthiness to financial institutions and acquire credit services. Availing farmer data also made it possible for farmers to be targeted with the right agronomic advice, which helped them to make informed decisions in their farming operations, resulting in yield increases.
It was evident from the analysis that the Data4Ag project also employed a gender inclusive approach in profiling farmers, as both male and female farmers had equal chances of been profiled. Furthermore, the increased access to credit was irrespective of gender. Also using the decision forest algorithm, a significant difference in returns was identified.
Towards this end in the other project countries, the same methodological approach will be applied with organisations and cooperatives, such as the Eastern Africa Farmers Federation (EAFF) in Kenya and the National Union of Coffee Agribusiness and Farm Enterprises (NUCAFE) in Uganda. Impacts from these projects are impending.