One Acre Fund - Creating a Chatbot With Limited Resources to Forecast Optimal Seeding Time in Sub-Saharan Africa
Name: Emiel Veersma & Alex Sananka
Title: Data Scientist & Senior Associate, Data and Analytics
Organization: One Acre Fund
One Acre Fund is a nonprofit organization that supplies smallholder farmers in East Africa with asset-based financing and agriculture training services to reduce hunger and poverty. Headquartered in Kakamega, Kenya, the organization works with farmers in rural villages throughout Kenya, Rwanda, Burundi, Tanzania, Uganda, Malawi, Nigeria, Zambia, Ethiopia, and India. One Acre Fund actively serves more than 1 million farmer families.
One Acre Fund offers smallholder farmers an asset-based loan that includes: 1) distribution of seeds and fertilizer; 2) financing for farm inputs; 3) training on agriculture techniques; and 4) market facilitation to maximize profits. Each service bundle is around US$80 in value and includes crop insurance to mitigate the risks of drought and disease. To receive the One Acre Fund loan and training, farmers must join a village group that is supported by a local One Acre Fund field officer. Field officers meet regularly with the farmer groups to coordinate delivery of farm inputs, administer trainings, and to collect repayments.
One Acre Fund offers a flexible repayment system: farmers may pay back their loans in any increment at any time during the growing season. Beyond their core program model, One Acre Fund also offers smallholder farmers opportunities to purchase additional products and services on credit. These include solar lights and reusable sanitary pads.
Data Science for Good
Most Impactful Ikigai Story
Climate change makes it more and more difficult for farmers in Africa to know when the rain will come — especially where there is low digital penetration. Not only the timing of the monsoon is affected, but also its duration. Hence they may plant seeds too early, and in this case risk germination failure due to insufficient rainfall to keep the soil moist enough , or too late, that there is insufficient water available for development during the growing season.
Planting at the right time has a very high impact on yield, and is one of the biggest levers we have yet to effectively pull when it comes to improving outcomes for farmers. Pilot studies by One Acre Fund and the University of Reading suggest that suboptimal decisions on when to plant can reduce yields by 7-14% for Kenya. Optimizing planting date therefore provides a clear pathway towards improving yields and reducing food insecurity.
The traditional method in the region is to plant seeds after four days of rain. One Acre Fund partnered with the University of Reading to analyze the effectiveness of this technique in light of the new weather conditions. We concluded that the changes in drought periods rendered this method irrelevant.
The farmers that our organization is addressing are rural and relatively hard to reach — they often do not have access to a smartphone. Hence we created a chatbot using USSD technology to send them direct messages, so they can receive relevant information and communicate back to us. The chatbot approach enabled us to automate and scale our communication effectively.
Dataiku helped us connect to different sources of data, like internal data, satellite data and survey information.We created one project to clean the satellite data, and another one to synchronize the chatbot with the usable datasets — all in just a few clicks.
Dataiku was pivotal to enable us to build the project without the help of any data engineers. Being a nonprofit organization with limited resources, it would have been much more difficult and uncertain to deploy this — whereas our traditional statisticians were able to process all data with Dataiku’s intuitive visual interface.
We used to communicate the four-day rule with the farmers: plant after four days of rain. With the Auto ML features of dataiku, we were able to come up with the most relevant forecasting model, tailored to every farmer. With the help of the Dataiku API we could connect the data and the model to the chatbot, so every farmer can access his personal planting date recommendation.
Business Area Enhanced: Marketing/Sales/Customer Relationship Management
Use Case Stage: Built & Functional
Thanks to receiving information at the right time via the chatbot, farmers are able to plant seeds at the optimal moment to maximize their output. In the long-term, based on the results of pilot studies (which resulted in an income increase of $10.6 per adopter) the initiative will generate an average of $5.58 per farm annually based on 52.6% adoption rate.
For One Acre Fund, the chatbot enabled us to reach hundreds of thousands of farmers and engage them in a scalable manner with our support initiatives. Since the solution fits into their daily habits, it has been widely adopted among our audience - now we will focus on fine-tuning our measurement of the effectiveness of our messaging and timing.
We also plan to leverage the chatbot for more use cases, enhancing the “customer experience” of our farmers and enabling us to expand our service offerings. We are currently deploying a fertilizer recommendation, which will support the farmer to apply the right type and amount of fertilizer.
Specific Value Brought by Dataiku:
This project wouldn’t have been possible at all without Dataiku! Our team didn’t have the relevant resources (incl. data engineering skills) to create and productionalize this workflow. For this project, we needed to:
Extract data from satellite data suppliers
Collect internal data from our ERP system and our survey tool
Analyse this data to come up with personal recommendations
Share these recommendations real-time with the chatbot
We might have been able to perform these tasks, but every one of them would have been much more complex and time consuming. For instance, it’s very fast to create APIs and link those with other software such as Infobip. Without Dataiku, it would have probably taken us days/weeks to come up with a working flow — hence we don’t think we would have made it to a good end with limited resources.