Jua
Harvard University | 2022-2023 | Master’s Thesis Project
Received the 2023 Outstanding Design Engineering Project Award.
Cultivating digital knowledge networks for smallholder farmers in East Africa.
Roles & Skills:
User research, problem definition, user testing
Data analysis, UI/UX, prototyping
Business strategy, market analysis
As part of the Master in Design Engineering (MDE) program at Harvard, students complete a thesis project addressing a real-world, societal challenge using design and engineering methods. This thesis project was done in collaboration with Rebecca Brand and won the 2023 Outstanding Design Engineering Project Award.
Visit our website here.
Context and Background
Our team united around a passion for climate change, social impact and open data. Given our diverse experiences and skillsets, we decided to focus our project on addressing the most pressing challenges that East African smallholder farmers face.
Climate change is already impacting our world and the livelihood of millions of people. By the end of the century, an additional 183 million people could face food insecurity and malnutrition. Looking closer at maize yields across different regions in the world, it is clear that Africa’s full agricultural potential remains untapped. In a region where 60% of the population is smallholder farmers, increasing their productivity in particular will be a major driver of growth.
Agricultural Value Chain
Smallholder farmers are the focal point within the agricultural value chain sitting between inputs & suppliers, outputs & markets, and lateral markets.
In a given agricultural season, smallholder farmers consider these various levers when making decisions about their land. And in turn, each of these levers influence the productivity and resilience of a farmer. These influences are inter-connected and often emerge as constraints for most smallholder farmers in East Africa. In the face of these uncertainties, we see a pattern in which empowering knowledge and access to information for smallholder farmers can support their long-term growth.
Access to information, including access to resources and data, can work to combat these uncertainties. A report conducted by the University of Washington found that:
Access to information has the potential to double the productivity of small-scale food producers when that information is tailored to local needs and educational pathways.
User Research & Stakeholder Analysis
To explore this space further, we conducted over 60 interviews with industry and academic experts and 40 interviews with East African smallholder farmers. We also had the opportunity to visit Kenya twice to co-create and develop this work alongside smallholder farmers.
Across these interviews, we heard:
1. Small-scale farming is highly collaborative. Farmers do not view their work as a zero-sum game and instead rely on each other for help and support.
"I have no problem with [sharing information with other farmers] because they share information with me. Why should I be selfish?”
2. People are thirsty for more information, but when they learn something new, they will perform their own research to make a more holistic decision. They are willing to try new things and take calculated risks, but ultimately need to try things themselves before they trust new information.
“By nature, I am a very skeptical person. You don’t tell me something and I buy it the same day. I do my research. I wouldn’t say no to information, but I want to test it.”
3. Nearly every farmer we spoke with watched or listened to two specific farming programs on local television and radio. They liked educational content where they could learn about what other farmers are doing.
“I like [Mugambo wa Murimi] because I can see what other farmers are doing in their farms.”
4. Most people that had smartphones used WhatsApp. However, many farmers we spoke with expressed that they would love to join a group, but they did not know how to join one.
“I use WhatsApp to talk to other people and other farmers nearly every day.”
Opportunity Area
Through these insights and extensive literature review and research, we believe that there is an opportunity to design an intervention that connects farmers to a community-building platform via WhatsApp and SMS in order to cultivate information exchange between farmers to empower their decision-making and planning for the future.
Introducing Jua
Given this identified opportunity, we developed Jua: a multi-sided, digital platform that leverages conversation intelligence to support information and resource exchange among smallholder farmers and agricultural service providers in East Africa.
Jua Communities: User Testing & Demand Validation
To test and validate our concept, we have created and facilitated WhatsApp groups, or Jua Communities: small groups of Kenyan farmers who did not know each other. Our aim was to validate if farmers would engage on WhatsApp with people they did not know, how farmers would utilize these groups, and to identify the level of facilitation that Jua needed to provide to maximize and sustain positive, collaborative engagement.
In total, we created two Jua Communities with 13 farmers across 10 consecutive weeks. There were a total of 148 messages exchanged between farmers with 10 of the 13 farmers sending at least one message.
Jua Communities: Prototyping
While initial user testing has been conducted with WhatsApp for Business, we have prototyped the end-to-end experience utilizing WhatsApp Business API to send text, image, file, and quick reply messages to users while storing responses in a spreadsheet. We tested this WhatsApp Business API interaction with farmers in person to simulate an onboarding process for a new user.
We also prototyped an automated agricultural resource recommender using natural language processing (NLP) models. We compiled a database of vetted agricultural resources and tested three NLP models of varying complexity. Given the accuracy of results and processing power, we opted for a Word2Vec model. We could then apply the Word2Vec model to actual Jua Community messages to identify the most relevant resource in our database to the conversation. Finally, we automated the process of sharing the resource back to our Jua Communities.
While the Word2Vec model was the optimal option for our prototyping, as Jua gains traction, we would switch to a more robust spaCy model.
Jua Insights: Demand Validation & Prototyping
Our Jua Communities gave us real data to prototype Jua Insights: a B2B solution that leverages farmer conversations to attract agricultural providers and buyers. Farmer conversation data becomes a strategic value add for providers and buyers looking to boost their effectiveness and expand their reach. As Jua Communities gains traction, we believe that there is an opportunity also expand farmers’ access to local resources and markets.
Jua Insights would be the aggregated and generalized insights from Jua Communities, delivering farmer conversation topics and behavioral insights to stakeholders along the agricultural value chain who may not have direct access to farmers.
We have conducted preliminary demand validation for Jua Insights with local service providers. A program manager for an agricultural non-profit in East Africa stated:
“These data assets are extremely valuable because they come from such remote areas where it’s really hard to access reliable data. I could be a client.”
