From Data to Impact: What We Built and Learned Through DataKind’s Global DataKit with Producers Direct

By Mallory Sheff, Director, Financial Inclusion, DataKind

In late 2025, DataKind partnered with Producers Direct to run a global DataKit event focused on understanding smallholder farmers’ needs through one of the largest peer-to-peer agricultural datasets ever assembled. By combining rare, farmer-generated data with a product-first approach to experimentation, the DataKit event surfaced insights at the intersection of climate resilience, agriculture, and financial inclusion – while also testing methods that could inform future data-driven products. This post shares what we built, what we learned, and how those insights are shaping what comes next.

DataKind and partner Producers Direct, a farmer-led nonprofit that supports over 1.3M smallholder farmers in East Africa and Latin America, kicked off a month-long DataKit event in late 2025. A DataKit is a virtual volunteering program where DataKind technologists work with curated data to answer key questions and share practical insights with social impact partners.

Our event objective? To leverage the power of our global data science community to unlock insights that could strengthen climate resilience, boost agricultural productivity, and expand financial inclusion for smallholder farmers.

Over the course of the fully virtual, self-paced event, data scientists from more than 60 countries joined us to wrangle, explore, and experiment with real-world data, generating insights with practical applications.

A Rare Dataset, Captured in Farmers’ Own Words

The data used for this event was shared by Producers Direct and comes from their now-closed WeFarm platform, a free, peer-to-peer SMS service that allowed farmers to ask questions and share advice directly with one another using basic mobile phones.

Over the course of four years, the WeFarm dataset grew to over 20 million rows of data, offering an unusually rich window into the challenges farmers faced and the opportunities they sought across East Africa during that time. This dataset is exceptionally rare: farm-level data captured in farmers’ own words, at scale, and across four languages – English, Swahili, Nyn, and Lugandan. Too often, digital tools and AI systems in agriculture are built without direct insight into what farmers are actually asking, worrying about, or responding to in real time – creating a gap between technology and lived experience.

Designing the DataKit for Insight and Application

Participants were guided by six key challenges co-created with Producers Direct. The challenges were developed to not just explore the data, but also generate key insights that could inform future collaboration and product development.

This approach reflects DataKind’s product-first strategy, where short-term engagements can inform repeatable, scalable data products that partners can build upon over time.

Testing Approaches That Could Power Future Products

Participants used this opportunity to test and compare innovative approaches that could underpin future data products. Under the data preparation and translation challenge, for example, volunteers evaluated multiple translation strategies, iterating with large language models such as Claude, open-source machine translation models including Meta’s NLLB-200 and other Hugging Face models, and Google Cloud Translate. Across other challenges, volunteers applied advanced techniques such as topic modeling and clustering analysis to surface patterns, trends, and insights embedded within the data.

Global Collaboration at Scale

Within a month, the DataKit event GitHub repository had grown to more than 60 submissions from 55 participants, producing a diverse set of approaches, validation methods, and assumptions, creating a strong foundation for synthesis and downstream product exploration.

What This Meant for Producers Direct

For Producers Direct, the value of the DataKit event extended well beyond analysis:

What the Data Revealed

Insights from the DataKit event revealed patterns in when and why smallholder farmers sought advice from peers through the (now ended) WeFarm SMS platform. Farmer engagement increased significantly during periods of extreme weather such as droughts, heavy rains, or unusual temperatures, when uncertainty and risk were highest. Many questions focused on protecting crops and livestock from pests and disease, often following changes in rainfall, highlighting the importance of timely, climate-responsive information.

Across the dataset, farmers consistently asked about a core group of staple crops and livestock, including maize, chickens, cows, and bananas, highlighting shared priorities across East Africa. There was a noticeable increase in questions around market access and market prices during the planting and harvest seasons, reflecting common cash-flow pressures faced by smallholder farmers.

Together, these insights point to significant opportunities to deliver the right information at the right time, helping farmers make better decisions, protect their livelihoods, and build resilience as climate conditions become more unpredictable.

From Insight to Impact: What Comes Next

For DataKind, this collaboration deepened our understanding of the interconnectedness of climate change, agriculture, and financial inclusion, reinforcing our commitment to using data to drive impact at these crossroads. By stress-testing data, methods, and assumptions in real-world contexts, we uncover opportunities for new data-driven products that help partners like Producers Direct – and the farmers they serve – adapt, thrive, and build long-term resilience in a changing climate.

DataKind and Producers Direct are continuing to explore ways to collaborate in the near future. If you’re interested in following this journey, or want to join a future DataKit event, please subscribe to our newsletter here.

Thanks to the volunteers and Courtney Epstein for their contributions to this event.

This program was made possible thanks to the generous support of Teradata Cares.

Images above courtesy of Producers Direct.

Join the DataKind movement.

Interested in sponsoring a project? Partner with us.
Interested in subscribing to our newsletter? Sign up.
Interested in supporting our work? Donate here.