Engineering · Case Study

From Paper to Digital: Building a Multi-Vendor Agricultural Marketplace

The SACCO Wallet project started as a simple digital wallet for a rural savings cooperative in Kyenjojo. It evolved into something much bigger: a multi-vendor agricultural marketplace where farmers can sell produce, traders can source goods, and SACCOs can manage member finances — all from a single app. This is the story of how we built it and the technical decisions along the way.

Phase 1: The Digital Wallet

The initial requirement was straightforward: digitize the SACCO's paper-based savings and loan records. Members needed to check their balance, see transaction history, and make deposits via mobile money. We built this as a PWA with a Supabase backend — lightweight, mobile-first, and offline-capable for rural users with unreliable connectivity.

Phase 2: The Pivot to Marketplace

After launching the wallet, the SACCO members asked a question that changed the project's direction: "Can we also sell our produce through this app?" The cooperative's members were farmers — maize, beans, coffee, bananas — and they were selling through middlemen who took significant margins. If we could connect farmers directly to buyers, the economics improved for everyone.

We redesigned the app into a multi-vendor platform with four user roles:

• Farmers: List produce for sale, set prices, manage inventory
• Traders: Browse and purchase produce in bulk, arrange transport
• Vendors: Set up digital storefronts for retail sales
• Stores: Larger outlets with inventory management and sales tracking

GPS-Based Land Measurement

One of the more technically interesting features: farmers can measure their land size using GPS directly in the app. They walk the perimeter of their field holding the phone, and the app records GPS coordinates and calculates the area. This data feeds into the AI crop recommendation engine — knowing the exact land size lets the system suggest crops that are viable for that specific parcel.

The implementation uses the browser's Geolocation API with a custom path-tracing algorithm that records points at intervals, filters GPS noise, and computes the enclosed area using the shoelace formula. The accuracy isn't surveyor-grade, but it's good enough for crop planning and gives farmers a number they can reference when applying for SACCO loans.

AI Crop Recommendations

We integrated NASA's satellite climate data API to pull real-time precipitation, temperature, and soil moisture data for the Kyenjojo region. The AI recommendation engine combines this climate data with the farmer's land size, historical crop performance, and current market prices to suggest the most profitable crops to plant. It's not a perfect prediction — agriculture has too many variables for that — but it gives farmers a data-informed starting point instead of guesswork.

The Market Page

We designed the marketplace page (accessible at /market in the app) as a Jumia-inspired mobile marketplace. Farmers list produce with photos, prices, and availability. Buyers can filter by crop type, price range, and location. Transactions are settled through the SACCO wallet — the same account members use for savings and loans. This integration means the SACCO can offer financing for purchases, and the marketplace transactions are visible to the cooperative for transparency.

Design Language

The visual design evolved from a generic fintech aesthetic to a warm, agricultural identity. We chose a cream background (#FAF8F4) with dark forest green (#1a4731) as the primary accent. Navigation uses pill-shaped components — a design choice that feels approachable and non-technical for users who may be using a digital marketplace for the first time.

What We Learned

Building software for farmers in rural Uganda taught us that feature complexity is the enemy of adoption. Every feature we added had to pass a simple test: could a farmer who has never used a smartphone before understand this within two minutes? We removed features that failed this test, even when they were technically impressive. The GPS land measurement stayed because it's immediately useful and requires only walking around a field. The AI recommendations stayed because they answer a direct question: "What should I plant?"