Most agricultural technology is designed for farms that do not exist here. The useful systems start from the smallholder’s real constraints.
There is a version of agricultural technology that photographs well: sensor arrays, drone imagery, dashboards showing soil moisture across a thousand hectares. It is real, and it is largely irrelevant to a farmer working two acres.
The systems that make a difference at smallholder scale are far less impressive to look at. They deal in prices, payments, inputs and records — and they work on a phone that cost under thirty dollars.
Design for the handset people actually own
Smartphone penetration is rising, but a service that requires one excludes a significant share of the people it is meant to reach. Worse, it excludes them unevenly: often older farmers, often women, often those in exactly the areas where the service would matter most.
This has concrete consequences for how these systems get built:
- USSD and SMS are primary channels, not fallbacks bolted on afterwards.
- Interactions must survive being interrupted and resumed, because sessions time out and credit runs out.
- Every message costs someone money, so brevity is a functional requirement.
- The system must work with a shared handset — a household may have one phone between several people, which affects how identity and consent work.
A service designed for USSD first and extended to smartphones later almost always reaches more people than one built the other way around.
Market information is only useful if it is trusted
Price information services are among the most common agricultural interventions and among the most frequently abandoned.
The reason is usually credibility. A farmer who acts on a quoted price, transports produce to market, and finds the real price materially different will not use the service again — and will tell others. One bad experience outweighs a great deal of accurate data.
Building for trust means being honest about uncertainty: where the figure came from, when it was collected, which market it refers to, and how much it typically varies. A service that says "this is yesterday's average at this market, and it moves by about this much" is far more useful than one presenting a single confident number with no provenance.
Records are the route to credit
The most valuable thing many of these systems produce is not advice. It is evidence.
A farmer without a formal financial history is invisible to a conventional lender. A farmer with two seasons of recorded input purchases, delivery volumes and payment history has something a lender can assess. That record is often the difference between borrowing at cooperative rates and borrowing informally at much worse ones.
This makes data ownership a real question rather than a policy footnote. If the record only exists inside a private platform, the farmer's history is only portable at that platform's discretion. Systems that let a farmer export or carry their own record are doing something meaningfully different from those that treat it as proprietary.
Cooperatives are the natural integration point
Working with individual farmers directly is expensive to run and hard to sustain. Cooperatives and aggregators already handle collection, weighing, grading and payment — which means they already hold most of the data worth digitising, often in paper ledgers.
Digitising that layer tends to produce faster benefit than any farmer-facing app: accurate weight records reduce disputes, prompt payment records reduce mistrust, and aggregate volume data makes planning possible. The farmer-facing service then has something real to sit on top of.
Connectivity assumptions break in the field
A collection point with usable coverage on the day of the survey may have none during the rains. Systems that depend on a live connection at the moment of weighing will fail at exactly the point where accuracy matters most, and the fallback will be a notebook that never gets entered.
Local capture with later synchronisation is not a nice-to-have here. It determines whether the digital record and the physical reality stay in agreement.
What we have found matters most
Across this work, a few things recur. Simple channels reach more people than sophisticated ones. Trust is built by being accurate about uncertainty rather than confident about estimates. The record is often worth more than the advice. And the cooperative layer is usually where digitisation pays back fastest.
None of this requires novel technology. It requires designing for the conditions that actually exist.
Where we fit
We build farm and cooperative management systems, produce collection and payment platforms, market information services, and the USSD and SMS interfaces that make them reachable. The work is generally about reliability and reach rather than sophistication.
More detail is on our agriculture and agritech page, or tell us about your cooperative or programme.