For twenty years, satellite imagery in African agriculture was a donor pilot. A programme funded a mapping exercise, a report was produced, the funding cycle ended, and the maps aged into a drawer. Something changed quietly in the last few years, and it changed because of accounting rather than technology. Earth observation data started appearing on the cost side of real commercial transactions: an insurer pricing a drought payout, a lender deciding whether a smallholder’s field history supports a loan, an exporter proving to a European buyer that a coffee lot did not come from land cleared after a cutoff date. The imagery itself is close to free, since the public constellations publish openly and have done for years. What is being bought is not the pixel. It is the interpretation, delivered to a decision, on a deadline, with someone’s money resting on it. That makes satellite data an agricultural input in the ordinary sense, something purchased each season because a downstream process cannot run without it. The commercially important question, and the one most founders in this space get wrong, is not what the data can show. It is who is compelled to pay for it.
Key Takeaways
- The raw imagery is a commodity. Public constellations publish openly, so no venture wins by having access to pictures. The margin sits in interpretation delivered to a specific decision on a deadline.
- The farmer is usually not the customer. The paying party is the insurer, the lender, the exporter, or the buyer, because those parties carry a risk or a legal obligation the data resolves.
- Regulation is the strongest demand engine in the category. Deforestation rules on exports turned plot-level geolocation from a nice-to-have into a condition of market access for coffee and cocoa.
- Insurance and credit are the second engine. Parametric payouts and field-level history let institutions price risks they previously refused, which expands thin insurance markets rather than merely digitising them.
- The Payer Question is the filter: name who loses money if the data is absent, and you have found the customer. If nobody loses money, you have found a pilot.
Why is the data suddenly worth money after years of pilots?
Because obligations arrived. For most of the pilot era, satellite analysis produced insight that was interesting to everyone and mandatory for nobody, and insight without an obligation attached is a grant line, not a market. Three obligations changed that. The first is regulatory: European rules requiring proof that commodities did not originate on recently deforested land put plot-level coordinates and time-stamped forest cover into the compliance file for Uganda’s coffee exports, which means an exporter without that data does not have a slow business, it has no access to that buyer at all. The same logic is spreading through carbon border measures and due-diligence rules on supply chains, each converting an environmental claim into a documentation requirement someone must fund. The second is financial: an insurer writing a parametric policy needs an index it can trust and cannot dispute, and a rainfall or vegetation measure from orbit is harder to argue with than a claim form. The third is operational, and it is the slowest but the most durable: buyers of agricultural output increasingly want to know their supply base, and a company that can show where its volume comes from wins contracts against one that cannot. In all three cases the data stopped being a way to understand agriculture and became a way to complete a transaction, and that is the transition every category makes on its way from donor project to industry.
Who actually pays, and why is it almost never the farmer?
Because the farmer’s willingness to pay is real but small, and the downstream party’s exposure is large. A smallholder with two acres can benefit genuinely from a well-timed planting advisory, and will pay something for it, but the sum available is measured in the price of a few kilograms of produce. The exporter who loses an entire European contract for want of geolocation on a supply base carries a loss measured in shipments. The lender deciding on a portfolio of thousands of loans carries a default cost that a modest improvement in yield prediction meaningfully reduces. The insurer carries a risk it will otherwise decline to write. This asymmetry explains most of the business model confusion in the sector: ventures build a farmer-facing app, price it at what a farmer can pay, and discover that the unit economics never close because acquiring and serving a dispersed smallholder base costs more than the ticket supports, a problem the last-mile agro-dealer economics already describes well. The ventures that work invert it. They sell to the institution, embed the data into a decision that institution already makes, and reach the farmer through that relationship rather than around it, which also solves distribution, since the exporter, cooperative, or lender already has the roster the app was trying to build. The farmer still benefits, sometimes considerably, but the invoice goes to the party whose money is at risk. That is not a compromise of the mission. It is the only structure in which the service survives the end of a funding cycle.
What is the Payer Question?
The named framework of this essay: four checks that separate a satellite data business from a satellite data pilot. It is deliberately blunt, because this category attracts more good intentions than most.
- Who loses money if this data does not exist? Name the party and the amount. An exporter losing market access, an insurer unable to price, a lender writing off avoidable defaults. If the honest answer is that everyone would be somewhat worse informed, there is no customer, only an audience.
- What decision does it enter, and when? Data has value at a moment: before a planting window, before a claim is settled, before a shipment is documented. Arriving after the decision has been made is the same as not arriving. Most pilots fail here rather than on accuracy.
- What is it replacing, and at what cost? Field agents, self-declaration, manual audits, or simply refusing the business. The comparison must be against the incumbent method’s real cost, including the risk it fails to catch, not against an ideal of perfect information nobody was buying.
- Who verifies it when it is disputed? A payout denied, a lot rejected, a loan refused. Every one of these ends in an argument eventually, and a service that cannot show its method under challenge will lose the account the first time it matters. Verification is the product’s spine, not its documentation.
The typical verdict: compliance and insurance use cases pass all four comfortably, credit passes three and struggles on verification, and farmer-facing advisory passes the second and fails the first. That ordering is roughly the order in which the money is arriving.
What should founders, cooperatives, and policymakers do about it?
Build for the obligation and own the ground truth. Founders should stop competing on imagery, which is a commodity, and compete instead on the layer nobody can copy from orbit: the plot boundaries, the farmer identities, the cooperative rosters, and the historical records that turn a picture into evidence about a specific person’s specific field. That ground layer is slow and unglamorous to assemble, which is exactly why it defends a business, and it is the same insight drone economics surfaced when the hardware commoditised faster than the operating know-how. Cooperatives hold more leverage here than they exercise, because they already possess the membership data that makes compliance possible, and a cooperative that organises its own plot registry becomes the indispensable counterparty in every export deal rather than a supplier to be verified around. That is a negotiating position, and it should be treated as an asset on the balance sheet in the way carbon documentation is already becoming one. Policymakers should notice that compliance data and land records are converging, and that a public plot registry serving both would lower costs across every export chain at once, which connects the land tenure question to market access rather than only to collateral. The risk to name honestly is exclusion: a documentation requirement is a barrier, and smallholders who cannot be geolocated will be dropped from supply chains quietly by exporters minimising their own risk. Whoever solves inclusion at the registry layer will hold the most valuable position in African agriculture within a decade, and will have built it, appropriately, from the ground rather than from the sky.
FAQ
Is satellite imagery expensive?
The public constellations publish openly and cost nothing to access. Commercial high-resolution imagery costs money, but most agricultural use cases run on free data. The expense is analysis, ground truth, and delivery.
Why do farmer-facing agriculture apps struggle?
Because acquisition and service cost more per farmer than a smallholder can pay for advisory alone. The ones that work reach farmers through an institution that is already paying for a different reason.
What is driving demand right now?
Export compliance, mainly deforestation and due-diligence rules, plus parametric insurance and agricultural lending. Regulation is the strongest single driver because it converts a preference into a requirement.
What is the defensible asset in this business?
Ground truth. Plot boundaries, verified farmer identity, and history tied to specific fields. Imagery is available to everyone; the link between a pixel and a named person’s field is not.
What is the main risk to smallholders?
Exclusion. If documentation becomes a condition of market access, undocumented farmers get quietly dropped. Inclusion at the registry layer is the problem worth solving.
Related Reading
- Uganda’s Export Ban, Value Addition, and EUDR
- Agricultural Drones and the Economics of the Field
- Insurance as a Greenfield Market in East Africa
- AI, Agro-Dealers, and the Last Mile of Agriculture
Sources and Evidence
- Copernicus Sentinel open data: the free European constellation underpinning most agricultural monitoring.
- NASA Harvest and USGS Landsat: long-run open imagery archives used for historical field analysis.
- European Union Deforestation Regulation guidance: geolocation and traceability obligations for covered commodities.
- CGIAR and index insurance research: evidence on parametric agricultural insurance design and basis risk.