Two companies raised money this month to make smallholder farming legible, one by satellite and one by registry. Making an action observable changes who can be paid for it. It does not settle who gets paid, and that is a separate question with a separate answer.
Two financings this month make the same underlying bet on smallholder agriculture, from opposite ends.
In Ghana, Sommalife, now trading as Vitara, raised backing from Acumen for its TreeSyt platform, which registers farmers and tracks crops from farm to buyer. The company reports 150,000 farmers registered and 40,000 active suppliers, with about 90 per cent of the rural farmers it aggregates from being women, and a target of linking 60,000 farmers to markets with this funding on the way to a million by 2030.
In Paris, CarbonFarm closed an oversubscribed €7 million Series A led by Colam Impact with Proparco among the new investors. The company runs satellite-based monitoring, reporting and verification for rice methane, combining AI-powered satellite imagery, field data collection and greenhouse gas modelling to quantify emissions at field level, and has expanded into 19 countries since its previous round.
Both are selling the same thing, which is not software. It is the conversion of an unobservable action into an observable one. A farmer who drains a paddy at the right moment, or sells nuts she actually collected from trees she actually tends, is doing something that was previously impossible to verify at a distance and is now, to a useful approximation, visible.
The interesting question is what that changes, and the answer is narrower than the pitch decks suggest.
Holmström's 1979 analysis of moral hazard and observability gives the precise result. Where a principal cannot observe an agent's action, the contract has to be written on something correlated with it, usually an outcome, and the agent ends up bearing risk they cannot control. The informativeness principle follows: an additional signal should enter the contract if and only if it carries information about the action beyond what the existing measures already convey.
Satellite verification of alternate wetting and drying is close to a textbook case. The outcome, methane abated, is invisible, delayed and confounded by weather. The action, draining a field on a schedule, is the thing the farmer controls. A signal that reports field water conditions directly is therefore informative in exactly Holmström's sense, and it permits a contract written on the action rather than on a noisy outcome. That is a real gain and it is why the capital is arriving.
What the result does not say is anything about the split. Holmström's framework establishes that a better signal expands the set of feasible contracts and reduces the risk premium the agent must be paid. Who captures that gain is determined by bargaining power, outside options and competition among principals, none of which a satellite observes.
This is the gap between the two claims that get run together in this market. Verification makes a payment possible. It does not make it generous.
Three things sit outside what any verification layer resolves, and all three decide the outcome.
Who holds the contract with the buyer. A traceability record proves provenance. It does not confer the right to negotiate. If the farmer's counterparty is still a single aggregator, better information about her crop strengthens the aggregator's position in the downstream negotiation at least as much as it strengthens hers in the upstream one.
Who bears the cost of being measured. Registration, phone ownership, data collection and the labour of complying with a protocol are real costs borne locally. In Holmström's terms they are a reduction in the agent's payoff that must be compensated before any of this is worth doing, and they are usually treated as a project overhead rather than as a farmer cost.
What happens to the farmers the system excludes. A registry of 150,000 with 40,000 active suppliers is, read the other way, a system in which most registered farmers are not currently selling through it. Verification regimes concentrate buying on the verifiable, and the residual market for everyone else can get thinner and worse rather than staying as it was.
CarbonFarm's own chief executive puts the first of these more usefully than most of the sector does, arguing that scale will not come from replacing local value-chain players but from equipping them. That is a claim about where the contract sits, and it is the right thing to be arguing about.
Development finance has attached risk-sharing guarantees to several of these structures, and that detail deserves more attention than the equity cheque.
A guarantee is a statement about what the signal cannot yet do. If verification were sufficiently informative and sufficiently trusted, the project would be financeable on its merits and a guarantee would be redundant. Its presence prices the residual uncertainty, and it indicates that buyers of these credits still do not fully believe the measurement, or do not believe the institutional chain carrying it.
That makes the guarantee a measurable thing in its own right. Its size, relative to the project, is a running estimate of how much of the verification problem is unsolved. A sector in which that share falls over time is one in which measurement is genuinely maturing. A sector in which it stays flat while deal volume rises is one in which public money is substituting for trust rather than building it.
Four quantities, none of which appear in current reporting, and all of which are obtainable.
The farmgate share. What proportion of the final sale price reaches the producer, before and after verification is introduced, measured against a comparison group rather than against the same farmers' recollection of last year.
The cost of compliance, borne locally. Hours, airtime, travel and equipment, converted into a share of the price received. This is the deduction that makes the difference between a scheme that pays and one that merely registers.
Payment latency. Carbon payments in particular arrive long after the practice change, and a farmer financing a behaviour change out of her own working capital for a year is lending to the scheme at a rate nobody has calculated.
Participation and attrition by year. Not registrations. The share of registered farmers transacting in year one, year two and year three. The gap between 150,000 and 40,000 is the single most informative number in either of these companies' disclosures, and its movement over time is the real test.
For DFIs, the conclusion is that the guarantee, not the equity, is the instrument that reveals whether verification is working, and its trajectory should be reported as a sector indicator. For project developers, it is that informativeness and fairness are separate properties and only the first is engineering. For agribusinesses, it is that a traceability system changes your negotiating position as well as your compliance position, and the two may not point the same way.
The satellite resolves the field to a few metres. What it cannot see is the conversation at the buying station, and that is where the price is still decided.
If this reads close to a decision you are weighing, we would like to hear about it. A short note is enough to start.
The Lab is an independent research team, not an advocacy organisation. We do not campaign, endorse products, or accept commissions where a finding is expected in advance. Where a partner's own theory of change is what needs testing, that is a legitimate engagement, and the Lab tests it honestly.