HERUFI

Agricultural Capital Series · Report 2 of 3

Matching the Business to the Instrument

Five agricultural business models, the balance sheet each one has, and the capital each one is offered

1 of 5
Agricultural business archetypes with the economics venture equity requires
76%
Fall in African agritech equity from its 2022 peak, while debt reached 41 per cent of all African technology funding
29x / 8%
Leverage on the largest loans against the share of borrowers there who have never held a formal loan
3.3% / 4.6%
Non performing loans at women owned enterprises against enterprises owned by men over 35

Executive summary

Report 1 of this series established that the binding constraint at the bottom of the agricultural lending market is the cost of making a loan rather than the risk of losing it. This report asks the question that follows. Given that cost structure, what kind of capital fits an agricultural business, and what is it usually offered instead?

The capital that is most available fits the part of the sector that matters least for the outcome the sector says it wants. Of the five agricultural business archetypes set out here, one has venture economics. It is also the smallest part of the sector by revenue and the most indirectly connected to farmer incomes.

Key findings

1. Agritech equity fell about 76 per cent from its 2022 peak while debt rose. African agritech raised 168 million dollars in 2025 against 207 million in 2024, with equity falling from roughly 328 million in 2022 to roughly 80 million in 2025, a minority of sector funding for the first time. Across African technology as a whole, debt reached 41 per cent of capital deployed against 17 per cent in 2019. This is better read as capital finding the right instrument than as capital losing interest.

2. Leverage and additionality are the same fact seen twice. Aceli's data shows leverage rising from 5x to 29x as ticket size rises, while the share of first time borrowers falls from 84 per cent to 8 per cent. A facility managed on leverage will climb up market until it finances borrowers who had other options.

3. Women and young borrowers are the better credit. Non performing loans run at 3.3 per cent for women owned enterprises and 2.5 per cent for youth owned, against 4.6 per cent for enterprises owned by men over 35. They also borrow half as much, so they generate lower leverage. That is the mechanism by which a leverage target excludes them.

4. Part of the constraint is a shortage of structurers rather than a shortage of capital. The two transactions in this report that financed a balance sheet as the framework prescribes were structured by the same intermediary, and both were reported as firsts of their kind. A balance sheet with nobody able to build its instrument takes the instrument already on the shelf.

What follows from this. Report 3 holds the concessional budget fixed and models what happens to volume and to reach when it is reallocated from cover against credit losses to the cost of origination.

The matching problem

2.1 What venture capital requires, and what agriculture has

Venture capital needsAgricultural enterprises have
Margins that scale without proportional costTrading and processing margins of 5 to 20 per cent
Growth non linear in capitalGrowth proportional to hectares, tonnage and working capital
A liquidity event inside the fund lifeThin trade sale markets, almost no listings
Low working capital intensityWorking capital intensity as the defining feature
Risk concentrated in adoptionRisk in weather, price and counterparty

This does not make them bad businesses. It makes them bad venture equity, which is a different judgement. A processor running a 12 per cent margin and growing 40 per cent a year on rising tonnage is an excellent credit and a poor venture bet, and both are true at once.

2.2 Five archetypes, five balance sheets

Business modelWhat the balance sheet isCapital that fitsCapital usually offered
Input distributorInventory, receivablesInventory and receivables finance, local currency, revolvingGrowth equity priced on a software multiple
Offtaker and processorPlant, warehouse stock, offtake contractsWarehouse receipt and trade finance, plus long dated debt for plantVenture equity expecting non linear growth
Embedded agricultural lenderReceivables portfolioOn lending facility, first loss on the portfolio, equity only for the platformEquity funding the loan book itself
Digital advisory and marketplaceSoftware, farmer networkVenture equity, and this is the one case where that is trueDebt against revenue with no asset behind it
Asset financier and mechanisationThe financed assetLong dated local currency debt, securitisation once a track record existsFive to seven year fund money against a ten year asset
Figure 1
The author's assessment of fit between five agricultural business models and six sources of capital, scored none, narrow, partial or strong. This is a judgement, not a dataset, so every cell carries its reasoning: hover any square to read why. The dots repeat the score for readers who cannot rely on colour.

Only the digital advisory archetype has venture economics. It is also the smallest part of the sector by revenue and the most indirectly connected to farmer incomes. The capital that is most available fits the part of the sector that matters least for the outcome the sector says it wants.

Documented cases against each archetype. The framework is testable against the public record, so it is tested here. Attribution is deliberately weak: none of these companies succeeded or failed for one reason, and none of them publishes audited profit figures, so "financed well" below means financed in a way the balance sheet can service, not demonstrated profitability.

ArchetypeCaseWhat happenedReading
Input distributoriProcure (Kenya)Raised about USD 17m across debt and equity, reached more than one million farmers, entered administration April 2024 after failing to close a new round. Reported unpaid debts were modest against the total raised.A seasonal inventory business was funded by rounds rather than by a facility that breathes with the season. A round that does not arrive is not a facility.
Offtaker and processorTwiga Foods (Kenya)Raised more than USD 180m, built owned logistics. Cut 283 jobs in 2023, 59 in 2024, over 300 in 2025 alongside a holding company restructure and an announced shift to an asset light model. A subsidiary entered administration in August 2026.The asset light pivot is the company's own verdict that the balance sheet and the instrument had come apart. Other factors, including an expensive logistics lease, are documented.
Embedded agricultural lenderApollo Agriculture (Kenya)May 2026: Kenya's first private agriculture focused securitisation. KES 276m first tranche against a KES 370m receivables pool covering 23,839 farmers; IDH Farmfit Fund anchor, Kaleidofin structurer; programme target KES 2.37bn.The loan book funded by debt secured on the loan book, in the currency of repayment, with equity reserved for the platform. The archetype financed as the framework prescribes.
Digital advisory and marketplacePula (Kenya)USD 20m Series B, April 2024, led by BlueOrchard with the IFC and the GAFSP Private Sector Window. Distributes through more than 100 partner organisations rather than direct to farmers.Venture shaped economics, funded by impact and development investors rather than by classic venture funds. The instrument fits; the investor base is not the one the archetype predicts.
Asset financierSunCulture (Kenya)September 2026: USD 10m senior secured facility from the Mirova Gigaton Fund into an SPV acquiring SunCulture Kenya receivables, structured by Kaleidofin. A USD 15m Bridgin receivables facility was announced in September 2025.Growth constrained by receivables financing capacity, then relieved by financing the receivables. Equity funds the product, debt funds the assets.

One observation the funding totals do not surface. Kaleidofin structured both the Apollo and the SunCulture transactions, and both were reported as firsts of their kind. The binding constraint was not only the quantity of capital. It was partly the absence of intermediaries able to build the instrument a given balance sheet needed, which left that balance sheet taking the instrument already on the shelf.

2.3 The supply side and its binding constraints

ProviderBinding constraintReaches smallholders
Venture capitalNeeds a power law outcome and an exit inside the fund lifeOnly via the advisory archetype
Impact and patient equityStill needs an exit; agricultural trade sales are thinYes, but the cheque is too large for most
Local commercial banksOrigination cost at small tickets, and collateral rulesWhere an incentive changes the arithmetic
Specialist and social lendersHard currency funding against local currency revenueDirectly, via the offtaker channel
Development finance institutionsMinimum ticket size; diligence cost per dollar deployedOnly at two removes
Donors and philanthropyBudget, and a habit of funding risk cover not transaction costYes, and uniquely able to pay for origination

Read down that column and a division of labour appears that nobody designed. Only the local bank can reach the small end. Only concessional capital can pay for the thing that stops it.

2.4 What makes these models commercially viable

Four mechanisms, each an attack on the origination cost curve rather than on credit risk:

1. Aggregation. Underwrite the cooperative, outgrower scheme or supplier book rather than the farmer. One credit file, many borrowers. The largest single lever. 2. Embedded distribution. Originate inside a transaction that is happening anyway. The acquisition cost is already sunk by the commercial relationship. 3. Data substitution. Replace the field visit with offtaker purchase history, satellite yield estimates, mobile money flows or warehouse receipts. 4. Portfolio underwriting. Price a book rather than a name. Flat, predictable credit loss across the size range is exactly the condition under which this works.

And one that is not about cost:

5. Local currency funding. TCX research suggests one in eight low income countries sees a currency fall of 20 per cent or more in any given year. Of Aceli's 55 lenders, 45 are African domiciled and originated 96 per cent of loans in local currency, against a pre launch expectation that international impact investors would do half the volume.

Uganda in 2026 is the worked example. A borrower earning shillings and servicing a dollar facility saw that obligation rise by roughly a quarter in local currency terms between March and late October, with no change to the facility and no change to the business. The same event then raised the price of the local currency alternative, because the policy rate rose from 9.75 to 12 per cent in defence of the currency. Currency mismatch and opportunity cost are not two separate risks here. They are one event arriving twice.

2.5 The trade off nobody prices

Loan sizeFirst time borrowersLeverage
10k to 49k84%5x
50k to 99k35%7x
100k to 249k26%10x
250k to 499k16%19x
500k to 1.75M8%29x
Figure 2
Each point is one loan size band in Aceli Africa's portfolio. The vertical axis is the share of borrowers who had never held a formal loan. The horizontal axis is leverage: for every dollar of subsidy spent, how many dollars the lenders put in from their own balance sheets. The dotted path runs from the smallest band to the largest.
Figure 3
Nine metrics from Aceli Africa's first five years, each expressed as a percentage of what it projected, so measures on nine different scales can be compared. Two lines fall, and both are about money. Every line about reach rises.

Leverage, defined precisely, because the term is used loosely. Aceli spends donor money on two things: an origination incentive paid per qualifying loan, and cover against a share of losses. The loans themselves are funded by the lender from its own deposits and capital. Leverage is the ratio of the second to the first: USD 1m of donor spend against USD 10m of lender balance sheet deployed is 10x. "Mobilised commercial capital" therefore means capital that must be repaid to a commercial lender, not granted by a donor.

Leverage and additionality are mirror images. Large loans mobilise more capital per concessional dollar because they go to borrowers who were already bankable.

Aceli's results against its own projections read as two misses and several large beats:

MetricProjectedActual
Lenders2055
Loans1,5005,576
Average loan size400k79k
Capital mobilised600m440m
Leverage ratio15x10x
First time borrowers35%63%
Farmers and workers reached1.1m2.3m

The two misses are the same sentence as the beats. 10x remains more than double the roughly 4x average Convergence found across 340 blended finance transactions globally.

2.6 Who is actually the risk

Women ownedYouth ownedMen 35 and over
Average loan49k36k95k
First time borrowers77%79%56%
Non performing loans3.3%2.5%4.6%
Leverage ratio7.06.010.5
Figure 4

Average loan taken

Loans going bad

Never had a formal loan

Agricultural SME borrowers in Aceli Africa's portfolio, by who owns the business. Three measures on three scales, each panel labelled with its values. Read the middle panel, then the left one: that is where the trap closes.

The under served segment is the better credit. Because it borrows less, it produces lower leverage, which is how a leverage target excludes it without anyone deciding to. Aceli notes it is aware of no other sex or age disaggregated loan performance data at SME scale anywhere, which means the sector has been allocating on an untested assumption.


Appendix A. Method

How the archetypes were built. The five archetypes are a reading of what agricultural enterprises hold on their balance sheets, set against what venture equity requires. They are a framework for matching an instrument to an asset, not a taxonomy with a claim to completeness, and a real company can sit in two of them at once.

How the cases were selected and read. Each case is a dated, publicly reported event at a named company, chosen because the company is a clear instance of its archetype. Attribution is deliberately weak. None of these companies succeeded or failed for one reason, none publishes audited profit figures, and no causal claim is made from a financing structure to an outcome.

Appendix B. Limitations

1. One programme, five countries, one vintage. Aceli's benchmark data covers 2020 to 2023 in relatively deep regional financial markets. West and Central African lender economics are likely worse, which would strengthen the direction of the argument while making the levels wrong.

2. No view on any individual company or fund. The archetypes are a framework, not an assessment.

3. The loan performance comparison sits inside a screen. Aceli's figures come from loans actually issued to borrowers who passed a credit screen. The comparison by sex and age therefore says which screened borrowers performed better, not how the unscreened population would perform.

Appendix C. What would falsify this

  • A sharp recovery in agritech equity returns, suggesting the 2022 to 2025 retreat was a cycle rather than a correction.
  • An agricultural processor or input distributor delivering venture scale equity returns, which would mean the balance sheet reading here is wrong.

A shorter explainer of this report is published as Good Business, Wrong Money. Sources are listed in full in SOURCES.md. Data, model and notebooks are open and reproducible.