Following the evolution of climate adaptation through the businesses closest to the farmer

There is a familiar way of describing the climate adaptation challenge in agriculture.

Climate risks are increasing. Smallholder farmers are highly exposed. Solutions already exist: drought-resilient seeds, solar irrigation, cold storage, weather advisory, water-management systems, renewable-energy technologies and better post-harvest infrastructure. Yet adoption remains low.

From there, the discussion often moves quickly to finance.

Farmers cannot afford the technologies. Local enterprises cannot finance inventory. Banks are reluctant to lend. Adaptation capital is insufficient. Therefore, more finance is required.

There is truth in each of these statements.

But the explanation feels incomplete.

Consider a solar cold room placed in a horticulture-growing district.

The technology works. The temperature inside the room falls. Produce lasts longer.

Yet before that cold room becomes useful to a farmer, somebody has to decide where to place it. Somebody has to convince farmers to use it. Produce has to arrive at sufficient volume. It has to arrive in the right condition. Someone has to grade and record it. A buyer has to exist. That buyer has to collect at the right time. Farmers have to believe that waiting rather than selling immediately will leave them better off. The machine has to be maintained. And enough transactions have to take place throughout the year for whoever operates the cold room to remain in business.

The refrigerator may be proven.

The business around the refrigerator may not be.

That distinction is important.

It suggests that one of the central adaptation questions is not simply whether we have technologies that can help farmers adapt.

The harder question is whether we have built the businesses, incentives and market arrangements through which those technologies can become normal parts of farming.

This article explores that question by following the evolution of the last mile.

Not just the evolution of technology.

The evolution of who performs the work around the technology, who gets paid for it, who carries the risk when something goes wrong, and why the local MSME increasingly finds itself at the centre of all of this.

---

1. Start with the farmer, but do not stop there

A smallholder farmer does not normally wake up wanting “climate adaptation.”

The farmer wants something more immediate.

Water when the rain does not arrive. Seed that performs in a shorter growing season. More time to sell tomatoes before they spoil. Information about whether planting this week is wiser than waiting. Protection against losing the money invested in inputs.

Climate adaptation is the name we give to the broader outcome.

For the farmer, the decision is much narrower.

Does this particular solution improve my position enough to justify its cost, uncertainty and effort?

That immediately creates an important distinction between need and demand.

A farmer can urgently need irrigation and still not be a viable customer for a solar pump.

There may be no reliable water source. The farm may be too small. The crops grown may not generate enough incremental income. The farmer may have access to water but no reliable market for additional dry-season production. The pump may technically solve the rainfall problem while creating a repayment problem.

So the journey does not really look like climate problem → technology → farmer.

It is closer to climate problem → specific farm need → potentially relevant solution → commercially usable offer → trusted local delivery → farmer decision → correct use → complementary services → economic benefit → repeat use.

Every transition matters.

And increasingly, it is the local enterprise that is expected to make those transitions work.

2. The MSME is not really a distributor

The phrase “last-mile distribution” can be misleading.

It makes the task sound like logistics.

There is a product somewhere. There is a farmer somewhere else. The MSME’s role is to move the first to the second.

For many adaptation solutions, that is a very small part of the job.

Take solar irrigation.

Someone may need to identify farmers who could actually benefit, understand the water source, size the pump, explain the economics, arrange finance, transport and install the system, train the farmer, monitor its functioning, manage repayment, diagnose faults, source spare parts and process warranty claims.

Take climate-resilient seed.

The local business has to decide which varieties to stock before the rains arrive, when final farmer demand is still unknown. It may have to demonstrate an unfamiliar variety, explain why it differs from farm-saved seed, carry inventory risk and somehow ensure that stock arrives during a narrow planting window.

Or consider climate information.

Nothing physical needs to be transported. But somebody still has to turn a forecast into a farmer decision.

“Rainfall is likely to be below normal” is information.

“Delay planting by ten days, use this variety and reduce the first fertilizer application” is closer to a usable service.

That requires interpretation, language, local knowledge and trust.

The local MSME is therefore better understood not as a retailer but as the actor trying to convert a technical capability into a usable local offer.

That conversion is expensive.

And historically, markets have been poor at deciding who should pay for it.

3. How did one small business end up doing all this?

This is where evolution becomes useful.

Today’s last-mile structure was not designed on a blank sheet of paper.

It emerged from earlier agricultural systems.

And those systems were built to solve different problems.

For much of the twentieth century, governments played a major role in agricultural research, extension, irrigation, seed systems and sometimes crop purchasing.

The structure was often vertical.

Researchers developed improved practices. Extension systems explained them. Public agencies distributed certain inputs. Marketing boards or cooperatives aggregated selected crops. Irrigation departments constructed schemes. Export systems coordinated logistics for high-value commodities.

Whatever their weaknesses, these arrangements contained many functions that today are casually placed under “last-mile delivery.”

The local retailer did not necessarily have to create the entire market.

Other institutions were already paying for parts of it.

Then the architecture changed.

4. Liberalisation created sellers faster than it created service markets

From the 1980s onward, governments across many African economies reduced direct involvement in agricultural markets.

There were good reasons.

Large public distribution and extension systems were expensive. They were often bureaucratic, inconsistent and poorly responsive to local needs.

Private markets could potentially stock products more efficiently and respond better to farmer demand.

Private seed businesses emerged. Agro-dealer networks expanded. Equipment distributors entered. Traders and aggregators grew.

But something subtle happened during this transition.

Functions that could be attached to a sale found natural commercial owners.

Functions that could not easily be charged for often did not.

A dealer can earn a margin on seed.

Who pays the dealer to spend three afternoons demonstrating a new drought-resilient variety?

A distributor can earn from selling a pump.

Who pays for travelling to farms that eventually turn out to have unsuitable water sources?

A technology provider can charge for hardware.

Who maintains a competent technician in a district with only 20 installations?

A farmer may value weather information.

Who pays for transforming a national forecast into local-language, crop-specific advice?

These functions did not disappear when public systems weakened.

They simply became harder to see.

Many eventually landed with the actor nearest the farmer.

The MSME.

This is one reason I find the phrase “missing last mile” slightly misleading.

The last mile is not always missing.

Sometimes it is present but economically overloaded.

5. The margin pays for the product. Who pays for creating the customer?

Imagine a rural agro-dealer.

A familiar seed variety sells regularly. Farmers know it. The dealer understands approximately how much to stock.

A newly released climate-resilient variety is different.

It may genuinely perform better under emerging climatic conditions. But the farmer does not know that yet.

Someone has to explain it.

Perhaps run demonstration plots. Perhaps allow farmers to observe it for one or two seasons. Perhaps explain why “drought tolerant” does not mean “drought proof.”

The dealer bears some of these costs today.

The value appears later.

And once farmers understand the new variety, any competing dealer may benefit.

The person paying for the learning cannot capture all of its benefit.

So the rational dealer often stocks what already moves.

The pattern repeats across adaptation.

The market needs somebody to create trust before sufficient demand exists.

But businesses normally want sufficient demand before investing heavily in creating trust.

This creates a chicken-and-egg problem.

Low adoption means few demonstrations. Few demonstrations mean weak local evidence. Weak evidence makes farmers cautious. Caution keeps adoption low.

6. Small transaction. Large delivery job.

There is another structural problem.

The value of individual farmer transactions is often small.

The work surrounding those transactions is not.

An agent may travel two hours to register ten farmers. A technician may travel three hours to repair one pump. A demonstration plot costs almost the same whether twelve or eighty farmers eventually purchase seed. A cooling operator needs staff whether the room contains 30 crates or 300.

This changes how we should think about the “last mile.”

Distance matters.

But density often matters more.

Suppose an enterprise can serve 150 customers within a 10-kilometre radius.

The same enterprise serving 150 customers scattered across 150 kilometres has a completely different business.

Same technology. Same number of farmers. Same climate need.

Very different economics.

That is why some solutions scale first in high-value horticulture clusters, organised dairy systems, major markets or strongly aggregated farmer groups.

It is not necessarily because climate risk is greater there.

The commercial density is greater.

The route can support itself.

7. Technology adoption has always followed infrastructure around the technology

We sometimes talk about innovation as though good technologies naturally diffuse.

History suggests otherwise.

Railways became more useful where feeder roads connected to stations.

Electricity became transformative when appliances, wiring, repair services and payment systems developed around it.

Mobile phones became dramatically more useful when mobile money, agent networks, airtime distribution and affordable handsets evolved around the network.

The same is true for agricultural adaptation.

A solar pump is not really a standalone product.

It sits inside a system containing water availability, installation, irrigation practice, crops, labour, finance, maintenance and output markets.

Cooling sits inside harvesting, grading, aggregation, storage, transport, buyer standards and payment.

Climate advisory sits inside forecast production, localisation, interpretation, farmer trust, available choices, ability to act and the eventual market outcome.

The solution is often only one component in a chain.

The farmer experiences the performance of the whole chain.

8. This is why technically successful technologies can look commercially unsuccessful

Consider cooling.

SokoFresh’s experience in Kenya is particularly revealing.

A cold room can maintain temperature perfectly and still fail to produce a successful transaction.

If mangoes enter the system already over-mature, cooling cannot make them export quality again.

If the buyer rejects them, refrigeration technically performed its function.

Commercially, the chain failed.

The company’s evolution therefore moved beyond “renting cold storage.”

It increasingly incorporated aggregation, harvesting protocols, logistics and market linkage.

That evolution matters.

The solution did not improve only because the refrigerator improved.

The business expanded its boundary until it controlled more of the conditions required for the refrigerator to create value.

This is a recurring feature of successful adaptation businesses.

They start by solving one technical problem.

They discover that their economics depend on neighbouring problems.

Eventually they bundle several of those problems together.

This is not mission creep.

Often it is the business model discovering what the actual product is.

9. Perhaps the farmer was never buying cooling

A farmer paying for cold storage is not really purchasing a lower temperature.

The farmer is buying time.

More precisely, the farmer hopes that additional time will create a better commercial outcome.

That might mean less spoilage, a later sale, access to a different buyer, a higher price or the ability to aggregate enough produce for transport.

If none of those things happen, the farmer may rationally stop paying for cooling.

This sounds obvious once stated.

But it changes the business model.

The customer’s value proposition is not the physical function.

It is the economic consequence of the physical function.

The same applies elsewhere.

A farmer does not really buy irrigation. The farmer buys the possibility of more reliable or differently timed production.

A farmer does not really buy climate information. The farmer buys a better decision.

A farmer does not really buy drought-tolerant seed. The farmer buys a different distribution of possible harvest outcomes.

Once we frame solutions this way, many adoption failures become easier to understand.

10. Farmer economics and MSME economics are two separate equations

Suppose a cooling service saves a farmer produce worth $8.

The service costs the farmer $3.

That is a good transaction for the farmer.

But suppose serving that farmer costs the operator $4 once labour, handling, customer acquisition, maintenance and asset costs are allocated.

Then the same transaction is bad for the enterprise.

Farmer value exists.

MSME viability does not.

The reverse can also happen.

A distributor may earn an attractive margin on a product that farmers purchase through subsidy or aggressive credit.

But if the farmer’s eventual benefit is weak, repurchase will disappear.

The MSME can look viable during the acquisition phase while the underlying customer proposition is unstable.

Both equations therefore have to close.

For the farmer, the real question is whether the expected incremental benefit is large enough after price, additional labour, behavioural change and perceived risk are taken into account.

For the enterprise, the real question is whether revenue covers product or asset cost, customer acquisition, logistics, service, working capital, downtime, default and seasonality.

A scalable adaptation market requires both sides to remain positive.

Climate impact does not automatically make either equation work.

11. The third equation: timing

Agriculture adds another complication.

Cash does not arrive when costs arrive.

Seed must be stocked before planting.

A solar pump may be purchased months before crops are sold.

A cold room incurs costs throughout the year despite seasonal utilisation.

An aggregator may pay farmers today and receive buyer payment 30 or 60 days later.

So even businesses with theoretically positive margins can fail.

The question becomes: how long does someone have to finance the gap between doing the work and receiving the value?

This is where working capital matters enormously.

But it is also where finance is frequently misdiagnosed.

Working capital can solve the problem of a dealer who knows that viable inventory will sell in three months but needs money to buy it today.

It cannot solve the problem of a dealer who does not know whether anyone wants the inventory.

Credit can solve the problem of a productive asset whose customer cannot pay the full price upfront.

It cannot solve the problem of an asset whose productive use has not been established.

Finance can bridge time.

It cannot manufacture economics.

That distinction may be one of the most important ones in adaptation finance.

12. Why giving everybody cheaper money can sometimes make things worse

Suppose a farmer cannot afford a solar pump.

We provide cheap credit.

Adoption rises.

That looks like success.

But suppose some farmers subsequently discover that their water source is unreliable, the productive season is shorter than expected, the market for vegetables is weaker than projected, the pump needs repairs or crop income arrives less regularly than monthly repayments.

The finance did not solve the economic problem.

It allowed the economic problem to surface later.

This does not mean financing is unimportant.

It means financing must follow a serious assessment of how value is created.

A financing constraint and a weak economic proposition can produce similar visible symptoms—low sales, insufficient inventory or delayed expansion—but require completely different interventions.

13. Services evolve differently from products

Another reason the MSME question is difficult is that different adaptation solutions create very different business models.

Seed is a seasonal inventory business. The dealer earns once per sale. Trust, timing and stock risk matter.

Irrigation equipment is a higher-value transaction, but customer screening, installation, financing and after-sales service become much heavier.

Cold storage is a utilisation business. The asset remains with the operator and the enterprise survives only if enough paid usage occurs.

Climate advisory is an interpretation business. Physical distribution becomes cheap, but human explanation and local relevance remain expensive.

Insurance is a risk-pooling business. The visible local transaction may be small, but the underlying financial architecture sits elsewhere.

Treating all these as “last-mile distribution” obscures more than it reveals.

The core question is simpler and more useful:

What recurring work does this solution create, and is anyone paid adequately to perform it?

14. Climate information gives us an unusually clean example

Digital technology was supposed to transform agricultural advisory partly because information could reach millions of farmers at extremely low marginal cost.

And it has transformed reach.

But reach and use are not identical.

A forecast can arrive on a phone without changing a single farm decision.

Someone still has to make it relevant.

What does a 60% probability of below-normal rainfall mean for a maize farmer here?

Should planting be delayed?

Should a different variety be chosen?

Should fertilizer application change?

These decisions require context.

The deeper issue is that “weather information” may never have been the natural commercial unit.

The useful unit may instead be weather plus agronomy, weather plus insurance, weather plus credit, weather plus market information, or weather embedded inside a buyer or input relationship.

The technology becomes viable by becoming less standalone.

15. This keeps happening: successful models expand sideways

Look at several adaptation businesses over time and a pattern emerges.

Cold storage becomes cooling plus aggregation.

Irrigation becomes hardware plus finance plus installation plus support.

Seed becomes product plus demonstration plus advisory.

Climate information becomes forecasts plus interpretation plus other farm services.

Insurance becomes embedded in input, lending or value-chain transactions.

Why?

Because farmers experience outcomes rather than technologies.

And outcomes are produced jointly.

The business initially thinks, “We sell X.”

Reality eventually teaches it, “Our customer receives value only when X works alongside A, B and C.”

The stronger models then ask a harder question.

Which of A, B and C must we control, which can a partner control, and who should pay for them?

That is business-model evolution.

16. Some problems were solved by changing who owns the asset

Cold storage provides one example.

Selling a $20,000 or $30,000 cold room to individual smallholders is usually unrealistic.

So businesses such as ColdHubs and SokoFresh changed the unit of consumption.

The farmer does not buy refrigeration equipment.

The farmer buys cooling by crate, kilogram or period of storage.

A large capital expenditure becomes a small operating expense.

This solves one problem extremely well.

Affordability.

But notice what happened to the risk.

It did not disappear.

It moved.

The farmer no longer carries asset-ownership risk.

The operator does.

If the room remains empty, farmers lose nothing.

The operator still pays.

This is the basic trade-off in many as-a-service models.

The more flexible the proposition becomes for the farmer, the more utilisation risk migrates toward the enterprise.

That means the next problem becomes volume.

So the business evolves again.

The room is placed closer to produce markets. Traders are brought in. Produce is aggregated. Crops with complementary seasons are combined. Buyers are linked. Transport is added.

Each adjustment is an attempt to make utilisation less random.

17. Put the asset where transactions already happen

This is another useful principle.

Innovation programmes often ask:

Where is climate need greatest?

A business has to ask another question:

Where can a repeatable transaction happen?

These locations overlap, but not perfectly.

A cold room in the most climate-vulnerable village may have insufficient throughput.

A cold room inside a busy produce market may serve more traders than farmers but operate far more regularly.

That creates uncomfortable trade-offs.

The commercially strongest starting point may not be the most vulnerable customer.

But the market may need to develop from viable nodes outward.

This is how many infrastructure systems historically expanded.

Dense usage supported early economics. Capability accumulated. Networks gradually extended.

Trying to begin with the hardest-to-reach user can sometimes force permanent subsidy rather than produce a scalable delivery system.

There is no universal answer here.

But pretending the trade-off does not exist makes programme design weaker.

18. After-sales service exposes one of the deepest MSME problems

Now imagine that the pump has been sold.

Everyone celebrates adoption.

Three months later it fails.

Who owns the problem?

The farmer calls the local dealer.

The dealer may not have the spare part.

The distributor may not have authority to open the equipment without affecting the warranty.

The manufacturer may sit in another country.

A technician may need to travel hundreds of kilometres.

Suddenly the “last mile” runs in reverse.

And the adaptation benefit disappears until that reverse chain works.

A pump that operates 11 months a year may be highly useful.

The same pump failing during the ten days when seedlings desperately need water can destroy an entire season.

So uptime matters more than annual averages suggest.

This is why some businesses have evolved from selling hardware toward building repairable products, distributor training, spare-parts systems, remote diagnostics and explicit warranty reimbursement.

The key innovation is not merely “train technicians.”

It is:

Give the technician parts, authority, information and a way to get paid.

Capability without an economic role rarely survives.

19. Low adoption and poor service can reinforce one another

This produces another trap.

Few farmers purchase the technology.

Therefore the district has few repair jobs.

Therefore maintaining specialised technicians and spare parts locally is unprofitable.

Repairs become slow.

Farmers hear about failures.

Purchasing the technology begins to look riskier.

Fewer farmers adopt.

The market never reaches sufficient density for good service.

This is a classic coordination problem.

Each participant is waiting for conditions that another participant can only justify after adoption increases.

This may be one of the circumstances in which temporary external support is genuinely useful.

Not to subsidise the system forever.

But to help it cross the minimum-density threshold where commercial service becomes possible.

The question for development finance then changes from:

How much subsidy does this business need?

to:

Which temporary cost must be absorbed until enough transactions exist for the market to carry it itself?

Very different question.

20. Projects solved access, but sometimes froze the market in pilot mode

There is another piece of evolution we need to confront.

When markets could not deliver new technologies quickly, governments and development organisations understandably stepped in.

They subsidised equipment, distributed inputs, paid for demonstrations, financed pilots, trained technicians, covered extension and purchased products in bulk.

These interventions produced enormous value.

Many adaptation markets would not exist without them.

But they also shaped incentives.

An enterprise may discover that one institutional contract is worth more than acquiring 2,000 farmers individually.

Farmers may reasonably wait because similar products were previously subsidised.

A technology provider can prove reach inside a funded geography without learning what customers will repurchase independently.

A technician may be trained during a three-year project but encounter too few paid jobs after the project closes.

The pilot succeeds.

The market disappears.

This is why an installation count can be dangerously reassuring.

A project can create a delivery event without creating a delivery economy.

21. Evolution teaches us to ask what remains after support disappears

Suppose a programme supports 1,000 solar pumps.

Instead of asking only whether all 1,000 were installed, ask what happens next.

Who finds customer 1,001?

Who demonstrates the pump?

Who finances the distributor’s inventory?

Who services pump 743 in year four?

Who pays the technician?

Where are spare parts held?

How is an unsuitable farmer screened out?

What happens during a bad harvest?

Who carries default?

If the answer to several of these questions is still “the programme,” then scale has not yet been transferred to the market.

This does not mean the programme failed.

Perhaps market formation was never its objective.

But it does mean we should describe the achievement accurately.

Distribution is not necessarily adoption.

Adoption is not necessarily effective use.

Effective use is not necessarily a sustainable delivery market.

22. High-value agricultural chains solved many of these problems long ago

This is where history becomes particularly interesting.

Look at highly coordinated agricultural export chains.

Exporters often control or strongly influence seed, agronomy, quality, harvesting, aggregation, packhouses, cooling, transport, certification and market access.

Why?

Because the buyer cannot tolerate one component failing independently.

If European supermarkets require precise quality and delivery schedules, leaving every stage to loosely coordinated spot-market transactions creates too much risk.

So successful chains internalised coordination.

They did not build a cold room and hope the rest of the market organised itself.

They built a system around the buyer.

The modern adaptation startup often faces a harder task.

It may introduce one technology into an open domestic value chain where no actor controls the surrounding system.

The local MSME then has to reproduce some of the coordination of a vertically integrated exporter without owning the farm, buyer, transporter or market.

That is an extraordinary amount of responsibility for a small enterprise.

23. Which is why the buyer often becomes more important than the technology

Consider irrigation.

A farmer begins producing vegetables in the dry season.

Production rises.

This is technically successful adaptation.

But everyone in the district starts producing the same vegetables.

Prices collapse.

The irrigation investment becomes financially disappointing.

The pump did exactly what it promised.

Again, the chain failed somewhere else.

This is why buyer-led arrangements can be so powerful.

A credible buyer changes the behaviour of multiple actors simultaneously.

The farmer sees a market. The MSME sees predictable transactions. The financier sees cash flow. The aggregator sees throughput. The input provider can plan demand. The buyer can communicate quality requirements backwards.

This is market pull.

The technology is no longer pushing its way towards farmers.

A commercial outcome is pulling coordinated activity toward itself.

24. But anchors create their own dependencies

It would be easy to conclude that every adaptation model needs an anchor buyer.

That would be too simple.

Anchors solve coordination by concentrating power.

One major buyer can provide predictability.

It can also withdraw, change procurement standards, renegotiate prices, reject output or change sourcing geography.

Similarly, one development programme can anchor demand.

One technology provider can anchor maintenance.

One government subsidy can anchor farmer purchases.

Anchors make markets easier to coordinate precisely because other actors become dependent on them.

So the relevant question is not merely whether an anchor exists.

It is what the anchor is solving, who becomes dependent on it, and what happens if it disappears.

25. Farmer behaviour becomes easier to understand when we stop treating caution as ignorance

One temptation in innovation work is to attribute low adoption to lack of awareness.

Sometimes that is correct.

Often farmer behaviour is more rational than it first appears.

Imagine two choices.

The first is to sell tomatoes today to the broker outside the farm gate. The price is low, but payment is immediate.

The second is to pay to store tomatoes, wait three days and hope that a better buyer arrives.

For an investor, the second choice may have higher expected value.

For a household needing food, school fees or debt repayment tonight, the first may be superior.

The broker is not merely exploiting the farmer.

The broker is providing liquidity and certainty.

Similarly, farm-saved seed may yield less, but it requires no immediate cash.

A solar pump may produce compelling lifetime savings, but debt payments create a fixed obligation against volatile agricultural income.

The farmer therefore values reversibility.

Try a little. Pay after harvest. Avoid long commitments. Maintain an alternative.

Unfortunately, MSMEs need the opposite.

Advance orders. Repeat customers. Regular utilisation. Predictable repayment.

We have found one of the central collisions in the system:

Farmer resilience often requires flexibility. MSME viability often requires predictability.

Neither side is behaving irrationally.

26. Business-model innovation is frequently about reconciling that collision

Many models that have evolved successfully can be interpreted this way.

Pay-per-use gives the farmer flexibility but leaves the operator carrying utilisation risk.

PAYGo removes the large upfront payment but shifts repayment and servicing risk to the provider.

Seasonal repayment brings payment closer to the agricultural cash cycle.

Aggregation allows farmers to remain small while demand or produce becomes predictable at group level.

Anchor buyers improve visibility of future transactions.

Embedded advisory avoids charging farmers separately for information.

Insurance moves catastrophic risk away from the individual farmer and into a pool.

Warranty reimbursement prevents local repair providers from absorbing a cost that properly belongs to the manufacturer.

Each model changes who carries what.

That is often more important than simply changing price.

27. Perhaps the real innovation is risk architecture

We usually describe innovation through products.

Solar pump.

Cold room.

Drought-resistant seed.

AI advisory.

Weather platform.

But from a market perspective, some of the most important innovations are invisible.

Who owns the equipment?

Who pays upfront?

Who gets paid later?

Who absorbs weather risk?

Who absorbs utilisation risk?

Who guarantees repairs?

Who holds inventory?

Who carries buyer rejection?

Who pays for customer education?

Who finances receivables?

Who is accountable for quality?

These are design choices.

A technology can remain identical while adoption changes dramatically because these arrangements change.

This is why adaptation requires business-model innovation as much as technical innovation.

28. And sometimes someone other than the farmer must pay

Climate information provides a particularly strong example.

A farmer may value the information and still be unwilling to pay a separate fee for it.

That does not necessarily mean the service lacks value.

It may mean the value is spread across several actors.

A buyer benefits when farmers receive better agronomy.

An insurer benefits when farmers reduce preventable risk.

An input company benefits when farmers successfully use its product.

A lender benefits when irrigation stabilises income.

A government benefits when climate shocks create less need for emergency support.

Sometimes the right business model is therefore not farmer-pay.

It may be B2B2C, buyer-embedded, finance-embedded or publicly supported.

The person receiving the service does not always have to be the person paying directly for it.

That is not automatically subsidy.

It can simply mean locating the payer where enough economic value is captured.

29. This changes how we think about the MSME financing gap

We can now return to finance.

Yes, MSMEs need working capital.

They need asset finance.

They need receivables finance.

Some need guarantees.

Some need concessional capital.

But before designing the financial instrument, we should ask what exactly is being financed.

Financing inventory with proven demand is fundamentally different from financing market discovery.

Financing a cold room with predictable throughput is different from financing an empty facility and hoping farmers appear.

Financing receivables from a credible buyer is different from financing speculative production.

Financing a distributor’s van is different from subsidising years of customer education.

Development capital is most powerful when it knows which problem it is absorbing.

Otherwise concessional finance can preserve a weak business model rather than help a stronger one emerge.

30. What some stronger models appear to have learned

Across very different adaptation solutions, a few patterns recur.

They are not universal answers, but they are useful clues.

First, stronger models tend to bundle what the outcome actually depends on. Cooling works better with aggregation. Irrigation works better with installation and after-sales. Seed works better with demonstration. Advisory often works better when connected to agronomy or finance. Production becomes more reliable when linked to buyer demand.

Second, they reduce fragmentation by aggregating before trying to serve. A village group, cooperative, produce market, trader network, processor or anchor buyer can reduce the cost of the last mile.

Third, they make risk explicit. Warranty risk, utilisation risk, buyer risk, weather risk and inventory risk do not disappear. Someone has to carry each of them.

Fourth, they treat market-building work as real work. Demonstration, farmer education, product localisation and technician development cost money. Calling them ecosystem activities does not make them free.

Fifth, they build around a transaction, not only around a need. Climate vulnerability can tell us where adaptation matters. It cannot by itself tell us where a business can survive.

And finally, they evolve. Early business models are hypotheses. Stronger businesses change channels, payment structures, service models and partnerships as they learn where the original assumptions were wrong.

31. This is where MSMEs become particularly important

Large technology companies can develop equipment.

Banks can provide capital.

Governments can build infrastructure.

Research institutions can develop agronomic solutions.

But many adaptation outcomes still depend on highly local information.

Which farmer has water?

Which crop is normally grown here?

Which trader actually pays?

Which farmer group is trusted?

Which road becomes inaccessible during rain?

Which technician can reach the village?

Which variety do farmers recognise?

Which market day produces enough throughput?

Which payment practice will farmers accept?

That information is difficult to centralise.

MSMEs possess or can develop it.

They are not simply cheap distribution channels.

They are mechanisms for contextual adaptation.

That is why strengthening MSMEs matters.

But it also explains why merely giving them products and expecting them to sell is insufficient.

32. The uncomfortable implication: not every adaptation solution should scale everywhere

Climate discussions naturally push toward large numbers.

Millions of farmers.

Thousands of installations.

National rollout.

Continental platforms.

But first-principles thinking points in another direction.

A solution may work extremely well for commercially oriented horticulture farmers within 20 kilometres of a produce market, with access to shallow groundwater, growing crops with staggered seasons, served through one aggregator and connected to a reliable buyer.

It may fail completely 100 kilometres away.

That is not necessarily a failure of the technology.

It means the relevant market is the solution–crop–geography–customer–enterprise combination, not “African smallholder farmers.”

The more precisely we identify those combinations, the stronger the route to scale becomes.

Paradoxically, trying to reach everybody immediately can prevent us from learning where the economics genuinely work.

33. Evolution before scale

This brings us back to evolution.

We often ask too quickly:

How do we scale this?

Evolution asks something else first.

What did the model have to become before it was capable of scaling?

Did the pump business have to add finance?

Did finance have to become seasonal?

Did the cold room have to become an aggregation business?

Did aggregation have to become an offtake business?

Did weather information have to become advisory?

Did advisory have to become embedded inside another transaction?

Did the seed company need local demonstrations?

Did the distributor need warranty reimbursement?

Did the farmer need pay-per-use rather than ownership?

Each adjustment contains information.

It tells us what the original model misunderstood about the system.

That is why failed pilots are valuable when studied properly.

They are not simply failed implementations.

They are experiments revealing where responsibility, cost or risk was incorrectly placed.

34. We should follow the friction

A useful way to study adaptation markets may therefore be to stop beginning with the technology.

Begin with the friction.

A cold room is empty.

Why?

Farmers do not use it.

Why?

They sell immediately.

Why?

Immediate buyers pay cash.

Why does storage not produce a better transaction?

Perhaps no committed buyer exists.

Why will the buyer not commit?

Quality is inconsistent.

Why is quality inconsistent?

Harvesting and grading are fragmented.

Who could coordinate them?

Perhaps the cooling operator.

Would the cooling fee pay for that?

Probably not.

Who benefits from better coordination?

Farmer, buyer, operator and perhaps trader.

Can the business model capture enough of that combined value?

Now we are learning something.

We have moved from:

“Farmers need cold storage.”

to:

“A commercially viable cooling service may require a mechanism to coordinate quality, aggregation, buyer demand and farmer liquidity around a shared asset.”

That is a much more useful problem statement.

35. The same grinding process works for every adaptation solution

Take solar irrigation.

Why is adoption slow?

Price.

Why is price difficult?

Large upfront cost.

Provide finance.

Now adoption should rise.

But some customers still struggle.

Why?

Agricultural cash flows fluctuate.

Change repayment schedule.

Some still struggle.

Why?

Incremental production is not sufficiently profitable.

Why?

Crop selection or output market.

Now the intervention has moved from financing technology to productive-use economics.

Then equipment fails.

Now the problem becomes after-sales.

Then spare parts are slow.

Now the problem becomes local service density.

Eventually we realise:

The business is not financing solar pumps.

It is enabling reliable irrigated production.

The pump is one component.

This is what learning slowly does.

It prevents the first visible constraint from becoming the permanent explanation.

36. Climate adaptation may therefore require a different kind of capital

If this diagnosis is right, capital for adaptation MSMEs cannot be designed only around asset purchase.

Different stages need different forms of support.

Some enterprises need inventory finance. Some need receivables finance. Some need capital to own shared assets. Some need risk-sharing mechanisms. Some need temporary support for demonstrations and customer acquisition. Some need technical assistance for unit economics. Some need finance tied to buyer contracts. Some need funding for spare-parts networks. Some need support to build management information systems before banks can underwrite them.

And some probably should not receive more capital yet.

They need to learn whether their basic value proposition works.

Providing debt before that point can make experimentation harder rather than easier.

37. The goal is not to eliminate subsidy

Another temptation is to divide models into “commercial” and “subsidised.”

Reality is messier.

Roads are publicly financed.

Agricultural extension is often publicly financed.

Meteorological systems are publicly financed.

Research is publicly financed.

Export infrastructure has often received public support.

Insurance markets depend on regulation and sometimes premium support.

Commercial ecosystems routinely sit on top of public goods.

So the question is not whether adaptation can become completely subsidy free.

It is which functions are genuinely commercial, which create wider public benefits, and where temporary support should end.

If farmer education creates benefits for an entire market, perhaps public or philanthropic support is justified during market formation.

If a profitable distributor can fund normal sales activity, subsidy should not replace it.

If a guarantee enables a bank to learn how to lend to a new class of viable MSMEs, it may crowd in finance.

If the guarantee has to remain forever because the underlying borrowers cannot repay, it is performing a different function.

The distinction matters.

38. A useful test: remove the support

For any adaptation model, imagine removing one supportive element.

Remove concessional finance.

Remove the anchor buyer.

Remove the donor-funded field agents.

Remove the subsidy.

Remove the central technical team.

Remove free weather data.

Remove the project vehicle.

What stops?

Whatever stops reveals the function that the market has not yet internalised.

Then ask a second question.

Should it?

Perhaps that function is legitimately public.

Perhaps the buyer should pay.

Perhaps the manufacturer should pay.

Perhaps the farmer should pay.

Perhaps technology can make it cheaper.

Perhaps aggregation can spread the cost.

Perhaps no viable arrangement exists yet.

But now the discussion is precise.

39. A different way of thinking about scaling adaptation

We often imagine scale as replication.

Find a model that worked in Kenya.

Replicate it in Uganda, Ghana and Tanzania.

But adaptation markets are highly contextual.

Water depth changes.

Crop calendars change.

Road density changes.

Farmer liquidity changes.

Mobile-money penetration changes.

Public extension changes.

Buyer structure changes.

Trust relationships change.

The underlying principle may travel.

The operating model may not.

Cold storage may need aggregation everywhere, but who aggregates will differ.

Climate information may need bundling, but the bundle could be credit in one market and input distribution in another.

Irrigation needs maintenance, but the service architecture may be a dealer network in one country and a centrally managed technician system in another.

What scales therefore may be the logic, not the exact business.

40. What would an MSME-centred adaptation strategy look like?

It would start much earlier than financing.

For each adaptation solution and geography, it would first ask what the farmer is actually trying to accomplish and what technology could contribute to that outcome.

Then it would ask what else must be true for that technology to create value.

Which local functions are required?

Who performs them today?

Which functions have no viable owner?

What would the local enterprise have to do?

What does each activity cost?

Which costs occur before revenue?

Which risks are controllable by the MSME?

Which risks are being pushed onto it from somewhere else?

Who benefits from the work?

Who could pay?

How dense must transactions become?

What utilisation is required?

What must the buyer commit?

What happens after installation?

What happens after the project ends?

Only then do we ask:

What capital is missing?

That sequence would produce a very different adaptation-finance portfolio.

41. Maybe the problem is not that MSMEs are weak

It is easy to describe MSMEs as lacking capacity.

Weak management.

Weak financial records.

Weak balance sheets.

Weak technical capability.

Sometimes that diagnosis is correct.

But it can obscure the structure in which they operate.

Imagine asking a five-person enterprise to educate a market, customise an imported technology, finance inventory, extend credit to farmers, install equipment, carry warranty complaints, maintain technicians, aggregate produce, manage seasonal utilisation, wait sixty days for buyer payment and absorb weather-driven volatility.

Then concluding:

“The MSME is not investment ready.”

Perhaps.

But perhaps we have also constructed a role that would strain a much larger company.

The more useful question is:

Which of these responsibilities genuinely belongs inside the MSME, and which should be carried elsewhere?

42. The strongest systems distribute responsibility, not just products

The cases that appear more resilient tend not to eliminate complexity.

They organise it.

Manufacturers remain responsible for warranties.

Local technicians perform repairs.

Buyers communicate standards.

Aggregators coordinate volume.

Financiers handle credit.

Insurers pool defined risks.

Public systems finance some advisory functions.

MSMEs handle the local relationships and operating work they are uniquely positioned to perform.

Farmers pay for value they can reasonably capture.

No actor is asked to carry every uncertainty.

That may be the deeper architecture adaptation markets need.

43. What evolution finally tells us

We started with an apparently simple problem.

Climate risks are increasing.

Solutions exist.

Smallholder adoption remains limited.

MSMEs lack finance.

After following the chain, the story becomes more complicated.

And more useful.

Many adoption problems are not permanent features of smallholder markets.

They are unresolved design questions.

Who aggregates?

Who demonstrates?

Who owns the asset?

Who carries utilisation risk?

Who finances inventory?

Who guarantees repairs?

Who translates information?

Who commits to buy?

Who absorbs weather risk?

Who pays for market building?

Who remains after the pilot?

Different businesses have answered parts of these questions differently.

Cold-storage businesses changed ownership models.

Irrigation companies bundled financing and service.

Equipment providers built distributor repair systems.

Climate-information providers moved toward bundled services.

Input systems invested in demonstration.

Buyer-led chains embedded extension and aggregation.

Insurance transferred particular risks away from individual farmers.

None of these changes made the underlying technology dramatically more “climate adaptive.”

They made the delivery system more complete.

44. The climate adaptation gap may partly be an enterprise-design gap

This is where I currently land.

Africa may not only have an adaptation-finance gap.

It may have an adaptation enterprise-design gap.

We have invested heavily in discovering technologies.

Increasingly, we are investing in financing them.

But between invention and finance sits a less glamorous layer.

The businesses that have to make the technology usable every day.

Their work is fragmented.

Their margins are thin.

Their customers are dispersed.

Their revenues are seasonal.

Their risks are often transferred from larger actors.

And many of the activities required to establish a new market—demonstration, trust, localisation, service readiness and demand aggregation—do not naturally pay for themselves during the early years.

Yet without this layer, the technology remains a pilot.

45. So perhaps we should stop asking only, “How do we get the technology to the farmer?”

A better question may be:

What has to become true for a local enterprise to keep delivering this solution after everyone else leaves?

Does it know which farmers to serve?

Can it find them cheaply enough?

Does the farmer create enough value?

Does payment match the farmer’s cash cycle?

Can demand be aggregated?

Are complementary services present?

Does someone credible buy the resulting output?

Can failures be repaired locally?

Does the MSME control the risks it is expected to carry?

Can it earn from the work it performs?

Does it survive the low season?

Can it replace the asset?

Would it continue if the project disappeared?

If the answer is yes, we may have something more important than a successful climate project.

We may have the beginning of a market.

46. The technology is only the first move

Innovation conversations naturally gravitate towards what is new.

AI.

Sensors.

Solar.

New seed genetics.

Remote monitoring.

Digital platforms.

These matter.

But some of the most consequential adaptation innovation may look far less exciting.

A distributor paid for warranty repairs.

A seasonal repayment schedule.

A buyer agreeing specifications before planting.

A cold room placed beside an existing produce flow.

An agro-dealer able to order small quantities of a new variety.

A weather service bundled with something farmers already purchase.

A field agent serving 300 farmers rather than 30.

A working-capital facility that pays farmers today while the exporter pays in 45 days.

A public programme paying explicitly for market-building activities that no individual business can capture.

Small changes.

But this is how systems evolve.

One friction becomes visible.

Someone changes the arrangement around it.

The next constraint appears.

The model changes again.

Eventually the technology stops feeling like innovation.

It becomes infrastructure.

That, ultimately, may be what successful climate adaptation looks like.

Not thousands of farmers trying a climate solution because a project brought it to them.

But thousands of ordinary local businesses finding it worthwhile to keep the solution available, functional and useful because farmers keep finding it worthwhile to use.

When that happens, the last mile is no longer the last mile.

It is simply the market.