The Missing Score: Why the World's Most Productive Farmers Are Still Invisible to Finance

The $200 billion agricultural credit gap will not close with more lending programmes. It will close when someone builds the FICO score for farming.

In 1956 William Fair and Earl Isaac had an idea that would reshape the global economy: use data to predict whether a borrower would repay. Their product was a literal scorecard, made of cardboard, filled in by loan officers and totted up by hand. Within three decades their company had partnered with Equifax, Experian, and TransUnion to produce the FICO score — a three-digit number, between 300 and 850, now used in 90% of consumer-lending decisions in America.

The effects were transformative and measurable. In Georgia, where your correspondent spent time reporting on the country's credit revolution in 2019, the introduction of a modern credit bureau had increased the value of loans from under 10% of GDP to 56% within twelve years. Average interest rates fell from 20.2% to 12.6%. Loan officers who once made ten decisions a day — visiting flats, verifying identities, checking purposes — found that an algorithm could make ten decisions in under a second.

The lesson from Georgia, and from every market where credit scoring has taken hold, is unambiguous: when lenders can systematically assess risk, capital flows. When they cannot, it does not. The corollary is equally clear: the biggest credit gaps in the world exist not where borrowers are riskiest, but where risk is hardest to measure.

Which brings us to the 285 million smallholder farming households across Asia, Africa, and Latin America who collectively produce nearly a third of the world's food — and who remain, in 2026, almost entirely invisible to the formal financial system.

The $200 billion question

The numbers are by now well documented, if stubbornly resistant to resolution. Smallholder farmers require approximately $323 billion in annual credit. They receive roughly $95 billion. The gap — over $200 billion — has persisted for years despite billions in development spending, hundreds of microfinance institutions, and a generation of fintech startups claiming to have cracked the code of alternative credit scoring.

The reason for the persistence is structural. The FICO model works because it draws on decades of standardised financial records: payment histories, outstanding debts, credit mix, account longevity. These records exist because consumers in developed markets interact with formal financial institutions — banks, credit-card issuers, mortgage lenders — that report data to centralised bureaus. The entire system is self-reinforcing: you borrow, you repay, you build a record, and the record makes you eligible to borrow more, at better rates.

For smallholder farmers in Bangladesh, Laos, Cambodia, or the Philippines, none of these preconditions obtain. Seventy percent of Bangladeshi poultry farmers have never held a bank account. Fewer than 30% of Lao farmers access formal credit. Cambodian farming households number nearly two million, yet the country has no agricultural credit bureau. Philippine cooperatives serve over 12 million members, but cooperative membership generates no standardised data that a lender can use.

The startups that have attempted to fill this void have, for the most part, reached for what was available rather than what was predictive. Tala, a California-based lender, claimed to use over 10,000 data points from a borrower's smartphone — contact naming conventions, travel patterns, app usage — to assess creditworthiness. Others have mined social-media activity, psychometric quizzes, or mobile-money transaction histories. As your correspondent noted in a previous survey of these efforts, the approaches resemble nothing so much as the pre-scientific era of American credit bureaus, when marital troubles and newspaper mentions passed for risk assessment.

The fundamental problem with these proxies is that they measure digital behaviour, not economic behaviour. Whether a farmer in Negros Occidental names her contacts with first and last names tells a lender nothing about whether she manages her crops well, responds to seasonal variation, or honours her obligations to a cooperative. The correlations are passable — exotic, even — but they are no match for the predictive power of actual performance data.

Enter the production cycle

What has changed since 2019, when this newspaper last surveyed the credit-scoring landscape, is the recognition that for hundreds of millions of agricultural borrowers, the relevant data are not financial at all. They are operational.

Consider a contract broiler farmer in Bangladesh. Every 42 days he completes a production cycle: receiving day-old chicks, managing feed, monitoring mortality, maintaining biosecurity, and settling accounts with an integrator. Within that cycle he generates a remarkably dense dataset. Feed conversion ratio — the efficiency with which feed becomes live weight — is recorded daily. Mortality is counted each morning. Vaccination schedules are tracked. Settlement payments are documented.

A farmer who consistently achieves a feed conversion ratio below 1.50 across multiple cycles is demonstrating the kind of disciplined operational management that, in a consumer-credit context, would be equivalent to a long record of on-time payments. A farmer whose mortality rates remain below 5% cycle after cycle is showing risk management competence. A farmer who logs activities daily and settles accounts promptly is exhibiting precisely the behavioural consistency that FICO's five variables attempt to capture through financial proxies.

The same logic applies, with different signals, across crop farming. A coffee grower on the Bolaven Plateau in Laos who plants within optimal seasonal windows, participates actively in cooperative quality programmes, and delivers consistent yields year after year is generating a credit profile — it simply has never been captured, standardised, or translated into a language that financial institutions understand.

This is the insight that underpins a new generation of agricultural credit intelligence platforms — companies that score farmers not on their digital footprint but on their actual farming behaviour. Among the most advanced is AGXL, a UK-registered company operating across Southeast Asia, which has built what it calls the OrganicCreditScore: a 0-to-10,000-point rating generated from production data, engagement patterns, and operational consistency across both poultry and crop systems.

Timing and catastrophe

The timing could hardly be more consequential. As this newspaper went to press in April 2026, the Strait of Hormuz remained effectively closed, trapping nearly two million tonnes of fertiliser behind a blockade that has sent urea prices surging over 30%. The blockaded region accounts for roughly 30% of globally traded fertiliser and 20% of liquefied natural gas — a critical feedstock in nitrogen fertiliser production. The FAO has warned that the window to prevent a full-blown food crisis is measured in weeks. The World Food Programme estimates that if the strait does not reopen by mid-year, an additional 45 million people will join the 300 million already struggling to feed themselves.

Compounding the supply shock is a strengthening El Niño. The last major event, in 2023-24, brought the worst drought in a century to southern Africa, killing thousands of livestock and pushing over 30 million people into food assistance. This year's event will layer on top of accelerating baseline warming, intensifying extremes in precisely the regions — South and Southeast Asia, sub-Saharan Africa — where smallholder agriculture is most concentrated and most undercapitalised.

In this context, the absence of agricultural credit intelligence is not merely a market failure. It is a crisis-management failure. Governments attempting to distribute scarce fertiliser subsidies have no systematic way to identify which farmers will generate the highest yield per kilogram of subsidised input. Microfinance institutions managing agricultural portfolios cannot distinguish between borrowers who will weather a price shock and those who will default. Development agencies deploying hundreds of millions in climate-adaptation funding lack the farmer-level data to target interventions or measure outcomes.

From cardboard to cloud

The parallel with Fair and Isaac's cardboard scorecards is instructive. In 1956 the idea that data could predict borrower behaviour was novel. The data available were crude — telephone ownership, job tenure, residential stability. The models were simple. But the principle was sound, and it transformed consumer finance within a generation.

Agricultural credit intelligence is at an analogous inflection point. The data are different — feed conversion ratios rather than payment histories, cooperative participation rather than credit-card balances, biosecurity compliance rather than debt-to-income ratios — but the principle is identical: observed behaviour, measured over time, predicts future reliability.

AGXL's approach is notable for what it does not do. It does not mine smartphone metadata. It does not scrape social-media profiles. It does not administer psychometric tests. It scores farmers on the thing that matters most to a lender: whether they farm well, consistently, and reliably. The platform operates across poultry systems in Bangladesh and crop systems in Laos and the Philippines, generating institutional-grade credit intelligence sold to the entities that actually make lending decisions — banks, microfinance institutions, cooperatives, integrators, and development finance institutions.

The company is not alone in this space — CGAP and Mercy Corps AgriFin have convened working groups exploring alternative credit models for smallholders, and the farm analytics market is projected to grow from $1.4 billion in 2023 to $2.5 billion by 2028. Asia-Pacific agricultural credit is expected to grow at 8.1% annually through 2033. But the distance between working groups and working infrastructure remains considerable. What the sector needs is not more research on whether behavioural data can predict agricultural creditworthiness — the evidence is clear that it can — but deployed systems generating scores that lenders actually use.

The Georgian lesson, revisited

Your correspondent, revisiting the Georgia story seven years on, is struck by how rapidly credit infrastructure transformed an economy once the basic architecture was in place. Georgia went from no credit bureau to near-universal coverage in barely a decade. Lending quintupled as a share of GDP. Interest rates fell by a third. The mechanism was not complicated: reliable data reduced the cost of assessing risk, which reduced the cost of lending, which increased the volume of lending, which generated more data.

The same virtuous cycle is available to agricultural finance in the developing world. The smallholder farmers of Bangladesh, Laos, Cambodia, and the Philippines are not unbankable. They are unscored. The distinction matters enormously. The first implies a deficiency in the borrower. The second implies a deficiency in the system.

Every 42 days, a broiler farmer in Gazipur completes a production cycle that generates more verifiable performance data than most small-business owners in the developed world produce in a year. Every harvest season, a coffee grower in Champasak demonstrates, through measurable behaviour, whether she is a reliable counterparty. The data exist. The predictive logic is sound. What has been missing is the institution that captures it, standardises it, and translates it into a score that a bank in Dhaka or an MFI in Vientiane can act on.

That institution is now being built. Whether it scales fast enough to matter in the current crisis — and the next one, and the one after that — is a question that 285 million farming families are waiting to have answered.

AGXL GROUP LTD is an AI-powered infrastructure technology company providing institutional-grade credit intelligence for smallholder farmers in emerging markets. AGXL is live in Laos, and the Philippines. Learn more at agxl.ai.