“Determining who’s a good credit risk used to be a challenge”
Nicaragua’s forested mountains are perfect for growing coffee. Some 92 percent of it is grown by smallholders, with families eking out a living on a few hectares. While world coffee prices have slumped over the last decade, specialty beans fetch a premium. However, farmers often need to invest in their farms to grow the high-quality specialty beans. Finding a loan on fair terms to pay for the necessary improvements can be difficult.
Aldea Global is one of the few places smallholders can obtain microcredit at affordable interest rates. The small farmers’ association in the heart of Nicaragua’s coffee country has seen demand soar. “We’ve grown in recent years,” confirms Warren Armstrong, Aldea Global’s General Manager. “Determining who’s a good credit risk was a challenge, as was the credit application process taking too long.”
Traditionally, one of Aldea’s eighty loan officers would visit farmers whenever they requested a loan. Loans are usually repayable within just six months, but if the farmer put in a new application, this entailed another visit. Coffee farms are often located up remote mountainsides, only reachable by Jeep or motorbike. To be able to meet the growing demand, Aldea Global asked Rabo Foundation to help it find a more efficient way to analyze their portfolio.
The Foundation has worked with Aldea Global for many years. It provides financing and supports the association’s training and trading activities that help farmers boost incomes and embrace sustainability. “We‘ve been able to add value this time because we have in-house data expertise available,” says Albert Boogaard, Head of Innovation at Rabo Foundation.
The Foundation brought in data, analytics and AI specialists from Rabobank’s Digital Transformation Office (DTO) and DLL, a company owned by the bank specializing in global vendor finance. They are now running a pilot at Aldea, using predictive credit risk modeling to speed up the lending process. Aldea proved an ideal candidate for the pilot as it had already digitized large amounts of customer information. Frank van den Eerenbeemt of Rabobank’s DTO: “Aldea’s data on things like individual farms’ irrigation methods and yields were very valuable. We’ve shown them how to optimize these using AI.”
Filtering out inefficiency
Machine learning was new to Aldea, but the new system still builds on the organization’s strengths. Aldea’s decades of experience are now formalized as business rules. The models help loan officers make fast, informed decisions about approving or declining a loan application, or determining if a farm visit is needed.
Farm visits will continue: part of the local culture, they help ensure default rates stay low. Crucially, they also allow Aldea to collect the new data needed to update the system. Rather than hurrying to visit when a farmer applies for credit, loan officers can now plan to visit several farmers who all live in a particular part of their catchment area, saving time and travel expenses.
Diverse data streams
The initial goal is for loan officers to look after 350 customers each, up from 250. Processing applications should speed up too, from five days to four for new customers, and even faster for existing ones. While these goals may sound modest, they can be made more ambitious in the future. Scalability has been built into the project. “Aldea can also improve the machine learning model by adding more data sources,” says Tanya Lagoda, Credit Scoring Modeler and Analytics Consultant at DLL. “Think satellite data, weather information and soil composition details. These will make it even more effective.”
“That way we’ll be able to help even more farmers move forward,” says Aldea’s Armstrong, clearly pleased at the prospect.