Financial institutions in Africa have historically relied on credit bureau scores, or the financial footprints of customers, to make lending decisions. In the absence of these, social demographics—such as gender, employment status, income level, etc—are considered before granting loans. 

A major shortcoming of these conventional lending methods is that it puts millions of people who are unbanked or informally employed at a disadvantage, which is why Africans have some of the lowest levels of access to credit in the world, especially those living in remote areas.

Over the past decade, new models based on artificial intelligence and machine learning tools have emerged as an alternative way to assess credit risk.

“Assessing creditworthiness without a formal history is a major problem in financial services in Africa while social demographic data isn’t sufficient to make good credit decisions,” notes Eunice Gatama, Director for Africa Business at Yabx, one of the fintechs driving alternative credit scoring trends in emerging markets.

Get the best African tech newsletters in your inbox

The startup, which is incubated by Comviva and part of India’s Mahindra Group, uses machine learning to analyse mobile money wallet records and combine these with other sources such as credit bureaus and utility bills. 

That data is used to construct risk profiles for potential borrowers without credit histories and detects borrower activity such as how much money is invested in a small business or used to meet personal needs. This solution is available for banks and microfinance institutions in markets where credit bureau coverage may be limited. 

“We enable financial service providers to create profitable unsecured portfolios that are accessible through easy loan application on mobile devices,” Gatama tells TechCabal in an interview. “A lot of elements can be deciphered from data that is collected.”

As well as micro and small consumer loans, Yabx provides small business loans, unsecured working capital loans for mobile money agents, smartphone purchase financing, and a savings product. Its solution enables loan decisions to be taken instantly.

Yabx claims to have executed the credit score of over 100 million borrowers—50% of them Africans—across 15 emerging markets in Africa, Asia, and Latin America. According to the company, it processes over 100 billion data records in a month across partner networks.

In Africa, it has operations in Tanzania, Uganda, Malawi, Somalia, Mauritania, and Côte d’Ivoire, where it’s partnered with leading telcos, e-commerce and payment service players, banks, and other financial institutions to deploy several digital lending products. It makes money based on the performance of loans it sources for banks, through revenue-sharing agreements with other lending partners and white-label services targeted at banks that wish to launch and market their own digital lending products.

Given the penetration and increasing use of mobile money, Gatama argues that in Africa, mobile wallets are currently the best source of alternative credit scoring. In its latest annual State of the Industry report, GSMA reveals that the value of mobile money transactions passed the $1 trillion mark in 2021 with Africa accounting for nearly 70% of the total amount of transactions recorded.

The ease of access to a mobile phone in Africa enables many people to access financial services that they would have otherwise not been able to access. 

“More than half of Africans remain unbanked but most own mobile phones that are also used to access financial services,” Gatama says. “This creates an opportunity to offer credit products to the mass market, especially people banks currently can’t onboard.”

An expansion into 11 more African countries is currently in the works, with priority given to markets like Kenya, Togo, Benin Republic, and Zambia, where mobile money is widely used. “We find countries where mobile money is established easier to set up,” Gatama says.

She doesn’t consider Yabx’s machine learning model to be infallible, despite its effectiveness so far in keeping non-performing loan (NPL) rates of lending partners at single digits but expects further improvements going forward.

“The way machine learning works is the more you train your model, the better it becomes at prediction,” Gatama explains. “Over time, our models will get stronger as more information becomes available.”’

Outside that, regulation, data privacy concerns, and reliability of aggregated data are some key concerns for Yabx as it looks to achieve its big vision: simplifying access to finance for the more than 2 billion unbanked population globally using the digital footprints through mobile devices.

If you enjoyed reading this article, please share it in your WhatsApp groups and Telegram channels.

Michael Ajifowoke West Africa Reporter, TechCabal

Get the best African tech newsletters in your inbox