Financial services for the poor in a data-rich world
When I was a Peace Corps volunteer, I lived in a village of subsistence farmers in Central America. Many of the women wove intricate baskets which they sold when they could. After selling a basket that was an entire year in the making, my friend Amelia came to me with an enormous handful of cash and asked me to hold on to it for her. Instead, I offered to take her to open a bank account. The next morning, we walked almost 10 kilometers through the mountains, followed by a two-hour bus ride down a bumpy dirt road to the nearest bank branch.
I was nervous. Standing there next to this brave barefoot woman, I realised then that banks are not designed for people like Amelia. What’s more, banks don’t seem particularly interested in serving customers like Amelia, despite the fact that she had $1,000 in cash in her hand.
What is Financial Inclusion?
Financial services are important for everybody. They enable us to invest in things like education, housing, and our businesses. They also help us to weather unexpected shocks, such as illness or the loss of a job or property. For those living on the economic edge, financial services can play a crucial role in the fight to escape poverty. However, according to the World Bank, 37% of the population in developing economies still do not have a formal account with a bank, mobile money provider, or any other regulated financial institution.
Financial inclusion is simply about ensuring that all people have access to a range of financial services (i.e. savings, credit, insurance, and money transfers) that are accessible (i.e. physically, in price, in administrative requirements, and in ease of use), safe, and meet their needs.
This sounds great, but for this to be possible, financial service providers (FSPs) need to be able to provide these services sustainably. Quite frankly, FSPs need to see that profit can be made in these new markets. Serving low-access, low-income clients can be expensive and risky, because they offer low margins to FSPs that have built their businesses on the urban rich. The key, as demonstrated by the success of products like M-PESA in Kenya, is to achieve scale, which technology and data are uniquely positioned to do.
The role of machine learning and alternative data sources
The financial services sector is fundamentally driven by data in ways other industries are not. Banks use statistical methods, that are rightfully types of machine learning, to run their businesses every day. However, to become more inclusive to customers that do not have a long history in the formal financial system, the industry has to adapt its practices to make the same decisions they have always made, but based on alternative data sources, i.e. data that are not from traditional sources such as a credit bureau or formal customer history.
Alternative data sources are richer, more diverse, and more complex than traditional data. Advances in machine learning techniques can help FSPs make the most of these new sources of data that are becoming ubiquitous due to the use of mobile phones, social media, drones, remote sensors, cameras, and satellites.
Some examples of machine learning applications that are evolving quickly with new sources of data are:
Credit risk assessment: Credit risk modelling on new data give FSPs the confidence to lend to "thin file" customers. These include GIS-mapped sensor data for credit scoring in agriculture (See Harvesting) or phone usage data to predict credit repayment behavior (see Branch and Jumo).
Insurance risk assessments: Modeling actual behaviour data to predict, and even lower, risks, which enables appropriate pricing of insurance products. This includes internet of things (IoT) and sensor data for car insurance (See MTN in South Africa) or “wearables” data for health insurance.
Fraud detection: Machine learning techniques are widely used to efficiently and accurately detect anomalies in client behaviour and signs of fraud. The models continue to improve as new and diverse sources of data are introduced. For insurance companies, this also translates into lowering the time and cost needed to settle insurance claims.
Identity verification: Advances in facial recognition and machine learning using social media data, help FSPs verify identities and other customer data, giving FSPs the confidence to work with new customers who have previously been blocked because they are unable to respond to traditionally onerous Know Your Customer (KYC) requirements. (See Smile ID and Lenddo)
In addition to these examples, there are also chatbots, imagine recognition, and real language processing that have different implications for further improving customer experience and improving efficiencies for FSPs. For further discussion on these topics, see a report by FIBR and video series from insight2impact.
As the world becomes increasingly data rich, machine learning will increasingly turn customers like Amelia and the millions of people like her around the world into a new and viable customer base. In developing economies throughout Africa, Latin America, and Asia, what we may term “alternative” data today is simply the data of tomorrow.
At Ixio, we work hand-in-hand with businesses and development partners across Africa to implement practical machine learning techniques. We see endless possibilities to create win-win solutions for both FSPs and their customers.
Please get in touch with us if you are interested in exploring any of these Financial inclusion applications.