Data Gold

Using Daily Lives to Unlock Financial Access

by Shivani Siroya

THE MIDDLE CLASS is growing faster in Africa than anywhere else worldwide. By 2060, it’s projected to reach more than 1 billion, putting Africa’s poor in the minority; in Kenya, which our team just visited, the middle class already makes up nearly 45 percent of the total population. This is all largely good news, except when you consider that most of the people who fall into this emerging class — defined by the African Development Bank as those who live on $2 to $20 USD a day — still don’t have access to the financial mainstream. Most have mobile phones and some discretionary income but few have banks, let alone credit cards, that can help them take risks and grow their businesses and better their lives.

This is because most, having just entered the middle class, have little to no formal financial history. Absent any of the traditional indicators, banks will rarely take a risk on them. So InVenture decided to build credit scores for them, examining the vast amount of daily life data on their smartphones to prove their creditworthiness. Our Android app, Mkopo Rahisi — operating in Kenya since March 2014 — approves customers for a line of credit and disburses loans in less than a minute.

InVenture’s Android app, Mkopo Rahisi, builds a living credit score for customers using the rich data available on their smartphones. Approved customers receive loans through the app nearly instantly. (Photo by Ami Gosalia, InVenture)

In emerging markets, mobile credit scoring has some key advantages over traditional scoring methodologies. For one, it’s often more convenient for borrowers who otherwise have limited options for proving their financial potential. After downloading Mkopo Rahisi and allowing the app to access their device, they receive a decision on credit nearly instantaneously, based solely on data from their daily life and habits — no paperwork or trips to the bank. Perhaps most importantly, analyzing the data from a potential borrower’s smartphone allows us to see a richer, more complex picture of his or her life than we might get from a paper trail or even an in-person interview. In many cases, this helps us approve customers for credit who might otherwise be rejected. As we structure data from a user’s social networks, mobile money ledgers, and GSM data, we can build relationships among seemingly disconnected bits of information that can tell us — with astounding accuracy — whether he or she is likely to pay back our loan.

We refer to these bits of structured information as “data nuggets” because they’re like gold to us, helping us open financial access to deserving people who might otherwise be overlooked. To understand both how important these nuggets are and what they really mean for a person’s life, consider the story of Jenipher, who we met last week on a visit to Nairobi.

Analyzing the data from a potential borrower’s smartphone allows us to see a richer, more complex picture of his or her life than we might get from a paper trail...this helps us approve customers for loans who might otherwise be rejected.

At 65, Jenipher has an enviable amount of energy and ambition. She’s been running a small restaurant in Nairobi’s central business district for more than a decade, waking up at four o’clock every morning to take the bus from her home in another part of the city. She had always dreamed of growing her business — moving to a larger location or at least being able to serve more customers per day — but she was never able to qualify for the bank loan that could help her do it. She didn’t have paperwork to demonstrate her own or her business’s financial health.

The story might end there if not for Mkopo Rahisi — and if not for Jenipher’s son. One afternoon, while explaining to him that she wanted capital to buy vegetables in bulk, her son took her phone from her and downloaded Mkopo Rahisi. Before she could protest, the app had approved her for credit and disbursed the funds to her mobile money account. She used the credit to buy vegetables, and a few days later, she began paying back her first loan.

So how did InVenture know that Jenipher would be a good borrower, and prove what the bank could not? For starters, our data shows us that good borrowers tend to have strong relationships with a wide network of people. We can begin to assess this trait by examining a customer’s communication habits: borrowers with high potential like Jenipher will have phone conversations with a greater than average number of unique contacts, and will reach out to their network more often than their network reaches out to them.

Jenipher, pictured at right, attended an Mkopo Rahisi borrower event in Nairobi. She runs a food stall in Nairobi’s central business district and uses her Mkopo Rahisi credit to buy ingredients in bulk. (Photo by Ami Gosalia, InVenture)

We then contextualize these nuggets to determine what they’re really telling us about a borrower’s life. A borrower’s phone use, for instance, can also give us an idea of his or her wealth, since in most emerging markets, airtime isn’t cheap. This is important for the obvious reason: a person who has the financial resources to spend a lot of time on the phone probably has the financial resources to pay back a loan. Not surprisingly, we’ve found that good borrowers tend to have purchased a higher than average amount of airtime than bad borrowers. Their calls also tend to be longer, and a lower percentage are made in off-peak hours, when airtime is cheaper.

To balance out our communication nuggets, we also cross-reference calling habits with SMS habits, since texting is more affordable and used more widely than calling in emerging markets. We’ve learned that, similar to calling, good borrowers tend to send more texts than they receive. They also send more texts overall than bad borrowers on average. Jenipher’s texting habits, like her calling habits, help us measure the strength and variety of her network, and, in turn, her likelihood of repayment.

It might seem that we know a good bit about Jenipher, at this point — maybe even enough to know that she’ll be a good borrower — but in fact, this single set of indicators is only a small slice of the data that Jenipher provides us and a small fraction of what goes into our credit scoring algorithm. By balancing and contextualizing a variety of data, we ensure that no single indicator defines Jenipher’s creditworthiness and avoid catch-all lending practices that might deny her a loan. In future posts, we’ll explain some of the additional signals that factor into our score. But first, we’ll tell you how Jenipher’s story ends.

Jenipher has taken out 21 loans and counting since she began working with us last September. She requests — and receives — her credit on her morning bus commute, before dawn, when bank branches aren’t even open. She buys bulk vegetables and other ingredients, collects her revenue, and makes her first installment payment just a couple days later. And because of the money she’s saved, she’s now, at 65, moved her business into a larger location in Nairobi.

Like all good stories, her end is another beginning.

Want to help change financial services for 2.5 billion people?
InVenture is hiring!

InVenture is providing modern credit for a mobile world.
Follow us on Twitter