A very important development in financial inclusion is the use of big data and artificial intelligence – specifically machine learning – to make it possible to lend money profitably to otherwise underserved segments of the population. A useful overview of this topic is the report by the Omidyar Network, “Big Data, Small Credit”, which was published in June 2016.
However, the challenge of making unsecured loans on a profitable basis should not be underestimated and, so far, there is little or no information on how the companies pioneering these advanced technologies are performing. The algorithms they are using need to be seen in operation over a reasonably long time period, ideally across economic cycles. There are also other important factors which will determine whether a lending business is profitable such as the cost of customer acquisition, operational efficiency and the management of collections.
Some companies are focused on online solutions and some companies are focused on mobile solutions, for example:
- Online focus – Lenddo, an early pioneer in this space, was founded in 2011 and has raised $14m of venture capital funding as of the end of 2016. The initial focus of Lenddo was to provide loans to the emerging middle class in developing countries. This has evolved into working with financial institutions (and other firms) as partners. Lenddo claims to have patented technology based on 4 years of lending experience that has included the collection, analysis and processing of billions of data points. The company is operating in several countries including the Philippines, Colombia and Mexico. (Lenddo’s scoring and verification factsheets are available on the company’s website).
- Mobile focus – Branch International was founded in 2015 and has raised $9m in venture capital funding as of the end of 2016. The company is focused on mobile lending and, with customer permission, uses a variety of data to assess credit worthiness – handset details, SMS logs, social network data, GPS data, call logs and contact lists. Loan amounts are up to $500, from 2 weeks to 1 year, and have an APR of 15% to 180%. Typically, a customer will be offered a very small amount for a short period at first, and then will be able to borrow more at a later stage if they repay the first loan on time. Based in San Francisco, the company is making loans in Kenya.
Two companies using artificial intelligence to make loans in emerging markets which have reported some performance information are Kreditech and MyBucks.
Kreditech was founded in 2012 and is based in Germany. It operates in 5 core markets: Poland, Spain, Czech Republic, Russia, Mexico and, as of June 2016, had processed almost three million loan applications and made 750,000 loans. Revenues have been growing rapidly, reaching over $40m in 2015, its third full year since launch.Kreditech have raised around US$150m in equity funding from leading private equity investors. Most recently, in March 2016, they received $11m from the IFC (part of the World Bank) because of their work with the underbanked. This successful fund-raising suggests that the business model is working. Over 300 people are now employed by Kreditech, many of them data scientists. Around 20,000 data points are used for each loan application, with decisions being made independently from credit bureau. The company uses a paperless, and fast application process with immediate scoring and pay-out – with scoring taking just 32 seconds.
MyBucks is also deploying artificial intelligence to make loans in emerging markets. This company was founded in 2011 and now operates in 13 countries following the acquisition of Opportunity Bank. In 2016, MyBucks listed on the Frankfurt Stock Exchange with an IPO price of €13.50 per share. As of February 2017, the price had risen to just over €16 per share valuing the company at €180m.The MyBucks annual report for year end June 2016 describes the company’s approach to credit scoring and fraud detection using artificial intelligence. In the report, the company noted that more than 750,000 loans had been made since inception, with a default rate of less than 8%. However, the impairment percentage on the gross loan book of €52m was 21%, illustrating the very high credit risk associated with this type of unsecured consumer lending. We should also point out that the advanced credit scoring was only being used in 6 countries as of June 2016, so it is probably too early to say how the systems are performing.
One issue that these new approaches to credit scoring do not explicitly take into account is that of affordability. In many markets, it is becoming more important to ensure that loans being made to customers are affordable and customers need to be treated fairly. This is a challenge that the companies making loans remotely through online or mobile channels will still have to face at some point.
We expect that over the next year there will be more information emerging on how advanced credit scoring is working, based on actual performance, and based on the ability of some of these early stage businesses to raise additional funding. The amounts that have been invested so far by fintech disrupters are still relatively small and incumbent banks and finance companies will also start adopting some of these technologies – note that Lenddo is already offering its credit scoring as a service to established players who have existing customer bases they can target.
You can find additional strategic advice and insights from Michael Pearson of Clarus Investments via http://www.clarusinvestments.com/?page_id=22