Imagine you are riding on the underground and you receive a text message: “Warning! Be careful! You are sitting next to an untrustworthy person.” Better yet – imagine the person sitting next to you receives such a message with your photo included.
Chills down your spine? For sure.
This is already happening in China.
Technology has given the human race unprecedented powers to predict things. And just like “The Force” from Star Wars, this phenomenon has its light and dark sides. Over 11 million Chinese citizens have already been scored and blacklisted. They’ve been prohibited from traveling, purchasing property, and stripped of other basic rights valued by their Western counterparts. On a more positive note, Chinese dating apps are now boosting the profiles of “trustworthy” individuals, utility companies are starting to offer them discounts on energy bills, and, of course, banks are offering lower interest rates and faster approvals. It is believed that the China social score will ultimately take into account even such issues as jaywalking and smoking in non-smoking sections, powered by AI-processed input from the 500 million CCTV cameras in China.
The underlying trends powering this phenomenon are Big Data and Artificial Intelligence. More data has been created in the last two years than in the entire history of the human race prior to that. It has been estimated that we are creating over 1 megabyte of new information every second for every human being on the planet. And we are rapidly developing new tools to make sense of all this data and find trends and correlations. Machines are way better than humans at crunching huge volumes of data, and most observers agree that we are going to achieve HLAI (Human Level Artificial Intelligence) within 10-20 years. To understand how this will influence consumer credit we need to understand what data can be used and how it can be made predictive.
Historically, lenders used data from a client’s application and credit scores to assess potential credit quality. However, there are serious problems with these data sources. Application data can be easily “gamed” by the applicant to improve his chances of getting a loan. Furthermore, over 70% of the world’s population doesn’t actually have a credit score. A number of companies have popped up globally over the last few years trying to use Big Data and AI to solve this Holy Grail challenge.
In their experimentation, contenders for global Alternative Scoring market leadership have tested out the following data sources:
- Device Data: The increasing ubiquity of smartphones with over 50% penetration in many Emerging Markets makes it a readily available data source. The smartphone device operating system contains over 50,000 data points, many of which are highly predictive of credit behaviour. For example, our portfolio company CredoLab, one of Southeast Asia’s foremost alternative credit scoring providers today, has been able to consistently achieve world-class Gini coefficients in excess of 0.4, and onboarded over 40 lenders onto their platform already. Another portfolio company of ours, AsiaKredit, one of the top digital lenders in the Philippines, also actively leverages device data in its credit process. Some other notable international players in this space include Tala, a Paypal-backed Emerging Markets digital lender, as well as Africa-focused companies Cignifi and FirstAccess.
- Financial Transaction Data: This includes bank account, credit card, as well as e-wallet ledgers. This data source seems to offer good predictiveness, but its availability is largely limited to the Developed Markets, as in the Emerging Markets bank account penetration remains relatively low. A further issue is the complicated client consent and bank-by-bank API integration that’s required in order to access this information. This has led to Alternative Scoring players in the Emerging Markets to deprioritize this data source.
- Social Media: There was a lot of excitement around this data source a few years back, but repeated trials found that this is not a good source for predicting credit behaviour. The consensus is that social media data is useful for KYC, but is low-Gini and very unstructured, making it hard to work with. Lenddo is among the leading players in this segment globally.
- Psychometric: A few companies, led by EFL (now merged with Lenddo), have implemented various tools that seek to access psychological and cognitive factors of the potential borrower, basically asking the borrower to fill out a psychometric test. While the data has been found predictive, it is also quite invasive and requires a lot of customer effort to collect, thus reducing the so-called “hit rate” and “conversion rate”.
Getting the so-called Predictor data-set is only part of the problem. In order to create a viable prediction, one needs to run a statistical regression analysis between two data sets: the Predictor set and the Credit Outcomes set. This is where most of the companies in the space had a lot of problems in the past, sourcing matching Credit Outcomes set.
A few players tried to solve this by becoming lenders themselves, to “train” their scorecard system. However, what all of them have invariably discovered is that a “generic” scorecard that they can create on the basis of their own experience doesn’t translate well to other lenders, sometimes even resulting in a negative Gini coefficient.
For example, CredoLab solves this challenge by developing custom individual Digital Footprint scorecards for each lender, and even each product/channel within each lender. A similar approach has been adopted by WeCash, the Chinese market leader in alternative scoring.
Privacy and data security are further issues that the Alternative Scoring industry faces. Most customers outside of China like to protect their privacy, and relevant legislation is in place to ensure consumer rights in this area. Anyone operating in this space outside of China has to be very conscious of this legislation, operating on an anonymized data basis. Over the coming years, we can expect some further regulatory tightening around this area globally.
Alternative Scoring has the potential to completely revolutionize global consumer credit, enabling lenders to lend to customers never accessed by credit before. Development finance institutions such as the World Bank and Asian Development Bank have actively called on lenders to drive financial inclusion and enhance credit access through using alternative data sources, including mobile device data. This financial enablement will empower billions of people in the decades to come.
I believe this is one of the most exciting things in retail finance that our generation is going to experience and encourage you to follow this space very closely.
Greg will be the trainer and panelist for the AngelCentral Deep Dive Series Webinar – Early Stage FinTech Investing in the Era of COVID-19: What’s Hot and What’s Not! Join us for the session here