5 Types of Credit Scores, Powered by Alternative Data – Pros and Cons

The expression “Data is the new Oil” has become a form of a cliché in recent years, as more and more companies incorporate data analytics, data science, Machine Learning and Artificial Intelligence into both internal business processes and as core value proposition of their products and services. This is true across all verticals, industries and company sizes, and for digital businesses, access to well structured, high quality data and proprietary know-how in extracting business value from it is no longer considered an option, but a core requirement for success.

This is however not necessarily true across all countries and regions in the world. The US, Europe (particularly – the Eastern European countries) and China are leading the way when it comes to top-level data scientists – both in terms of top-tier talent attraction and know-how incubation. This leaves the rest of the world at a slightly more disadvantaged position, as the lack of access to a significant in size, well-educated pool of skilled data professionals, makes the in-house development of cutting-edge technologies a constant challenge.

One of the trends we observed during FinScore’s 5-year journey in the financial technology industry in Southeast Asia (and the Philippines in particular) has in turn been the prolonged education period and effort required, when it comes to unleashing the power of alternative data. There are several factors contributing to it – the low bank account penetration, the lack of an established nationwide government credit register, as well as the relatively low coverage of the traditional credit bureau players on the Philippines market practically leaves more than 80% of the population of the country with no access to the traditional measures of creditworthiness. This in turn makes it impossible for banks and financial institutions to extend even basic services, such as bank account, credit card or minimal cash loans to a vast majority of the market,  a phenomenon also  known as the “chicken or an egg problem” of the financial industry in the Philippines.  

But what options do financial institutions have in the market, where no traditional reliable credit information is available? 

Here’s our list of the most commonly used alternative data sources, proven to have the highest predictive power for credit risk assessment worldwide, alongside with their upsides and downsides:

• Telecom / Telco / Mobile Network Operator Data – notably the alternative data that is considered to have the most predictive power for measuring creditworthiness, in the absence of past credit history. Mobile network data such as average bill size, good payer (for postpaid), time with operator or top-up frequency & amount (for prepaid) is invaluable as a proxy for credit reliability and best of all – telco information is not self-reported and cannot be amended. Borrower’s consent to use their telco data for credit scoring purposes is obtained during the loan application process, which ensures data privacy laws and regulations are being observed. While Telco credit scoring is relatively popular across the globe, FinScore is so far the only leading Telco credit scoring provider in the Philippines, with over 8 years of experience and more than 1,000,000 scores delivered to financial institutions in the country to date.

• Device Data – mobile device data is another popular method of gauging clients credit reliability in the absence of traditional credit bureau information. The data is normally accessed through an app that the Android or iOS user needs to install on their smartphone. While most people have smartphone devices nowadays, unfortunately this method doesn’t work particularly well for feature phone users. Another downside is related to data privacy regulations. Since the data, obtained through mobile devices, contains a lot of sensitive information (such as geo-location, contacts and, in some cases – photos), international regulations of major App Stores, such as Apple Store or Google Play have with time tightened the rules, related to the type of information that can be extracted from user mobile devices. While rich in quantity (device data often can gather up to 50k+ individual data points), the challenge for analytics companies and scoring providers is to extract the valuable insights from such a large pool of potential variables – a task that FinScore’ in-house analytics team’s experience and proprietary know-how has been utilized particularly well in various consulting assignments in Europe, Africa, and Asia.

• Social Media Data – in the past, some credit scoring providers have tried to obtain valuable information, related to creditworthiness from Social Media, such as Facebook and Instagram. Recent PR scandals have raised concerns with the level of detail, made publicly available by the companies, which in turn significantly decreased the amount of data being made available through public API’s. Another issue is the reliability of information and number of fake accounts on the web – since all the information is self-reported, it is relatively easy to game the system. Nevertheless, there are still factors and ways to use Facebook as a proxy for identity or basic fraud verification, which are worth considering, in the absence of traditional tools, such as a centralized identity register or national / government ID.

• Psychometric Survey Data – while the term sounds kind of modern and potentially a bit scary,  this type of data analytics has been around in Research for decades. Through a simple survey, usually containing no more than 10 questions, a “personality score” can be developed, which in turn can be used as proxy for creditworthiness for the borrower. The advances of the digital age have made it possible to combine the answers with parameters such as “speed of response” or heatmaps, in order to maximize the results. However, similar to Social Media data, responses to psychometric surveys are still self-reported and as we all know – individuals are not always necessarily upfront with their answers, even with the best intentions in mind. Additionally, filling out a psychometric survey as part of the loan origination process often leads to an increased churn rate, especially if the questions haven’t been introduced in the right format.

• Web Behavior Data – it is no secret that almost all websites we visit nowadays gather various information about us, in the form of “cookies”. Some of these are indicative of our behavior on the website, while others may include information about other websites we visited, searches we performed online or devices we are using. Some companies claim that Integrating this information in the loan decision process can lead to lowering the default rate of the borrowers, but in reality, compared to some of the better known alternative data types, such as Telco, there has been very few research and experiments that confirm this theory. GDPR in Europe, as well as general privacy concerns have also prompted businesses to be reluctant with the type of information gathered from their website visitors.