Providing access to credit to individuals and businesses sparks inclusivity and drives growth for the society and economy.
The existing brick wall that we are facing is that despite the strong loan appetite banks and financial institutions to the lower-class, they cannot proceed in fulfilling the loan because of lack of financial information.
Their options are constricted, resulting to their dependency informal sources of credit. In reality, these types of lenders, are unregulated by the government and can provide unjust terms and interest rates.
In the “Bridging the Financial Gap in the Philippines”, a talk delivered by FinScore’s Country Manager and Chief Strategy Officer Christo Georgiev at the Ultimate Fintech Forum 2021, it was mentioned that there is a gray area or a “gap” between the credit-invisible segments and the credit providers. The said segments are considered to be of high-risk and many loans to be unpaid. On the other hand, people that fall under this segment might be unbanked because they may have been utilizing their earnings in faithfully paying for their rent, water bills, electricity bills, school fees, etc. These bills payments can be considered as another category of alternative data —enterprise data.
He tackles that Alternative data has the potential to break down the brick wall and even build a bridge with it. Even though that the downside of Alternative data is it is massive and unstructured, data analytics companies such as FinScore, apply their data science capabilities can unlock the power of this data and transform it to meaningful business insights using cutting-edge technologies such as artificial intelligence and machine learning.
What is alternative credit scoring?
Alternative data credit scoring or alternative credit scoring is the umbrella term for credit scoring using non-traditional data sources and methods.
In traditional credit scoring, lenders manage their risk and reduction of bad loans by relying on credit bureaus that provide credit information of the consumers. They can learn if the loan applicant has been paying off loans on time, has ongoing loans, or has been blacklisted for not paying at all. But this only works for the ones who had a loan history. Besides this, they assess the socio-demographic profile of the borrower which can be found on the loan application form either in pen-and-paper or digital form. The lender may also use a credit scorecard, a type of credit model, that outputs a score telling how probable the borrower can repay the loan on time. With Alternative Credit Scoring, alternative data sources such as mobile usage, digital wallet usage, bill payment history, social media behavior, geolocation, and more can be aggregated to produce a potent credit score.
Machine Learning is a way to utilize modern technologies and statistical knowledge to let powerful computers find out the best mathematical model which produces useful insights based given data. It is a type of artificial intelligence wherein a computer system, with minimal human intervention, learns from data, recognizes patterns, and make decisions or predictions.
When Machine Learning is used for the purpose of alternative credit scoring, data scientists develop a predictive model, based on hundreds of alternative data points or variables. Compared to the static data approach of traditional financial and socio-demographic data from a lender, credit scoring powered by Machine Learning could allow banks and other lenders to increase revenue by approving more credit invisible applicants and more applicants whose traditional credit scores paint an incomplete picture of their creditworthiness. This can potentially increase acceptance rates and decrease loan defaults or non-performing loans.
How Does Telco Credit Scoring Work?
The mobile phone has become a physical extension of a 21st century human. Whatever model, size, or function it may have, each mobile phone owner has a SIM so he or she can connect with the rest of the world. They use it to make phone calls, send messages, access a wide array of mobile applications — from social media to digital banking. As time passes, a user would create his or her mobile footprint such as, but not limited to texting usage, data usage, voice usage, top-up patterns, SIM age, and more. These are telco data that is unique to each user.
At FinScore, we use over 400 telco data variables to create a highly predictive credit score. Our proprietary, Artificial Intelligence (AI) and Machine Learning technology developed by our data scientists is the engine that provides the firepower in making these happen. Each credit score pull is being delivered in less than a second.
Here’s a sample scenario…
A borrower applies for a loan from a lender. Then, he provides his consent to allow a third-party credit scoring company to collect and analyze his or her data for assessment. The lending company then shares the mobile number with the credit scoring company. In FinScore’s case, we ensure that the Terms and Conditions of our partner financial institutions are amended with a specific consent statement in a way that adheres the Data Privacy Legal Framework in the country.
The lending company makes an “inquiry” to the credit scoring system. With FinScore, a client can pull a credit score which is available in their lending system (API integration) or the FinScore ACE (Alternative Credit Evaluation) web-based Scoring Portal, depending on their requirement. Using the declared mobile number of the borrower, the credit scoring company then pulls Telco data from a partner Telecommunications company. The Telco data will be used for analysis via a scoring system that runs on AI and machine learning algorithms, establishing a profile of the borrower such as willingness to pay and capacity to pay.
To check if the borrower shows an ability and willingness to pay, the system scans through patterns and its consistency over time. Examples of good behaviors show very stable patterns of the number of SMS sent and calls made over time, and the amount of data usage. As for the borrower’s capacity to pay, the system can look at the top-up patterns such as frequency and amount of top-up.
This can increase the approval rates by 25%, reduce bad loans by 50%, and reduce loan decisioning and processing time by a large fraction.
What should we look forward?
Banks and financial institutions of any size have been increasingly having positive perceptions towards Alternative Data Credit Scoring because not only can it increase their profitability but also reduce costs due to its speed and efficiency.
The COVID-19 pandemic has further pushed such businesses to transform the way they assess borrowers’ creditworthiness. Besides making their digital banking and/or lending mobile application available to the masses, their lending process, especially loan underwriting and disbursement needs to be fast, easy, and secure to keep up with the demand of the market and to stay ahead of their competitors.
But, what about the borrowers? How can they benefit from AI credit scoring?
One that is worth telling is that they have better, easier access to credit. If they apply for a loan from a lender that uses this particular type of credit scoring method, there is a higher probability for him or her to be identified as a trustworthy borrower versus if he or she applies for a loan to an establishment relying on traditional methods alone.
It is also important to note that telco data, though it has becoming a less and less of an intimidating type of alternative data source, is not the only data category that can be used for alternative data credit scoring.
Cooperating with enterprises such as power, water, and internet companies who serve a large volume of customers and billing them in recurring periods are also significant ways to determine the creditworthiness of a borrower.