Fintechs have risen to the challenge of meeting new demands globally, with neobanks and challenger banks becoming household names in many locales.
Yet, such a transformation carries certain challenges, as procedures that were developed in an analog world have to be augmented in order to fit the digital landscape.
One such challenge is that of credit scoring. After all, lending is all about trust. The problem is finding the right signals about a given customer that allow you to determine (or deny) that trust.
This guide will walk you through not just the traditional approach and the philosophy behind credit scoring, but also how you can use open, public data about your customers to navigate the risky waters of underbanked regions.
What Is Alternative Credit Scoring?
In alternative credit scoring, alternative data helps support the loan applicant by requesting their payment history for other common services, such as phone bills or utilities.
This is a form of “lending trust”. If other companies trust this customer to pay their bills on time, the customer is deemed more likely to repay their loans as well. Proof of employment serves a similar role, as someone earning a regular salary is clearly less of a risk than someone who doesn’t.
Employment data can also be combined with sector data, allowing the lender to guess the income level of the debtor and make offers to them accordingly.
As you can see, this is a traditional approach fit for previous generations, where the credit score process relies fundamentally on services they were expected to use.
Enter social media and other online platforms….
The “online generation” is not only equipped with smartphones and a data plan; they are all over the internet, social media and subscribed to several online services.
Our online presence, otherwise known as our digital footprint, can say more than a thousand words about us. As such, this information can be a powerful form of alternative data for credit scoring models.
How to Collect Alternative Data for Credit Scoring
Digital footprint analysis can be extremely useful in credit scoring. In a nutshell, this is about performing a search for personal information based on someone’s internet usage to get a sense of who they are.
The reason it’s a treasure trove of data is that internet usage is increasing all over the world, even in markets with a high percentage of unbanked citizens.
Here are examples of the kind of footprint data you can get simply by enriching your basic KYC info:
- Device fingerprinting: The phone, computer or tablet that borrowers use to connect to the lending website contains tons of digital footprint info. Are they using private mode or an emulator? This could increase suspicion that they are not who they claim to be.
- Email profiling: Does the email address exist? Is it from a suspicious, disposable domain? Or one that doesn’t require any verification at sign up? Read more about reducing risk with reverse email lookup here.
- Phone analysis: Are they signing up with a real phone number? From a fixed line or a mobile? And do they use that number for messaging services?
- IP analysis: One of the oldest and easiest forms of security available: Search for the origin of the connection to your site. Is it from the right location? Or is it likely to be masked via Tor or a VPN?
Partner with SEON to minimize risk and reduce fraud rates in your business with ML, real-time data enrichment, and advanced APIs.
Ask an Expert
Within the context of credit scoring, you can turn that data into numerical values. As each market is different, the recommended approach is to look at your historical data (for good debtors as well as defaulters) and identify correlations between presence on different platforms and their quality.
The most obvious way to use social media data for credit scoring involves looking at a person’s friends. This is based on the axiom that “birds of a feather flock together”. The idea is simple to explain: Each person’s online friends are generally from the same socioeconomic class as they are. It is assumed that friends of good debtors tend to be good customers themselves and vice versa.
But there’s much more to consider.
It’s easy to use LinkedIn, for example, as a digital equivalent of employment proof, and LinkedIn users tend to be white-collar professionals who generally have better credit scores. Having a Spotify subscription or an active Airbnb account also hints at good spending habits, and so on.
The process in which you take the raw digital footprint information and use it to create a 360-degree picture of your customer is called social profiling – and it’s being deployed in a variety of fields that deal with risk.
How Does Alternative Credit Scoring with a Digital Footprint Work?
Fundamentally, digital footprints come most handy in markets where traditional bureau data is unavailable, other data sources are stale or inaccurate, or the target group is simply underbanked. All these situations make it difficult for lending organizations to meaningfully assess risk.
Furthermore, when it comes to mobile services, users expect a frictionless experience. The more hoops they have to go through, the sooner they get fed up and go to the competition.
This is one reason why there is good value to using the email address or phone number that’s provided by the user on registration to find them on social media and other platforms – thus creating their digital footprint. Let us explain…
Why Is a Digital Footprint Important for Alternative Credit Scoring?
For one, someone’s digital footprint proves that the person does exist and is probably legitimate – since they’re willing to give you the details they use in their online lives, as opposed to a fraudster, who would likely use a throwaway address or a burner number, which do not have the expected digital presence.
How Does Alternative Credit Scoring Work with Social Profiling?
Depending on the user’s country, you can reasonably expect that good debtors and bad debtors will have a different mix of sites and services for which they are signed up.
By looking at these links, you can assign different values for their presence on these various services, and factor them into your credit scoring.
For example, a fresh user who doesn’t have an account with their country’s largest messaging platforms could indicate a bad-faith actor, whom you can block or ask for further documentation – such as income statements.
On the other hand, someone who has active GitHub, LinkedIn and Skype accounts is likely to be a programmer, who might be easier to trust within certain contexts.
Such an approach can augment your credit scoring models in a way that allows you to take on customers who would otherwise look too risky because of their underbanked nature. Moreover, it also lets you screen out malicious actors even before they reach the stage of mandatory KYC verification.
Generate more accurate credit scoring to measure risk by leveraging alternative data, and real-time data from digital and social profiles
Ask an Expert
Challenges for Credit Scoring Bureaus
The first problem traditional credit bureaus face is that data can be falsified or stolen. We are talking about two clear forms of fraud, namely identity fraud and synthetic ID fraud.
According to the US Federal Trade Commission, the latter is the fastest-growing form of identity theft. This takes place when criminals steal part of someone’s real ID (name, social security number or bank account) and combine it with fictional data.
So, how do fraudsters acquire data that helps fool credit scoring bureaus?
Well, one method is to use stolen data from children or teenagers who have a clean transaction history because of their age – which goes to show that fraudsters will stop at nothing to be able to borrow money from loan companies.
Digital Lending and Risk
One consequence of the increase in smaller, frictionless loaning, is that it puts lenders at risk.
Fraudsters are increasingly targeting microloan offers, using a combination of known attacks and techniques:
- Identity theft: Used to open a new account with stolen credentials without any intention of repaying the debt.
- Going AWOL: Fraudsters take advantage of loaning organizations’ need for frictionless experiences. They open accounts quickly, borrow funds, and vanish.
- Account takeover: Fraudsters phish for current account holder information and exploit it to borrow money for themselves.
- Synthetic identities: Fraudsters combine the personal information of real people into new synthetic IDs, lay low and use your service as normal. Then, when their credit opportunity is at maximum, they “bust out” – taking out a massive loan and disappearing.
In fact, every fraudulent dollar in a digital loan costs the lender $2.82, a much higher cost than other industries including ecommerce, retail and financial services.
Battling Multiple Forces at Play
Competition isn’t the only challenge for lenders. A growing trend towards frictionless user experience means legitimate borrowers want to create new accounts and access their money in no time. A delay in the application or a complex digital verification process can turn them away in seconds.
Digital lenders are therefore under increased pressure to battle fraud while trying to keep their heads above water. If they refuse a loan, there is no shortage of competitors who will happily take the risk.
Microlending companies in China are growing too fast for legislators to remain in control.
Clearly, there is an urgent need for frictionless solutions.
Running Their Own Background Checks
In fact, fraudsters are smart enough to test their data through their own background check systems before using it in the real world.
Since there are paid services and websites that let you perform credit checks online, the fraudsters can simply build their application file and use a stolen credit card to purchase access to their results.
Once they know the credit score of the person they will be pretending to be will be high, they can apply for a loan with a higher success rate than the average borrower.
In some cases, fraudsters even acquire the assessment criteria from background check companies, and simply manufacture the right user profile to meet all the right criteria.
Digital ID Profiling for Scaling KYC Processes
However, lenders may not realize it, but they are already in possession of a tremendous amount of information about their potential borrowers. In the digital world, any data point can be leveraged to paint a clear picture of who your users are.
As we’ve pointed out, you can gain a lot from aggregating the right information at the right time.
The Future of Alternative Credit Scoring
Technological advances have allowed banking transactions to be simpler, more cost-effective, and smaller.
Over the last few years, commercial banks and the new wave of digital-only financial institutions have therefore embraced microfinancing, a sector that was valued at $178.84 billion in 2020 and is projected to reach $496.90 billion by 2030.
There is some good and bad to this. For instance, it allows an increasing number of previously unbanked individuals to access financial services, particularly women in rural areas, as has been stressed by the UN on several occasions.
The challenge in this environment for lenders is two-fold: After the 2008 crisis, younger consumers opted out of using credit cards or taking loans, so they don’t seem like “good” debtors from the point of view of the traditional credit scoring system – similar to segments that are traditionally underbanked.
However, savvy cybercriminals know very well about the different security loops that are in place in lending, and fast digital loans look like attractive targets to them, which they can exploit easily.
Connecting the Dots
Traditionally, credit scoring has been calculated based on payment history, debt burden, requests for new credit, types of credit used, and the length of the credit history. This is now believed to be an outdated system that is holding back the economy.
The question is how to redefine credit and credit scoring. Millennials distrust big banks and increasingly turn away from credit card offers, which means historical payment or credit card data is less relevant than ever. On the other hand, the meteoric rise of buy now pay later services is driven by millennial and Gen Z spending habits, with 60% of the former and 57% of the latter using this type of scheme.
Lenders can leverage the power of big data analysis to generate personal fraud scores and new credit risk models, all thanks to the power of machine learning.
Key Takeaways of Alternative Credit Scoring
Building a solid, efficient online credit scoring model in the digital era is challenging yet entirely possible. The challenge is about trusting the personal information you acquire through standard KYC processes and avoiding fraud from stolen or synthetic identities.
The good news is that users’ digital footprint is increasingly large and available to aggregate with the right tools.
Using a phone, email or social media profiling solution, you should have no problem enriching data and digital footprints from users who are unbanked – or simply confirming their identity.
Once a digital footprint helps you know they are who they claim to be, it will be easier to understand if they should be granted access to your products, how much they could borrow, and how fast they can pay the funds back – all helping you to ensure your business can grow safely.
Sign up to download
In order to download this PDF, please sign up to our newsletter.
Thanks for submitting the form, click the button below to download our guide.
Click the button below to download our guide.