The economic prosperity of a country and its citizens hinges on responsible access to affordable credit. Monumentally important decisions like buying a home, getting a business loan or even a medical procedure become cost prohibitive for many without credit.

However, despite the high stakes, credit decisions are still largely based on a decades-old credit scoring system with anywhere between 10-20 basic variables mixed in for good measure. They’re largely static, meaning that they’re based on if/then functions and lack any flexibility or context. 

This practice often leads to consumers being categorised either based on their bureau data or excluded as "new-to-credit." Undoubtedly, it is logical for risk teams to prioritise credit bureau data as their primary source of reliable information, especially when dealing with individuals who have good credit scores. 

However, the coverage of credit bureau data in the country remains insufficient, which stood at 63.1% in 2019. Moreover, even among those considered credit active, only a meagre portion possesses credit scores that justify underwriting. 

Traditional credit scoring comes with its own set of challenges - 

• Partial hit rate and the resulting asymmetry between scorable and unscorable populations

• False positives without clearly knowing the expected Probability of Default

• Mostly based on bureau data which does not cover recent bounces and loan profiles as bureau data has a lag of 45 days

• Misses behavioural data- digital savviness, conscientiousness, digital footprint, online portals registered on, credit hungriness, utility bill bounces and payments

• Bureau scores are the same for all portfolio types irrespective of the type of loans, sourcing, and ticket size 

All of this translates into loan books that are both smaller and less profitable than they could be with better credit analysis. To combat these issues, lenders require new ways to enhance access and affordability — they need a risk model upgrade built for today’s consumer.

That’s where alternative data-based risk models come in! 

The catch-22 of modern credit is that in order to borrow, you need a score; but to generate a score, you need to have borrowed before. That’s where alternative data comes in - data from non-traditional sources such as mobile devices, social media interactions, app data, credit and debit transactions and much more. At FinBox, we call it the customer’s digital footprint that we derive from their devices. This includes - 

1. Device metadata 

2. Apps data 

3. Masked transactional SMS data 

4. Location data

Image 1

Alternative data-driven credit risk scores involve various data sources as shown in image 1

Based on these data sources, the FinBox Inclusion Score or FIS is derived. The FIS is an amalgamation of diverse customer segmentations (as seen in image 1)  with statistically defined weights, resulting in a credit risk score which is far more powerful than any traditional bureau score for underwriting digitally acquired portfolios.

The main goal of a good credit score is to distinguish between borrowers who are likely to be good and those who are likely to be bad. To that end, FIS offers a more precise differentiation within and across credit score categories, ultimately enhancing the accuracy of risk scores. 

In simpler terms, better credit scores will be achieved by concentrating higher numbers of risky borrowers at the lower end of the score range. By combining traditional credit bureau data with alternative data, the relationship between the score and the likelihood of default should improve. This improvement should lead to a noticeable decrease in default rates among higher and mid-level credit score bands, while lower credit score bands may experience an observable increase in associated default rates. 

The FIS Score adds significant value to underwriting for all digital lenders, through the features listed below - 

1. Coverage: FIS has broad and consistent coverage; currently,  coverage stands at about 95%, and is available and works with the same efficacy for NTC and Bureau Rated customers; making all customers aptly scorable. 

2. Specificity: FIS contains detailed data elements about an individual—data elements that provide part of a full picture of the borrower (e.g., on-time and late payments over a significant time series, or specific asset or income data);

3. Accuracy and timeliness: FIS uses primary data from customers' devices and hence is updated observing bounces/transactions even from seconds ago, making it much more effective to observe the recent behaviour of the customers. This makes data used to generate FIS real-time, as opposed to traditional bureau data that is updated every 45 days. 

4. Predictive power (‘signal’): FIS contains information relevant to the behaviour that a lender is trying to predict and encompasses behavioural aspects as well (like social media activity, e-Nach payment patterns and bounces etc, and gambling apps on devices). By itself, FIS boasts a univariate Gini of 34.1% and a combined Gini of 40.1% with Bureau data, on average, which is unseen in digital portfolios. 

5. Orthogonality: FIS acts as an additive to traditional bureau data; this means that using it will improve the predictive accuracy of any new score by improving the signal-to-noise ratio. 

6. Customisation: FIS creates custom scores for different portfolios, which would ensure more value for the portfolio. 

When combined with any bureau score, FIS gives an additional lift to lenders’ underwriting strategies, and in the process, is better able to approve more worthy borrowers. 

As illustrated in the table below, X is the average delinquency rate of the portfolio. Even the worst borrowers in the bureau bucket fit into the best bucket of FIS. And as a result, the risk is 0.7X compared to 1.35X when it lies in the best bucket of the bureau and the worst bucket of FIS. 

In simpler terms, even the riskiest borrowers as per bureau scores find redemption with FIS, resulting in more approvals and lower defaults for lenders, and one step closer to financial inclusion. 

FinBox's DeviceConnect, which generates the FinBox Intelligence Score (FIS), is a user-friendly SDK integrated into your app, ensuring a 100% hit rate for all digital customers. We prioritise data ethics and have a strong commitment to it (you can find more information here). 

The data collected is in the form of anonymous metadata, guaranteeing that no personally identifiable information (PII) is ever transmitted from the user's device. This data collected undergoes processing to evaluate various factors, including the probability of customer payment default, detection of fraudulent applicants, and the enrichment of customer segmentations. 

The FIS is real-time and provides quick insights. It is based on data with an impressive track record, boasting a 99% correctness rate. 

FinBox stands out in the market, with an extensive set of over x0 engineered behavioural features and a proprietary data modelling pipeline that sets the industry standard

Our implementation process is designed to be easy and flexible for our customers. It can be seamlessly integrated into a lender's IT infrastructure using FinBox's adaptable API and SDK, reducing time-to-approval and simplifying the client onboarding process.

The usage and importance of alternative data can be observed by examining the expenditure trends over the past few years. As depicted in the chart below, spending increased from $232 million in 2016 to a staggering $1.71 billion in 2020. This exponential growth is truly remarkable and can be described as absolutely astonishing. 

For lenders looking to stay on the right side of risk, the FIS might prove to be beneficial - not just for growing their loan books, but by being a significant part of the larger financial inclusion story in India!

To know more about what we do at FinBox, check out our resources page. 

You can also get in touch with us here !