Onboarding in retail credit has to be reworked: Scienaptic's Joydip Gupta

Scienaptic AI, a credit-decision solutions platform offers technology to enable lenders to make more accurate and efficient decisions throughout a customer’s life cycle — from initial application to management. And it helps them revisit legacy onboarding and credit underwriting processes. In recent times, the Reserve Bank of India has called attention to the runaway growth in retail credit and the risks arising from it. Joydip Gupta, head of Asia Pacific at Scienaptic AI interacted with Raghu Mohan via email on the issues involved. Edited excerpts. 

 

Do you think it is time to review the on-boarding process for retail credit given the high incidence of defaults, more so in the unsecured segment? 

Absolutely. The surge in unsecured lending, coupled with rising defaults, highlights the pressing need to revisit and refine the on-boarding process. Traditional assessment tools, primarily reliant on credit bureau scores and static income documentation, often fall short in capturing dynamic financial behaviour and recent stress signals. Modern lending ecosystems now have access to richer data through alternate data sources such as Account Aggregators (AA). These platforms provide seamless access to a borrower’s financial data -- bank statements, insurance records, and investment histories -- with their consent. Integrating AA data into on-boarding workflows allows lenders to get a real-time, granular view of an applicant’s financial health.

Additionally, artificial intelligence (AI) can analyse these vast datasets to identify patterns and predict credit-worthiness far more accurately than traditional methods. By combining these solutions with insights from alternative data sources -- like utility bill payments, mobile recharges, or transactional data -- lenders can make sharper on-boarding decisions, especially for those with ‘thin-files’ (customers with limited or no credit histories) or new-to-credit borrowers.

 

Can you run us through some of the blind spots in the know-your-customer (KYC) process, especially when it comes to those with ‘thin files’? 

‘Thin-file’ customers pose unique challenges for lenders. While the KYC process verifies identity and address, it often fails to capture behavioural or transactional insights that are crucial for assessing financial health. For example, a borrower may have minimal credit history, but consistent digital payment patterns, regular mobile bill payments, or predictable cash inflows. There’s a growing risk of synthetic identity fraud, where real and fake information are blended to create seemingly legitimate borrower profiles. This issue requires advanced detection techniques, including AI-based anomaly detection models that can flag inconsistencies across datasets. Additionally, banks and lending institutions also need to perform KYC periodically. In today’s fast-paced environment, KYC details get out-of-date very quickly. Performing periodic KYC is expensive, requiring institutions to adopt 360 degree digital KYC techniques. 

 

How is it that data reported to credit information companies (CICs) is ‘dated’ even as credit dispensation is nearly instantaneous?

  While digital lending platforms now enable near-instantaneous approvals, most lenders still rely on batch-reporting models, where credit data is shared with CICs monthly or fortnightly. Once data is reported, the CIC internal processes around data quality checking and uploads adds to the cycle time. 

In case there are quality issues, the cycle has to be repeated, adding to the delays. During this reporting lag, borrowers can access multiple credit lines across different lenders, creating the risk of over-leveraging without timely detection. 

Additionally, inconsistencies in how data is formatted, verified, and shared between lenders and CICs contribute to further delays. 

One promising development is the promised rollout of Unique Loan Identifiers (ULI), which offer a standardised way to track individual loans across the financial ecosystem. Coupled with real-time data-sharing protocols and AI-powered monitoring tools, ULIs can enable lenders to flag risky borrower behaviour proactively, minimising exposure to defaults. Demand for credit data is instantaneous due to real-time lending requirements. However, its supply continues to come with a lag. For the credit ecosystem to truly match the speed of digital lending, CICs, lenders, and regulatory bodies must collaborate on enhancing data reporting infrastructure and adopting real-time, automated data pipelines. 

Last August, the Reserve Bank of India said that scores will be updated faster as CICs will get fortnightly reports from lenders (the deadline for the latter to comply with this format is FY25). Your views on the development. 

This will reduce information asymmetry. Fortnightly updates will allow lenders to detect early stress signals, identify patterns of risky borrowing, and take pre-emptive action. However, for this to take effect, lenders must restructure their processes and technology to follow a fortnightly reporting cycle. Any exceptions to the process, for example, repeat uploads due to data quality issues, must also be fortnightly. CICs will also need to accept the data, perform their quality checks, complete data upload process, and provide the feedback loop to lenders in a fortnightly cycle. Once the teething issues are sorted out, the system will definitely benefit the digital lending ecosystem. 

 

But given that bureaus’ data is based on past behaviour, how different will customer assessment turn out to be?  

Lending of the future will increasingly focus on real-time and forward looking behavioural indicators, in conjunction with traditional bureau scores. The future lies in hybrid assessment models that combine historical bureau data with real-time insights from AAs, behavioral analytics, and AI models. Alternative data sources (like e-commerce transaction history, mobile payments) can provide additional signals of financial discipline.