How fintech is transforming the underwriting process in India

fintech

As modern-day consumers leave behind their digital footprint, understanding user behaviours and financial transactions becomes more important in building unique credit models, optimized for different segments. Fintechs are able to deliver customised credit solutions, with better and faster customer engagement, using the latest innovative methods like Big Data, Artificial Intelligence (AI) and Machine learning (ML). Fintech is changing the credit and finance industry.

Until a few years ago, underwriters had to go through loan papers manually to decide the creditworthiness of a person. The manual process could take up many days or even weeks to complete. Since the entry of new-age Fintech companies, the dynamics of the underwriting process have changed considerably. In today’s digital world, it is humanly not possible to look at so many data points, correlate and decide on the creditworthiness of borrowers.

The availability of digital data at the micro and macro level helps in training credit models with insights that were not available during the days of the manual process. The modern-day approach to lending involves creating algorithms that categorize borrowers based on their risk profiles. Training credit models with machine learning helps with more precise predictions, free of human errors.

While the traditional underwriters reject loan applications of customers who are new to the process due to insufficient data, Fintech companies make use of alternative credit scoring to benefit these customers. This process involves extracting data of the borrower from multiple sources and analyzing them into buckets or segments. These segments can now be used as touch points to slowly help the individual secure a loan in the future through the use of automated follow-ups, remarketing and cross-referencing.

Alternative Credit Scoring
Many people with a steady source of income – both salaried and self-employed – do not pass conventional bank loan processes due to strict and outdated credit underwriting processes. Credit rating fintech companies are taking a new approach by considering alternative data points and percentile scoring amongst similar borrower groups.

All these insights combined with an intuitive algorithm can lead to better credit decisioning over time. With this, lenders can reach their financial inclusion goals faster and serve different segments with product customizations based on employment type, demography or both.

For example, digital platforms lend to customer segments earning between INR 1.5-4.5 lakh which form a large number of Indian households named the next billion. Within the next billion, there are segments that are classified further based on salaried, self-employed or gig workers. Customised credit models help in assessing each of these unique segments and build products contextual to their users. All this in quick time due to the modularity and structure with which these credit models are built with.

An important type of customer who benefits from alternative credit scoring is the New-to-Formal-Credit (NTFCs). NTFCs are not new to credit. They are customers who were denied loans from formal institutions and had resorted to informal lenders such as loan sharks instead. The new-age credit scoring
helps such customers get a fair chance to access formal credit and avoids them getting into a debt trap.

Advance Data Sciences Models are proven to be stable
With simplified user experience, more advanced technology and language customization, platforms can offer better user engagement for customers to transact securely. User engagement provides multiple data points leading to a more refined understanding of customers. The study of user behaviours will help in improving prediction accuracy and hence augments the ability to build more products customized to suit the needs of different customer segments.

Also, Open banking, API integrations and strategic distribution models ensure that credit models are fed with unique data on an ongoing basis to make them more robust. These data are aggregated into machine learning algorithms. ML-trained models have proven to perform well even when there have been disturbances in the ecosystem, such as the Covid-19 Pandemic.

Credit Models are continuously evolving
Technology has created a level playing field for easy credit access by removing human biases. The data-driven approach combined with process automation helps in scalability along with customization to segments – specific to user requirements.

Data Sciences based credit models are not only easy to moderate but also can be replicated to offer products contextual to different segments. These models continuously monitor correlating variables and adjust weights accordingly, to churn out the best predictions on risk profiles. This is a continuous self-learning loop mechanism. The models are built in modular structures, which helps in customization based on various available data sets. Being modular makes them more scalable.

Another important aspect to highlight is the fact that many financial institutions fail to recognise how monitoring accounts in the post-lending stage acts as a key feedback mechanism to both the credit risk model as well as early warning systems. Digital lenders today use advanced tools to continuously engage with users as opposed to a one-time interaction during loan disbursement. AI-enabled chatbots and voice bots help in understanding customer expectations more closely and provide data-driven customer feedback for fintechs to act upon.

Fintech that stays on top of technology trends and makes data science-driven models a priority is able to achieve scale and make an impact on the underserved and unserved markets. Financial inclusion is a reality, thanks to the advancement in the newer techniques in credit modelling.

(This article is written by Monish Anand, Founder and CEO of MyShubhLife. The views expressed in this article are of the author.)

Related posts

Leave a Comment

This site uses Akismet to reduce spam. Learn how your comment data is processed.