Brands
Discover
Events
Newsletter
More

Follow Us

twitterfacebookinstagramyoutube
Youtstory

Brands

Resources

Stories

General

In-Depth

Announcement

Reports

News

Funding

Startup Sectors

Women in tech

Sportstech

Agritech

E-Commerce

Education

Lifestyle

Entertainment

Art & Culture

Travel & Leisure

Curtain Raiser

Wine and Food

YSTV

Amazon Web Services

Amazon Web Services

View Brand Publisher

How Bengaluru-based Perfios is automating credit underwriting decisions for 500+ global BFSI companies

How Bengaluru-based Perfios is automating credit underwriting decisions for 500+ global BFSI companies

Saturday December 19, 2020 , 8 min Read

When V.R. Govindarajan and Debasish Chakraborty realised that there was no tech solution that provided a 360-degree view of an individual’s finances across different financial institutions and products, the serial entrepreneurs decided to build one themselves. The duo started to work on a solution that could enable individuals to efficiently manage personal finances. That effort led to them starting Perfios in 2008 as a one-stop shop for personal finances Previously they had built Aztecsoft, a large Outsourced Product Development (OPD) company that went public in 2000 and was acquired by Mindtree in 2008.


In its initial years Perfios was a B2C Personal Financial Management (PFM) solution built on the proprietary Perfios Data Platform. As they made inroads into the sector, the founders realised that the platform could address the challenges faced by enterprises in the banking, financial services and insurance (BFSI) sector. “Real-time credit decision-making was an unexplored territory. It was almost non-existent. However, we felt that the Perfios Data Platform could make it happen,” shares Debasish, the Co-Founder and Director, Perfios.

Disrupting credit-decisioning

That thought led them to explore the enterprise side of offerings, and develop Perfios Insights -- a statement analysis and credit analysis product for loan applicants. The team launched Perfios Insights in 2012. And, in less than a decade, it has become a category creator and the de-facto digital solution for Loan application processing for banks, NBFCs and fintechs. Today, Perfios serves more than 500 financial institutions. Its data analytics platform is used by leading banks, NBFCs and fintech companies across 18 countries, including India, Singapore, Malaysia, Indonesia, Vietnam, Philippines, Hongkong, UAE, Middle East and Africa.

Leveraging a powerful combination of AWS technologies and home-grown solutions, Perfios Insights bridges the gap between data and insights to enable enterprise customers to automate data-driven risk- and price-reducing decisions. Perfios extracts, categorises and analyses thousands of document types in real time, helping financial institutions take decisions in stringent privacy and compliance environments. Perfios uses Artificial Intelligence (AI) and Machine Learning to automate what was once a cumbersome and human-intensive process.

Developing an integrated and comprehensive solution for financial underwriting

The Perfios platform has been designed to analyse e-statements, online bank statements as well as paper-based bank statements for credit-decisioning. Ramgopal Cillanki, Senior Vice President, Head of Engineering at Perfios, shares, “With the three capabilities, we have been able to offer an integrated solution that provides a consistent loan processing journey for Perfios clients.”


What makes the Perfios platform unique is that it supports 1700 bank formats for e-statements and data retrieval from the bank’s online websites. This makes it easy for banks and financial institutions to choose Perfios as the default bank statement analysis provider. “Nobody has the kind of coverage that we have. We support the remotest of banks to the smallest of banks and process the statements. We support national banks, co-operative banks, and even rural banks. This way the lender does not have to worry about the category of customers they can cater to.”

Decoding the technology layer of market-winning product

Perfios evaluated a number of technology solutions to complement its home-grown platform, says Cillanki. Finally, the team onboarded Amazon Textract - the fully managed machine learning service that automatically extracts printed text, handwriting, and other data from scanned documents that goes beyond simple optical character recognition (OCR) to identify, understand, and extract data from forms and tables. Amazon Textract not only quickly automates the manual document activities but also enables the processing of document pages at scale. Explaining what makes Amazon Textract valuable, Cillanki says, “The traditional scanned processing journey is often fragmented and poor. Lenders often receive bank statements that are account passbook copies printed using dot matrix printers. So readability is often an issue. Second, lenders receive documents which are photocopied versions of photocopies of original copies. All this means that we rarely receive documents that have been scanned in controlled environments. If not for Amazon Textract, it becomes very challenging to digitise these documents. Even if a single decimal point is missed, it could result in huge variations, thereby completely skewing the analytics.”


Cillanki also points out that Perfios has built additional capabilities on top of Amazon Textract to enable form extraction to suit specific demands. Explaining how this works, “If you hold accounts in multiple banks, you will observe that the e-statements vary not just in form but also description and process. But Perfios is not just able to read and extract the data but also categorise it.” Perfios’ home-grown solution and Amazon Textract have enabled a unique web flow and pipeline for processing these documents which has enabled them to bring the advantages at scale.

Powered by ML

In addition to extracting, digitising and categorising the data, the other key aspect that makes Perfios unique is its Machine Learning capabilities. “Once the documents are uploaded on the Perfios platform, it performs analytics based on the bank’s rule engine and provides information that helps make credit decisions,” shares Cillanki. Perfios has built a repository of 175k+ rule engines to cater to different financial institutions.


Machine learning algorithms are used in its flagship Bank Statement Analysis product to identify the categories of the transactions by mining the transaction narrations. “The variations, semantic variance and other offshoots of the narrations are all addressed by the machine learning algorithms to have an accurate categorisation,” shares Cillanki.


He explains that the neural networks are used in the customer analytics and risk analytics part of Bank Statement Analysis application. The likelihood of the applicant to default in the loan repayments are measured in probabilities and then converted to risk scores.This score helps the lending organisation in their credit underwriting process with better risk mitigation and pricing mechanisms.


Perfios also onboarded Amazon SageMaker. A fully-managed machine learning (ML) service, Amazon SageMaker removes the heavy lifting from each step of the machine learning process and enables developers and data scientists to build, train, and deploy high quality ML models quickly. “With Amazon SageMaker we were able to build a product that we call a Debit Score Card. Similar to a CIBIL score, the Debit Score Card arrives at a score based on the bank statements, spend patterns, income patterns, among other things, and helps assess the credit-worthiness of individuals without credit history.” He adds, “Ideally, if you had to build this product, the data engineering and data pipelining aspects of the solution would have required humongous effort for a company like ours. We wanted to develop the solution quickly and deploy at scale, and this is where Amazon SageMaker fit in perfectly,” says Cillanki.


He shares that with the maturing of AI and ML technologies, challenges related to data, variety and prediction can be solved more efficiently and effectively. “Today, by leveraging AI and ML technology we have been able to address a very critical challenge in the BFSI industry in a very unique way. Earlier the minimum time for the manual credit-decisioning process was a week. Today, by leveraging ML capabilities, our home-grown solutions and AWS managed services like Amazon SageMaker and Amazon Textract, we have been able to bring down that time to an hour. And, we are continuing to see a 25-30 percent improvement on an on-going basis.”

Advancing farther and further

Headquartered in Bangalore, Perfios has been seeing a steady and impactful growth, especially in the last three years. “We have been growing at a rate of 100-150% year-on-year and expect the growth rate to be at least 100% year-on-year for the next three years,” shares Govindarajan, Co-Founder and Director, Perfios. He adds, “The projected growth rate will be the result of the expansion of our product portfolio, a robust international expansion strategy, and the strengthening of our relationships with our existing customer accounts.” The startup raised a


Series A funding of $6.1 million led by Bessemer Venture Partners in 2017 and followed up with Series B funding of $50 million led by Bessemer Venture Partners and Warburg Pincus in 2019. Recently, in 2020, the startup was recognised by FICCI for its ‘formidable growth story’ and by CB Insights as one of the Top 250 Global Fintech companies using technology to transform financial services in the Credit Score and Analytics space.


Today, even as Perfios continues to increase and improve the depth and breadth of its product portfolios, it continues to expand its presence across the globe. “ We are looking at taking our strong partnership with AWS in the Asian region to other geographies too. AWS' scale and global footprint makes it easy for us to offer our Products & Solutions to our clients in these new regions in extra quick time. We also hope to explore and adopt cutting edge offerings from AWS in our suite of Products to enrich them,” says Cillanki, signing off.