The loan origination process can be manual and time-consuming, delaying individuals and businesses from receiving the money they urgently need. Yet, reducing the time to validate loan requests has been a major challenge for financial services providers because of errors in the evaluation process.
Launched in 2008, Perfios is a leading FinTech in India that recognized the need to automate the loan origination process and transform the traditional method of loan processing. To do this, the company offers its Perfios Insights solution to price and manage risk through its Data Aggregation and Analytics platform. By using a combination of machine learning (ML) techniques, the data platform extracts, categorizes, and analyzes unstructured data from hundreds of financial institutions, which then allows lenders to automate customer onboarding, decisioning, risk pricing, and fulfillment. Ramgopal Subramani, Chief Operating Officer at Perfios, says, “We are a product technology platform that deals with thousands of document types to help our customers automate data-driven risk-pricing decisions.”
From the start, Perfios chose to run its data platform on Amazon Web Services (AWS) and in 2016, it became one of the first few users to move to the AWS Asia Pacific (Mumbai) region, ensuring that financial data stayed in the country as per local regulations. Subramani says, “AWS offered us a broad set of ML services in addition to the data security to meet financial regulations.”
Traditionally, lenders would scan loan application documents using handheld devices while an optical character recognition (OCR) software in the backend extracted the data for analysis. However, the OCR technology couldn’t read data stored in tables or scans in mixed and skewed orientations, which resulted in errors during the extraction process.
To address this, Perfios is using Amazon Textract to automatically extract data from various configurations accurately, including bank statements in more than 200 different formats. As a result, Perfios ensures that there are fewer inaccuracies and manual reconfigurations required—reducing the average scan processing time by 20 percent. “The consistency of Amazon Textract in the way it extracts the data is a key part of the efficiency gain because we avoid the need to rescan bank statements,” states Subramani.
Once Amazon Textract has extracted and categorized the data from scanned documents, a Perfios decision engine calculates a credit score indicating the likelihood of an applicant repaying a loan. The score is based on data from the bank statements and algorithms identifying income and spending patterns for a 360-degree view of an applicant’s financial health. The FinTech then tailors its algorithms to satisfy the 175,000 lending rules for the loan organizations it works with.
To expedite building, training, and deploying its credit scoring algorithms, Perfios uses Amazon SageMaker. Ramgopal Cillanki, senior vice president and head of engineering at Perfios, says, “With Amazon SageMaker doing the heavy lifting, we’ve saved about 12 months of development time.”
By automating the loan origination process, lenders can disburse loans in almost half the time set by the Small Industries Development Bank of India. The entire Perfios workflow, from data extraction and categorization using Amazon Textract to the decision engine calculations, takes no more than 35 minutes.
Perfios is using Amazon Aurora to aggregate data held on loan applications, making it easily available to consumers and financial services business via APIs on request. It is also using Amazon Simple Storage Service (Amazon S3) for cost-effective object storage and Amazon Virtual Private Cloud (Amazon VPC) to secure and monitor network connections to lenders’ backend systems.
Today, Perfios is serving more than 200 lenders, including the top 10 banks and numerous financial institutions and FinTechs in India. By using Amazon Textract, Perfios is processing 2 million financial document scans per month, with the number growing by 15 percent month on month.
India is seeing a new generation of borrowers who have a large digital footprint but no credit history. With Perfios, credit decisioning no longer needs to be subjective and lenders can make data-backed decisions.
Subramani concludes, “India is changing rapidly, and its citizens are becoming increasingly aspirational. AWS AI and ML solutions make it possible for Perfios to drive the financial evolution, giving people and businesses faster access to finance.”
With ML on AWS, Perfios analyzes bank statements in more than 200 formats and processes loan applications in just 35 minutes. Perfios enables lenders to automate customer onboarding, decisioning, risk pricing, and fulfillment using its Data Analytics and Aggregation platform. The FinTech is using Amazon Textract to extract and categorize data from bank statements and Amazon SageMaker to score applications.