We Are Cancelling Manual Rekeying
Most purchases made today require only a handheld device and a credit card. “Most” purchases. And then there is the home loan which meanders into a room dawning a feather pen and long scrolling parchment, demanding 40 days of your time. At least, that is the perception.
The reality is that over time technology has and continues to make the home lending process easier. The parchment of the 1800’s made way for typed documents which made way for electronic document delivery and e-signatures. You could even say that the home-buyer documentation process is modern (ish) and has been for some years now.
But those modern capabilities did not necessarily translate to a better customer experience and that is because the documentation processes in the mortgage middle-office (where processing and underwriting analysis occur) modernized at a slower pace due to excessive manual data entry, data correction, and data rekeying. Hand cramps anyone?
The advent OCR technology improved the document-to-data process, but there was, and still is, room for improvement, especially when it comes to non-standardized documents (think self-employed borrower documentation). Even today, rekeying data is a common occurrence depending on lender technology sophistication despite advancements in optical character recognition (OCR) technology.
Indecomm is cancelling rekeying and taking a new approach to document processing.
Since the transition to the digital framework in mortgage is anything but linear, Indecomm took a comprehensive approach and leaned on subject matter experts in areas such as machine learning (ML) automation, SaaS technology, OCR capabilities, AWS technology, and mortgage operations to address the most time-consuming aspect of middle-office operations: document-to-data extraction.
With an end goal of accelerating the classification, reading, and processing of documents; saving costs in the crucial middle- and back-office stages of mortgage operations; and ensuring accuracy, Indecomm developed a new approach: its Intelligent Document Extraction (IDX), which recently won the AWS prize for best innovation in Machine Learning. The solution was deployed by Indecomm to form the basis of its proprietary Intelligent Data Extraction or IDX. IDX works alongside two other pillars to form Indecomm’s “Genius Tech Stack”:
1) Source Connect – an API-based linkage between a lender’s LOS and Indecomm’s systems, and ;
2) A rules-based engine returning conclusive loan-decisioning outputs to Indecomm’s users.
Thus, IDX is the custodian and data extraction center of its Genius suite. IDX is powered by AWS Textract – with the AWS Lambda’s output feeding into IDX’s Machine Learning-based data extraction model.
IDX performs the modular functions of:
- Document identification and classification
- Data extraction
- Data enrichment (if needed based on the document’s quality)
- Data certification
The architecture of IDX, powered by AWS and Indecomm’s proprietary ML-based model, has powered improvements within its Genius suite of products and improvement for the borrower experience. Machine learning specifically aids lenders in the transformation from manual operations to a repeatable and scalable processes. Virtual machines can contribute significant savings on metrics like process efficiency, quality, turnaround time, and scalability with or without a human to oversee the feedback loop. Not to mention, fewer hand cramps from data entry.
Using IDX, reading a document becomes a single, fluid action. In addition to its ML model and AWS Textract features, Indecomm’s IDX utilizes a variety of methods, including partial or complete image-to-text conversion, optical character recognition (OCR), and storage, among others.
Using Genius products which incorporate IDX gives your typing fingers a break. IDX is a repeatable, effective system in which all documents are parsed, assessed, and decoded for the information they can provide.
The Hands on the Clock Deserve a Break, Too.
It is true that the quantity of documents in a loan file varies depending on the borrower, and that number is occasionally too high. Be prepared to learn that a document might not be required after all! Just in case, some borrowers have been known to attach every document in their file to a loan application. This is a significant amount of extra time spent for a human agent sorting through a stack of proofs and will not in any way assist their loan application.
Fortunately, IDX can determine whether a document is required or not, saves time, and creates stacks of pertinent documents that are ready for reading. Processors and underwriters will benefit from significant time savings.
Extraction (IDX) allows computers (computer systems) to read documents of varied formats and readability levels as a preliminary step to judgments that are made. The characters on image files make comprehensible when the documents presented by borrowers are formatted appropriately. Anyone who works in processing and underwriting has probably made a few errors after hours of reading and staring.
Now, the information in the documents is ready to be unlocked thanks to IDX. Normally, reading the details and entering data manually takes hours but with IDX, it is reduced to minutes.
Data-driven Document Processing Powers Intelligent Decisions
Beyond the time savings, there is much to gain from solutions that improve the time and accuracy of document-to-data processing. Consider an income statement as an illustration. A borrower certifies that they get a specific sum of money each cycle of payments. However, based solely on this number, it is unclear if they would be able to afford mortgage payments. When deciding whether to lend to this applicant, a lender must additionally consider the applicant’s income tax bracket, credit lines, existing debts, and a number of other variables.
The relevant quantities (based on documentary evidence) for loan eligibility are evaluated during processing and underwriting to see if they adhere to a limited set of requirements. The application of a lender’s own best practices as well as the principles established by regulatory agencies is an act of intelligence and judgment.
Using IDX enables decisions that are based on formal, empirical criteria and quantitative methodologies pulled directly from the required documents. This automation removes risk associated with incorrectly entered data. In fact, AWS Textract achieved a document classification accuracy of 100% and a data extraction average of 99% for the average loan package.
In the truest sense, the devil is in the details!
Document processing is essential to many industries, with mortgage and lending being one such inescapable industry. Check out the case study on AWS’ website showcasing Indecomm’s solution and reach out to us to learn how IDX can benefit you!