Intelligent Document Processing for the Mortgage Industry
- Automated the classification and extraction of mortgage documents
- Increased worker productivity by 700%
- Reduced closing document processing times by 67%
- Reduced HUD-1 processing times by 67%
- Reduced human errors on HUD-1 forms
- Streamlined the conditions clearing process
- Redistributed staff from manual tasks to high-value projects
- Achieved data accuracy rates of 90% or greater
- Increased closing capacity without adding headcount
For more than a decade, this mortgage company has been providing specialized mortgage services and low mortgage rates to brokers around the United States.
As one of the largest mortgage wholesalers, the company receives between 15,000 and 25,000 documents each month from brokers during the conditions clearing process of a mortgage application. Once brokers upload the required documentation using the secure, online portal, an employee reviews each document to verify it was classified correctly before sending it to the underwriter. This manual process was time consuming and tedious. During busy periods, the conditions clearing queue could contain hundreds of documents—delaying the entire mortgage loan process.
The CIO wanted to automate the conditions clearing process using optical character recognition (OCR) technology. This would automatically classify separate, sort, and extract data from loan documents regardless of format—paper, PDF, or electronic formats.
“After researching OCR and intelligent document capture technologies, we selected Ephesoft Mortgage for its ease of use, rich capabilities and affordability,” commented the CIO.
He saw the potential to use Ephesoft in other areas of the closing process to classify and extract information on HUD-1 documents, streamlining the loan closing documentation process.
“We met Zia Consulting through Ephesoft,” he said. “It was clear from our first meeting that Zia had tremendous domain expertise. More importantly, they had implemented OCR solutions for other financial institutions and understood the unique demands for the mortgage industry.”
The experts at Zia worked with the company on a discovery and process assessment. Zia was able to work with their team to understand their specific needs and develop a project roadmap to implement OCR technology across multiple process steps including the conditions clearing queue, HUD-1, and closing documents. As part of a phased approach, Zia helped the company implement and customize Ephesoft for each process.
Zia trained the IT team on the software so that they could test and train the system to achieve accuracy ratings of more than 90 percent.
Focusing first on the conditions clearing process, Zia was able to leverage the Ephesoft Web API and incorporate advanced document classification and extraction capabilities directly into the portal. Unlike other OCR technologies, Ephesoft API allows seamless integration into other platforms for a one-step capture process. Zia customized the document upload section of the portal to map directly to common condition clearing document types. Now, when a broker uploads documents during the conditions process, the user sees the document type requested in the drop-down menu. Once the document is uploaded, it automatically goes through OCR document recognition confirming the document type.
The company has trained the system to achieve an accuracy rating of more than 90 percent for most document types. Today, the company only requires one staff member to review the remaining ten percent of condition clearing documents. As a result, they have redistributed seven staff members to higher value projects.
“We have seen a seven-fold increase in productivity by implementing OCR as part of the conditions process,” says the CIO. “This is a huge savings for the company and provides instant verification of documents eliminating any process delays during this stage.”
Advanced Closing Extraction
To streamline the closing process, Zia partnered with the company to co-develop a mortgage tool based on Ephesoft that is tailored to recognize and classify common closing documents eliminating the need for bar codes or separator pages. Since closing documents are standard across the industry, Zia created customized OCR guidelines and trained the system to classify more than 200 document types using documents from the company.
Closing documents sent to the company through mail/hard copy, PDF, scanning, or electroniccally, are automatically split into individual documents, sorted, stacked, or imaged using Ephesoft Mortgage. The company has fine-tuned the system to their specific company requirements for a document accuracy rating of 95 percent. The company estimates it has cut down the processing times for closing documents from 15 minutes to five minutes per loan.
“We have tested Ephesfot Mortgage with millions of documents and are excited to roll it out across the organization once it’s fully integrated it into our current systems,” says the CIO. “We anticipate tripling closing capacity without having to hire additional headcount.”
As one of the post important documents in the closing, HUD-1 forms are carefully reviewed and verified by the company before each closing in accordance to governance guidelines. Traditionally, this requires a loan officer to manually review the document and key in the information into a custom-built system that leverages sophisticated business rules to verify the calculations.
Working with Zia, the company plans to leverage Ephesoft Mortgage to extract and export data from HUD-1 and GFE documents. Zia is setting up the system to extract more than 1600 types of data contained on three-page HUD-1 forms. By integrating Ephesoft Mortgage directly into their current HUD analysis tool, the data can be quickly analyzed and reviewed by the loan officer. The system is testing at 95 percent accuracy and the company anticipates a 67 percent reduction in HUD-1 processing times.
“Data accuracy is critical on HUD-1 documents,” says the CIO. “Using OCR technology, we are able to meet compliance regulation and add another layer of protection into our process by significantly reducing, if not eliminating, human errors. This translates into thousands of dollars saved for the company.”