Ephesoft Mortgage FOR Financial Services
Financial Services Company
Utilizing the automated classification feature in Ephesoft Mortgage streamlined the identification of documents
- 50% decrease in employee hours spent on classification and verification of data
- 75% of backlog of nearly four million images processed in around four months
- Newly originated loans are classified, indexed, and uploaded to the repository automatically
The company is a dedicated sub-servicer of mortgage loans including loan due diligence and quality control, foreclosure trustee services for foreclosures in four non-judicial foreclosure states, and REO outsourcing. They employ more than 300 people and occupy seven offices in four states. They service mortgages and are required to transfer information and documents from one company and back to them. This translates into large amounts of workflow, data comparison, and data classification.
Along with massive amounts of mortgage assets and images that need to be classified—from seasoned pulls to new loans, all in various stages of their lifecycle—the company must also meet many compliance requirements.
Because of this, it’s very important that they are able to clearly identify and find the documents they have to ensure they are meeting regulations.
Additionally, the process was taking too long and needed to be drastically decreased.
Through in-depth discussion with Zia, the company felt they had found not only an Ephesoft third-party vendor, but a partner who was focused on creating the best long-term solution to address their critical business needs.
During the analysis, it was determined that they needed a solution that was scalable and configurable and had the ability to rename and load content into image repository.
They also needed to be able to pull data points and do a verification check against these images and data files.
Because pulls come from different providers, with varied image quality, naming protocol, and types, it was imperative that Ephesoft document classification be fine tuned as needed to ensure accuracy. As newly-originated loans come with only a month or two of servicing, the process is much simpler and more consistent because they are known documents. Before, the company sent newly-originated batches—of around 30–100 documents—overseas, taking several days to process. Now, small batches take only about 30 minutes to process in house. “It’s helped us tremendously,” said the Chief Technology Officer. “We have about five users who have processed around 75% of our backlog of four million images in just four months.”
In the past, pulls were being uploaded to an imaging solution but these images were not being classified correctly. When images were uploaded, they were labeled as FTP file uploads. Users would only see a generic name and indexes so they would be required to open every image the file to find what they were looking for.
Once Ephesoft Mortgage was in place, the company identified about 176 document types deemed critical or viable to identify—including everything from a letter to a mortgage statement—and updated their classification library.
Then, production images were brought down from their current repository, identified by investor, and each loan was run through the system for classification and export.
Now, as newly originated loans come in, they are classified, indexed by filename and images, and then uploaded to the imaging repository—automatically. The next step is to extract key data points in order to do a compliance validation check upon consumption of the loans. The next step in the roadmap will be to take the data file and run it against around 30 key fields in the images in order to check for discrepancies. Any discrepancies can then be flagged and sent through an exception process which will greatly benefit quality control in both post-closing and pre-funding.
Thanks to this solution, the company expects employee hours spent on these tasks to be decreased by around 50%.
In addition, the images are uploaded quickly to Ephesoft Mortgage, classified, data points are extracted, exported, and fed into the repository and into the appropriate folder. This not only cuts down on hours—allowing for improved SLAs—but also the overhead associated with quality control as only exception cases require personal attention.