CASE STUDY
89% Faster Loan Processing, Full ROI in Under 5 Months: Custom Document Intelligence for a Regional Mortgage Lender
Key Results
| Processing Time | Loan file processing reduced from ~45 minutes to under 5 minutes per file |
| Classification Accuracy | 95%+ document classification accuracy across 200+ distinct document types |
| Volume Handled | 4M+ pages processed annually with no headcount increase |
| ROI Timeline | Full investment recouped in under five months |
SERVICE AREA
INDUSTRY
CHALLENGE
Manual document review buried operations under four million pages per year
The challenge for a private mortgage banker servicing first-time homebuyers in multiple states through its branch network was that it produced more than four million pages of loan documentation annually across more than 200 separate document types. It is common for individual loan files to exceed 700 pages. Manually identifying, indexing, and routing each page became the bank’s most resource-intensive back-office function. While speed was an issue, so too was the fact that lending forms and government regulations are constantly changing and therefore create a classification risk. A classification risk is a misclassification that may expose the lender to regulatory issues or slow down the closing process. The operations group did not have the means to update the classification logic for the rate of change that regulations require, and hiring additional people to do this work would be cost-prohibitive due to the volume of documentation being processed. What they needed was a commercially available automatic solution that could handle large quantities of documentation, classify them correctly regardless of format changes, and reduce the amount of manual labor required by an overworked operations group.
SOLUTION
Custom document intake, classification, and routing application built for scale
SCIGON developed a custom document intelligence platform specifically tailored to meet the needs of the mortgage bankers’ unique loan processes. The system consisted of a high-volume production scanner combined with a classification engine that has been trained on the entire range of document types used by the mortgage banker. Bar-coded forms automatically moved through the scanning layer. All other documents were analyzed by the classification model, which then provided a confidence level-rated type for each page and routed complete loan files to subsequent processors — eliminating the need for manual indexing. From the outset, internal configuration was a major design goal. The mortgage bankers’ IT department can configure new document types internally when there are changes to standard forms and/or regulations requiring such updates, without having to go outside for professional services. SCIGON also configured the data structure to generate a database record for each document containing attributes for each document type. These records flowed directly into the mortgage bankers’ existing business applications, thereby completing the circle between document ingestion and loan processing. As a result, SCIGON delivered a commercial application built exclusively for scalability that removed the largest manual bottleneck from the mortgage bankers’ processes, allowed the firm to grow linearly with volume, and reduced time-to-decision on all loan applications.