Loan origination and boarding has largely been a manual and paper intensive process. Moreover, it has been observed that while loan production cost has increased at a CAGR of ~7%, , headcount productivity has decreased by more than half !
This means that loan origination & boarding process is not as efficient as it should be. This has serious implications for the business.
During the loan processing & underwriting, loan files containing dozens of documents totaling between 250 - 400 pages with thousands of data elements need to be verified and processed.
Manual handling of such documents and data is slow and error-prone.
As these documents move downstream, crucial errors might creep in - that may render the loan file unserviceable.
Such errors severely compromise loan quality and pose serious business risks to lenders and servicers.
In this paper we analyze the bottlenecks due to a manual & paper intensive process that leads to loan defects. Further, we investigate the key business risks that can stem from such poor quality loans.
Finally we build a case for mortgage document processing automation as a risk mitigation strategy.
Let us start by understanding a typical loan origination & boarding process from a bird's eye view.
Loan Origination & Boarding Process
Loan origination & boarding process is a complex one that involves multiple stakeholders & choke points.
From a bird’s eye view, this is what the entire process looks like:
Let us understand this in depth -
Mortgages are originated through multiple channels such as brokers, third party originators, correspondent lenders, co-issue, retail lenders etc. The origination process begins with the applicants submitting their employment details, financial statements, credit history etc for pre-qualification.
Once prequalified, applicants enter the loan processing stage where they need to submit various documents that usually run into hundreds of pages. These typically include - their rental payments history, bank account statements, tax returns, pay stubs or W-2s, gift letters etc.
Once the loan documentation process is complete, it goes through the underwriting, credit decisioning, QC, loan funding & closure stages. After loan closure, separate loan assets & MSR assets are created.
These assets undergo a post close audit when they are sent to a mortgage servicer for the sale of MSR. Alternatively, they undergo a due diligence process as a part of investor portfolio management.
At each step of the process - lenders, servicers and post close teams have to ingest bulk data in non-standardized formats (paper, scanned image, PDF, data feeds) & analyze them in a short span of time. This is a major bottleneck that can adversely impact loan quality and lead to serious business risks. Let us first understand the bottleneck.
Understanding the loan bottleneck
People intensive processes - leading to lower output
Mortgage origination & servicing operation has been inherently a people intensive activity. If we look at the break up of production expenses, then we see that 66% cost can be attributed to personnel expenses. Out of this, production & fulfillment staff accounts for 21% of the total cost.
Ironically, people heavy cost structure does not mean higher efficiency and better output. On the contrary, there has been a decline in average monthly productivity over the years. The below chart represents the average monthly productivity for retail production between 2003 & 2018.
It can be seen that during the period 2003 to 2018, there has been a steady decline of nearly 60% in the non-producing retail FTE productivity and monthly loan officer productivity.
Thus loan production being largely a people driven process is experiencing a significant drop in per person output.
Document indexing is time consuming & cumbersome
Document classification, sorting and validation is one of the most time consuming components of the loan origination & boarding process. On an average, it takes at least 14 days to complete activities like document indexing, validation and bookmarking just in the loan processing stage.
This is because loan packages consist of multiple types of documents running into hundreds of pages in different formats. In a manual process it is very difficult to judge the veracity of 500 complex pages of information in a short time and to do it repeatedly ! As loan origination and boarding is people intensive & suffers from declining productivity - manual document indexing becomes a costly proposition.
Data validation and updation is error prone
A big challenge with the loan package is that - a large portion of documents, even if digitized, are printed out during loan origination, closing and boarding stage. Any manual data entry or updates in the paper documents makes it difficult to do version control. This can cause a mis- match between data fields between paper documents and loan origination system.
Another challenge stems due to inaccuracies in data validation. A typical loan package of 250- 400 pages in formats ranges from digital documents like - scanned PDF, excels to paper documents like photo ID copies. Each such package consist of thousands of data fields that need to validated. This has 3 limitations.
Manual data validation has three core challenges:
- It is too tedious & error prone to manually verify all the data points.
- Second, when done manually, it is very difficult to track changes in customer data and update it across all the relevant documents and sub-systems.
- Third, federal & GSE regulatory compliance processes are embedded throughout the origination and boarding process. At each mandatory compliance step, new data must be collected and added to the core lending systems. With manual handling, updating data points as per compliance requirements becomes error prone.
Thus we see that manual handling of documents and data is cumbersome, inefficient & error-prone. The impact of this is poor loan quality.
Business risk due to loan defects
The problem of poor loan quality and business risks
Loan quality refers to mortgage loan files containing accurate and sufficient documentation that complies with pre- determined loan policies of an originating lender, loan guarantor, loan investor, and/or regulator.
Any aspect of the loan asset that does not conform to these pre- determined standards are called loan defects. Thus, loan quality is measured in terms of number and severity of such defects.
Higher the number and severity of defects, poorer is the quality of loans. It leads to two major business risks for all the stakeholders in the loan production and boarding value chain:
BUYBACK RISK FOR LENDERS
Defective loans could be sent back to lenders for correction. In worse case, there could be buyback requests from investors.
FINANCIAL RISK FOR LENDERS & SERVICERS
Lengthens the loan processing and boarding cycle, reduces per employee productivity of overburdened staff and increases overall production costs.
Thus we observe that bad loan quality due to defects is a major driver of business risks. In order to mitigate these risks, let us look at them in depth.
Buy-back risk: due to critical defects
Out of the three, critical defects are the ones that pose serious buyback risk as they render a loan package ineligible for sale. Consequently, a lender has to provision for losses associated with non-saleable loans. From a mortgage servicer’s perspective, critical defects translates into noncompliance issues and penalties levied by external auditors & regulators. Let's analyze some interesting trends.
Critical defect rate have increased over past year:
It is worth noting that overall critical defects have shown an upward trend in the last one year.
As per the latest survey conducted by ACES quality management, the overall critical defect rose by 25% from Q2’s 1.88% to 2.34% in Q3 2020.
Loan documentation package is one of the largest drivers of critical defects.
Loan documentation 2nd largest driver of critical defects
If we are to look into the composition of critical defects, then we see that loan documentation forms the second largest contributor at around 19% of all critical defects in Q3 2020.
Critical defects due to loan documentation grew by 80%
If we compare the growth in critical defects within each category, we see that critical defects due to loan documentation has seen an increase of 80% between Q2 2020 & Q3 2020 ! Thus loan documentation remains one of the major sources of critical defects.
Thus critical defects pose buy back risk for lenders and financial risk for servicers. Further we observed that critical defects have been on a risen during FY 2020-21, where in errors in loan documentation process were the 2nd most driver of such defects.
Financial risk due to operational overburden
Critical defects also pose formidable financial risks for lenders and servicers.
- Pricing adjustments associated with loan quality.
- Capital provisioning for defective loans
- Penalties for non-compliance
- Risks of possible early payment default and foreclosures
Another important, yet an overlooked aspect is the cost of reworking a defective loan.
With 30 year fixed interest rates still low at 3% and a housing inventory shortage, loans volumes are expected to stay high during FY 21- 22. Interestingly, ACES has observed that increased manufacturing related problems has a direct correlation to underperformance issues . This means, higher the volume of loans manufactured, greater the number of defective loans.
Thus loan production and quality assurance teams will continue to be overworked under a huge stockpile of loan volumes. Any re-work on the defective loan files, will further stretch the employees beyond their already surpassed capacities.
This will only result in further dip in productivity and rising loan production cost.
An analysis of historic loan production cost based on the data of MBA.org, it has been observed that loan production cost has seen a steady increase, from $ 3600+ in FY 09 to $ 7500+ in FY 20. This trend is expected to exacerbate further, if critical defects continue to bog down the existing resource bandwidth of lenders & servicers.
Loan defect mitigation: A case for document automation
Let us understand how does a mortgage document automation software reduce risk in the entire process.
In the above figure, there are three workflows sitting on top of one another. In the core data processing workflow, LOS is the primary tool for the loan processors and underwriters. It ingests data from point of sale and multiple data sources like banks, credit agencies etc.
At the same time, the document automation workflow ingests data from the point of sale. It offers functionality that LOS doesn't offer.
Document automation software automatically captures, indexes, classifies and stores documents. It then extracts data from structured and unstructured documents. As a part of the validation workflow, it compares these data points with those from the LOS.
It also collects & manages new version of documents that ensures consistency with LOS data. Most importantly, an automation software enables the lender/ servicer to identify exception and automate their resolution. This is critical because manual exception processing is slow, expensive and error prone. Finally, integrating a document automation software with LOS can build and deliver error free digital loan-packages to mortgage servicers, investors, document custodians and others.
So in a nutshell: HOW DOES DOCUMENT PROCESSING AUTOMATION MITIGATE RISKS ? Automatically & accurately captures & indexes various document types & formats. Extracts data from these documents & validates with LOS data. It also allows proper version control & automated exception handling. This minimizes the chances of missing data, wrong entry or manual data entry This reduces critical defects and eliminates buy back risk and financial risk for lenders & servicers
Document Processing Automation: Core Capabilities
Since we have understood the significance of integrating a document automation software with LOS to improve loan quality & reduce business risks. Let us look at the core functionalities of such a software. Document Recognition & Classification This involves automating identification of different doc & forms - 1003, pay stubs, deed of trust, W-2 etc. This is followed by indexing the pages mapped to the appropriate document types.Then, these documents are checked against a dynamic checklist to identify missing documents. Document Construction the system should provide a simple UI to view pages sorted by document type while separating out & labelling individual forms & documents. Data Verification & Exception Handling The capabilities for data extraction allow data to be collected from the mortgage file with both structured and unstructured data. Software platforms can extract more than 2000 data points from the ingested data set & assign a confidence score and a label (color code) to indicate whether it needs user review or not. Users can review these highlighted data points by simply clicking on the extracted value, to cross check the error.