Printer Friendly

The infrastructure predicament.

Panic is setting in! In my 40 years in the residential mortgage business, I have never witnessed a higher level of frustration and concern by C-level executives who run U.S. mortgage companies. This article provides a replay of a recent series of conference calls I have had with a number of chief executive officers who are very concerned about the state of the operational infrastructure that supports their respective businesses. [paragraph] These calls have been condensed to a three-way WebEx[TM] conversation involving myself (Terry) and fictional CEOs Drew Smart and Bill Legacy. As Frank Webb, star of the old TV series Dragnet (now I am really dating myself) would say, the names (except mine) have been changed to protect the innocent.

Bill Legacy: Thanks for arranging this call, Terry. Drew, how are you doing?

Drew Smart: Thanks for asking, but I have had better days. I just finished a board meeting, and my team and I were grilled about the explosion in our mortgage loan production expenses, the reduction in our transaction velocity and the mediocre customer service surveys we've gotten back from our borrowers over the past year. In our defense, my chief operating officer presented a series of figures that I'll share with you. Can you see my screen?

Bill Legacy and Terry Wakefield (simultaneously): Yeah, we see it.

Drew Smart: Here's Figure 1, average loan origination and production expense.

One of our key directors runs a business that manufactures wireless office equipment. When he saw this figure, he literally screamed: "How can any industry survive a cost increase of that magnitude? If my math is correct, costs have increased by nearly 230 percent since 2004. This is lunacy!"

We pointed out the impact of reduced production volume, escalating costs associated with the regulatory environment, and our existing infrastructure's dependence on humans to orchestrate work that leads to a very expensive hiring-and-firing cycle to adjust to volume cyclicality. He was not impressed and said, "We have a structural problem that needs to be addressed--or we should consider getting out of this business!"

Reluctantly, we moved on to Figure 2--average days to close.

Another one of our directors runs a logistics company that provides consulting services to local companies with large vehicle fleets. He's a little more even-tempered, and asked if we had statistics that track our transaction velocity, or average time to close, going back 10 years. Fortunately, we were prepared for that question.

From 2003 through 2009, our transaction velocity averaged 15 days from application to closing for refinance transactions and 24 days for purchase money transactions. After a few seconds of silence, he asked if we understood the impact of transaction velocity on revenue and customer service.

Our chief financial officer responded that the current, much slower transaction velocity has reduced our annual per-transaction revenue by approximately 50 percent. The director sat stunned.

Another director asked how this reduction in transaction velocity impacted our customers' perception of our service levels. I produced a report that showed only 60 percent of borrowers we gave mortgages to within the last 12 months responded that they would come back to us for their next mortgage. The board chairman and founder of our firm rose to his feet and barked in his Southern drawl, "Houston, we have a problem."

Bill Legacy: Drew, I hate to admit it, but the declining profit--I mean, growing losses--in our mortgage operation caused a similar experience at our board meeting two weeks ago. I did not present the same slides that you did, but I was told to have a plan in place to restore profitability to the mortgage operation within the next six weeks or the topic is going to shift to exiting the mortgage business.

It didn't help that I had to report we had just finalized a settlement in the amount of $220 million with the Federal Housing Finance Agency [FHFA] for mortgages delivered to Fannie Mae in the period of 2005 to 2008.

Our chairman remarked we are now officially a member of the "club" that has paid over $120 billion--and rising--in buybacks, legal fees, settlements and fines to various regulators and secondary market investors. He asked if this is the end of our exposure. I could not reassure him.

Terry Wakefield: Bill and Drew, I can assure you that you are not alone. We hear stories like yours every day. Some executives blame the Consumer Financial Protection Bureau [CFPB] and/or FHFA. Some blame the inevitability of volume cyclicality. Others blame government policies that encouraged the extension of credit to people who could not possibly repay their mortgages. Others blame greedy loan officers and unscrupulous third-party loan originators.

There is no question that the subprime crisis was fed by a chain of greed that permeated the entire spectrum of mortgage lending and servicing, but playing the blame game ignores the real problem.

Drew, your director is correct. The mortgage industry has a structural problem that relates directly to outdated operational infrastructure. Fortunately, a small but growing number of lenders are facing facts and admitting that their current infrastructure is to blame and must be reconfigured if they are going to prosper in this business.

Bill Legacy: Terry, I agree with you--but every time we explore options, we are faced with evaluating the capabilities of the traditional loan origination system [LOS] vendor community. Something is missing. What is it?

Drew Smart: Ditto. Where can we turn?

Terry Wakefield: Before we go down that path, let me take control of the WebEx and show you Figure 3--the infrastructure collapse.

This dives a little deeper into Drew's first figure. Like Drew, we subscribe to the Mortgage Bankers Association's [MBA's] Quarterly Mortgage Bankers Performance Report.

Historically, loan production direct labor expense represents approximately 60 percent of loan production expense. But, in first-quarter 2014, the industry reached a historic high of $3,315 per closed loan. In second-quarter 2014, things improved, but loan production direct labor costs remain over $2,600 per closed loan.

This stunning statistic points to the industry's fundamental infrastructure problem. Humans continue to orchestrate the work.

Since 2003, we have conducted detailed process architecture analysis in 34 different loan production environments representing all channels of origination. While there is a perception that individual lenders have some form of "secret sauce" that drives their respective loan production environments, our experience demonstrates just the opposite.

All lenders perform the same tasks, but not necessarily in the same order or by workers with the same titles. Let's face it--for the past several years, more than 90 percent of all mortgages have been purchased or securitized by government-sponsored enterprises [GSEs] Fannie Mae, Freddie Mac or insured by the Federal Housing Administration [FHA], So, if 90 percent end up with the GSEs or FHA, one would think that a common manufacturing process would exist to produce mortgages in the United States, and that platform would significantly reduce production costs over time.

Bill Legacy: Terry, I think I know where you are headed. A few weeks ago, I had a three-hour meeting with our COO, and we met informally with 20 of our most experienced loan production personnel. They were honored we would take time to have a frank discussion on how they did their jobs and what we could do to make them more effective.

After an hour or so, light bulbs started going on. It was clear that each of these capable individuals had their own workarounds to overcome the deficiencies of the systems they use to produce loans. These workarounds take the form of Post-it[R] notes on computer screens, notebooks in desk drawers or, worst of all, memorized tasks that reside in their respective minds. We have 220 people in this particular production environment. I finally asked the audience: "Do we have 220 different processes in place?" They all agreed.

Terry Wakefield: Bingo! Bill, you just hit the nail squarely on the head. How is it possible to control costs, ensure compliance and deliver great customer service when you have 220 undocumented and unsanctioned processes in place? It's impossible.

Drew Smart: Terry, you and I have had this discussion before. Lenders do not think of themselves as manufacturers. We tend to obsess over the complexities of mortgage lending rather than focusing on the common attributes of loan production. We're always emphasizing increased market share and topline revenue growth. It sounds like we need to shift greater focus to process definition and technology that can execute a uniform, well-defined process.

Terry Wakefield: Let me offer a relevant quote from Bill Gates here. He said, "The first rule of any technology used in a business is that automation applied to an efficient operation will magnify efficiency. The second is that automation applied to an inefficient operation will magnify the inefficiency."

That absolutely nails it. I cannot recount how many times we have witnessed lenders applying technology to an inefficient process. The outcome is as predictable as Gates' quote--inefficiency magnified.

Bill Legacy: OK, I am not going to quibble with Bill Gates. But I have six weeks to convince my board that we should stay in the mortgage business. Let's talk about an action plan that will make the case resonate.

Terry Wakefield: The first thing to do is to use proven tools to define the current state in your loan production environment. You're already off to a good start based on your three-hour meeting. I would involve those 20 people you have already gathered and break them into their respective functional divisions of labor. Interview them to find out 1) the names of the tasks currently performed; 2) who performs the task; 3) when pipeline fallout occurs; 4) how frequently the task occurs; 4) the calculation of adjusted task time to the fraction of a minute; and 5) the calculation of task cost to the penny.

My guess is that after three to four weeks, you will have concurrence from these 20 people that there are currently 300 to 350 different tasks taking place in the current state. These tasks result in a direct labor cost of approximately $2,600 per closed loan.

Once current-state definition is completed, you will have a valuable baseline of direct labor cost data that you can use to conduct return on investment [ROI] analysis on the technology investments necessary to achieve an optimized state.

Drew Smart: Terry, I am confident that we already have a good handle on our loan production direct labor cost. Our current direct labor loan production expense is right around $2,800 per closed loan. Can we bypass this first step and move on?

Terry Wakefield: Sure. Each time we have analyzed a loan production environment, we have stored the output into what we refer to as the INPLORE[TM] Task Level Database. INPLORE now contains detailed information on more than 3,700 specific loan origination and production tasks. Using INPLORE in collaboration with your production managers and supervisors, it is possible to create what we refer to as an optimized production time summary [OPTS], It is critical that senior compliance personnel are involved in this effort to ensure that the OPTS output represents a compliant, optimized production environment.

Documenting the optimized state focuses on three objectives:

* Removing non-value-added activities. We often hear frustration regarding workers checking and rechecking the work performed by others.

* Improving workflow, consolidating job functions and eliminating functional divisions of labor. Functional divisions of labor create expensive bottlenecks in cyclical industries like mortgage lending.

* Defining tasks that can be automated through deployment of software components not available in traditional loan origination systems.

Definition of an optimized process includes task level metrics similar to those used to define the current state. That is, the OPTS describes: the human or system that performs the task; the impact of fallout; task frequency; adjusted task time; and adjusted task cost.

Upon completion of process optimization, the number of tasks performed in the loan production environment is typically reduced to a range of 110 to 130 tasks--down from 300 to 350 in the current state. Of those 110 to 130 tasks, somewhere between 40 percent and 50 percent are automated, meaning that the task count performed by humans is reduced from 300 to 350 down to 55 to 78 tasks. Direct labor production costs are reduced to a range of $550 to $650 per closed loan, depending on the production channel.

Drew, in your case this would produce a savings of $2,150 to $2,250 per closed loan. So, if you are closing 20,000 loans per year, you will realize direct labor cost savings of more than $40 million annually.

Drew Smart: You definitely have my attention.

Bill Legacy: Mine, too. So, how does this process architecture optimization work translate to automation?

Terry Wakefield: That's the $64,000 question. Once the optimized state is defined at the task level, it must be thoroughly documented at a step level. Think of tasks as defining what work must be performed and steps as describing how each task is performed. Documenting step-level descriptions of task performance is hard and unglamorous work.

Bill Legacy: How is this possible, given the variability that surfaces as loans are produced?

Terry Wakefield: That's a great question, Bill. Task-level detail addresses this variability by defining required tasks that apply to all loans and random tasks that occur on an unscheduled basis or are required based on loan-specific circumstances. Variability is a reality in most manufacturing processes. The next time you're on the freeway, focus on the variability of cars you see. Managing this variability requires human definition of an optimized manufacturing process and a heavy dependence on technology.

Bill, to answer your question about automation, think of task-level detail as the sheet music that allows best-of-breed technology components to automate work orchestration and eliminate human interference with the optimized process. Because task-level detail defines the optimized process down to the step level, it also serves as the training manual for new employees and those who want to improve their value to the enterprise by learning cross-functional skills. Thus, task-level detail ensures that there is no disconnect from what workers learn and how they actually perform their work tasks.

Bill Legacy: So, task-level detail is the "code" that drives software performance. What are these software components that are not typically imbedded in an LOS?

Terry Wakefield: Before I describe those software components, let me emphasize that current LOSes do perform many functions very well. They have solid product and pricing engines, most are effectively integrated with settlement services providers and documentation preparation firms, and they can all ingest output from point-of-sale systems used by loan originators. So, I am not suggesting that you replace your LOS. However, the LOS needs to be demoted in importance so that the software components I am about to describe enable automated work orchestration.

Drew Smart: I am happy to hear that, because we just spent a lot of money changing our LOS.

Bill Legacy: We did the same last year--and to be honest, we have not experienced the productivity gains we expected.

Terry Wakefield: Bill, we hear your message from a lot of lenders. Getting back to Bill Gates, no technology will improve an inefficient process.

Bill Legacy: So, let's talk specifics.

Terry Wakefield: Before I describe these software components, I should mention that they are all commercially available and have proven their value in many industries. Inexplicably, they are not prevalent in the mortgage industry.

The first of these is business rules management software [BRMS]. Traditionally, business rules have been hard-coded directly into applications, like an LOS. Business logic is the most volatile aspect of a business application; however, the pains of maintaining rules in application code have become more obvious and significant as the pace of business demands more agility. Agility requires that you externalize business rules logic and manage this logic in a way that is easy to build and easy to integrate as decision services while still meeting your enterprise performance and scalability demands.

Drew Smart: The last word I would use to describe our loan production environment is agile. We are very dependent on our LOS vendor to effectuate change, and it costs us a lot of money. Have you had direct experience using business rules management software?

Terry Wakefield: We recently completed a project for a retail mortgage lender to build a prototype with one of our business partners. In a matter of days, one of our senior process architects authored hundreds of rules to automate dozens of mundane tasks contained in task-level detail. No programmers were required--just a gifted process architect with an intuitive understanding of business logic as defined in task-level detail.

Bill Legacy: I get the concept of extracting business logic from a hard-coded application. In fact, our head of IT met with a BRMS vendor at the last MBA technology conference, and he was impressed. Let's hear more.

Terry Wakefield: The next software component is business process management software [BPMS], All commercially available BPMS vendors deploy process-mapping tools to create process maps that drive automated work orchestration.

Going back to the prototype project, we took six tasks in task-level detail and used a leading BPMS vendor's process-mapping tool to create a process map for each of the six tasks. Using easy-to-use click, drag and drop functionality, the same process architect I referred to earlier completed the six process maps in two days. Again, no programmers involved.

Drew Smart: What do you mean by automated work orchestration?

Terry Wakefield: Simply stated, BPMS pushes work tasks to the right worker at the right time. Workers do not request tasks; tasks are pushed to workers as the BPMS engine determines when a task needs to be performed. Each time a task is pushed to a worker, the worker's performance is monitored in real time per the metrics in the OPTS and task-level detail. If a worker has not performed a specific task in a while and is unsure of the steps to follow, each task pushed to a worker is accompanied by a link to that task's task-level detail. So, there is only one source that guides each worker's performance--the BPMS process maps that emanate from task-level detail. If a task is not in task-level detail, it cannot exist in the production environment.

Bill Legacy: So, the problem of having 220 different processes disappears?

Terry Wakefield: Exactly, Bill--but let me go further. As certain tasks designated by compliance or management are performed, BPMS triggers automated internal and/or external communication through a set of management-prescribed alerts, escalations and scripted messages. The time-stamping of tasks, the tracking of worker performance and the monitoring of task outcome creates an audit trail of all task performance in the production environment. This level of auditability is precisely what the industry needs in order to proactively respond to the demands of the CFPB and other regulatory agencies.

Drew Smart: We have our first CFPB exam in two months, and we are more than a little nervous.

Terry Wakefield: That's a good segue to the third critical software component--enterprise content management software [ECMS]. To deal with the avalanche of paper received from external sources and produced internally, we advocate establishing a document management center [DMC]. It would focus on digitizing all data received from external sources using ECMS and depositing all digitized data in a borrower-specific virtual loan file.

Based on our experience, there are approximately 4,000 discrete data elements in a closed loan file. Advances in ECMS have made it possible to extract those 4,000 discrete data elements from the documents they reside on, and deposit those data elements into the virtual loan file along with an image of all external and internally produced documents.

Bill Legacy: Wait a minute. Are you suggesting that we process all of our loan transactions electronically? Is that even legally possible?

Terry Wakefield: E-SIGN [Electronic Signatures in Global and National Commerce Act] legislation enacted in 2000 permits a lender to conduct a mortgage transaction in a purely electronic fashion, provided that the borrower consents. The way communication patterns are changing, I am convinced that a growing percentage of borrowers will consent to an electronic transaction versus the painful fulfillment experience that currently plagues the industry.

Drew Smart: So, we may need two fulfillment environments--one that caters to borrowers who consent to an electronic transaction and another for those who don't.

Terry Wakefield: You're right, but I would look at it with a different twist. You already have the non-electronic environment. Why not launch a greenfield project to build the electronic environment, and expand it as more borrowers consent? You can control the expansion of the electronic environment by inviting borrowers as electronic capacity scales. I'd bet that within three years, you'll be processing more loans in the electronic environment than in the current one.

Bill Legacy: Terry, I have to get to another meeting in 15 minutes. Can we wrap this up?

Terry Wakefield: Sure, Bill. Let me conclude.

Assuming the borrower consents to an electronic mortgage process, let's look at the consequences. The DMC's role in digitizing all external documentation received from the borrower, and storing the images and extracted relevant data from those images in a borrower-specific virtual loan file, prevents any paper from entering the loan production environment and enables a paperless workflow. Production workers have access only to the virtual loan file as tasks are pushed to them. BPMS provides access to only that segment of the virtual loan file that is relevant to the task the worker receives.

Best-of-breed ECMS continually populates the virtual loan file with images and data extracted from external documents and all documents produced internally during the loan production process. BPMS continually updates the virtual loan file with data on worker task performance.

So, in concert with BPMS, ECMS enables a borrower-specific virtual loan file that includes an audit trail of:

* all of the approximately 4,000 discrete data elements aggregated during the loan origination and production process;

* the performance of all tasks contained in task-level detail, including the system or worker who performed the task, task duration, task cost, task outcome and task-specific alerts, escalations and messages;

* all internal and external communication, including voice, email, text messages and social media interactions that occurred during the loan origination and production process;

* all documents received from the borrower and other external sources, and all documents produced internally, including federal and state-specific disclosures and time stamps of when all documents were delivered to the borrower; and

* all secondary market transaction documentation and, if you also service loans, all loan servicing and loss-mitigation transaction data.

Basically, the virtual loan file provides 100 percent data transparency of the entire loan life cycle. This is profound compared with today's environment, where loan-level data is trapped in myriad systems and databases.

Now, the coup de grace. The virtual loan file allows secondary market investors to conduct their quality control of loans prior to the funding of the loan by the lender, basically eliminating buyback exposure. The tension between mortgage lenders and secondary market investors is totally unconstructive and has imposed an expense of more than $120 billion on the residential mortgage industry. The technology exists to dissipate this tension and re-establish trust through the deployment of best-of-breed technology.

Drew Smart: Pardon my skepticism, but do you have any additional information you can send me and our COO that goes into greater detail? If you do, and it validates this call, we'll want to meet with you soon.

Bill Legacy: Terry, send it to me as well.

Terry Wakefield: I'll send you both a PowerPoint[TM] deck later today. Thanks for your time today. I really appreciate it.

Terry Wakefield is chief executive officer of The Wakefield Company LLC, Mequon, Wisconsin. He can be reached at


            Average Days to Close

            Avg.    Avg.   Jan    Feb    Mar    Apr
            2012    2013   2014   2014   2014   2014

Refinance    49      45     44     40     37     37
Purchase     46      45     47     42     41     40
All          48      46     45     41     40     39

SOURCE: Ellie Mae Origination Insight Report, Sept. 2013


Total loan production expenses    $6,932
(per loan)

Origination expense at 1% and    ($2,500)
assumes an average loan amount
of $250,000

Loan production personnel         $4,423

Loan production direct labor      $2,659
expense at 60%

SOURCES: MBA's Q2 2014 Quarterly Mortgage Bankers
Performance Report, The Wakefield Company
COPYRIGHT 2014 Mortgage Bankers Association of America
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2014 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Title Annotation:TECHNOLOGY
Author:Wakefield, Terry
Publication:Mortgage Banking
Article Type:Discussion
Geographic Code:1USA
Date:Nov 1, 2014
Previous Article:Staying power: Denver-based Accenture Mortgage Cadence now has the deep pockets of a global company to power its growth and innovation.
Next Article:A manufactured market.

Terms of use | Privacy policy | Copyright © 2020 Farlex, Inc. | Feedback | For webmasters