Software That Fits Like a Custom Suit

I bought my first good suit off the rack. It was fine. The shoulders were a little wide and the sleeves a little long, and I adjusted the way I stood so it looked better than it was. Years later I had one made. The difference wasn't luxury. I just stopped thinking about the suit. It moved the way I moved.

Software works the same way, and most companies are wearing the rack version without noticing how much they've changed their posture to make it fit.

I've spent my career on the operating side of mortgage lending, consumer lending, real estate and manufactured housing. I started Guaranteed Home Mortgage Company in 1992 and grew it into a national lender. Today I run Millwright Logic, which builds software on top of the systems a business already uses. Here's what I've learned about where custom software pays for itself and where it doesn't.

The process bends to the tool

Off-the-shelf software is built for the average customer. That's its strength. It's tested, supported and cheaper up front. But the average customer doesn't exist. Every lender, servicer, property manager and community operator has its own way of doing things, and some of those differences are exactly why the business is good.

When the tool doesn't match the process, people don't stop working. They build a workaround. A spreadsheet gets exported every morning. Someone copies a borrower's information from one screen into another. A manager keeps a list of exceptions on a legal pad because the system has no place for them. None of this shows up on an invoice, so it rarely gets counted.

After a few years, the company has quietly reorganized itself around the limits of its software. New hires learn the workarounds as if they were the job. That's the slouch you develop to make the rack suit look right.

Where margin hides

If you want to find money in an operation, don't start with the big systems. Start with the gaps between them.

Manual re-keying is the most common one. In a mortgage shop, a loan officer takes an application in one system, a processor retypes parts of it into another, and a closer checks it again against the documents. Each pass costs time and creates a chance for a typo that turns into a condition or a delay.

Reconciliation is next. Servicing teams often spend hours each month tying out payments, escrow balances and investor reports across platforms that don't talk to each other cleanly. Much of that work is matching the same numbers in two places. A machine should do the matching and a person should review what doesn't match.

Handoffs are where things get lost. A web lead for a manufactured housing community goes to an inbox, then to a leasing agent, then to whoever sends the application. If one of those people is out for the day, the lead goes cold. Nobody decided to lose it. The process just had no owner between steps.

Exceptions are often the most expensive. Most software handles the normal case well. The trouble is the co-borrower who changed jobs, the home that needs a title correction, the resident who pays half by check and half online. Experienced people spend most of their time on these, and off-the-shelf tools usually offer nothing but a notes field.

Finding opportunities, not just cutting costs

Cost reduction is the easy pitch, but it's only half the value. The other half is what a business can do once it isn't fighting its tools.

A real estate operator who can see every inquiry, showing and application in one place can tell which listings and agents are actually converting. A lender who can launch a product variation in weeks, rather than waiting on a vendor's roadmap, can serve borrowers competitors are ignoring.

A worked example: TurnKey

The clearest example I can give is one I built myself. TurnKey started as our HomeBase platform and became a bespoke system for a public company client that operates manufactured housing communities.

Selling and leasing a manufactured home has more moving parts than most people expect. A prospect may want to buy with a home loan, may only want to rent, or may not see the right home listed yet. Each path needs different steps, and some states, Pennsylvania among them, have their own forms. Off-the-shelf tools covered pieces of this. None of them handled the whole thing the way this operator works.

So the system follows the operator's process. Applicants go through one residency and home-loan application, with a rental-only path and a "don't see the right home" option so nobody drops out just because inventory didn't match that day. Loan applications are delivered automatically into MeridianLink, so nobody retypes them. For each applicant, the system creates a document folder in ShareFile with a no-login upload link and writes that link back to the loan, so documents land in the right place without an email chase. Tenant screening runs through Settlement One. Insurance quotes come through Assurant's API. Rent Manager stays connected on the property side. Social Security numbers are masked and the application is hardened for security, which you have to take seriously when you handle loan data for a public company.

None of those systems was replaced. The custom work sits on top of them and removes the re-keying, handoffs and dead ends between them.

When custom makes sense and when it doesn't

I'm not going to tell you to build everything. I've seen companies waste real money doing that.

Buy the commodity. General ledger, payroll, email, a loan origination system, a core servicing platform. These are mature products solving problems that look about the same from one company to the next.

Build where you're different. If a process is specific to how you make money, if it touches customers in a way that sets you apart, or if your team spends hours every week working around a tool, that's where custom pays off. The best custom work usually sits on top of what you already own, as it does in TurnKey.

There are also times to wait. If the process is still changing every month, automating it locks in something you haven't figured out. If nobody can describe how the work is done today, the first project is writing that down, not writing code.

How to start small

The companies that get the most out of custom software rarely start with a big project. They start with one painful step.

Pick the workaround people complain about most. Measure it roughly: hours a week, errors a month, how often it delays a closing or a move-in. Then build the smallest thing that removes it. That might be an integration that ends the daily export, a rule that flags a file before underwriting, or a view that shows a leasing manager which prospects haven't heard back in two days.

Run it for a month. If the numbers move, take the next step. If they don't, you've learned something for a small price. And keep the people doing the work involved. They know where the problems are, and they'll decide whether a new tool gets used.

The fit is the point

A custom suit isn't about showing off. It lets you stop thinking about what you're wearing and focus on the meeting. Good bespoke software does that for a company. It fits the way the business runs, carries the routine work, and gives experienced people their time back for the files, deals and customers that need judgment.

Most operations I see have margin sitting in re-keyed data, manual reconciliations, dropped handoffs and unhandled exceptions. Finding it doesn't take a massive transformation. It takes someone willing to look closely at how the work actually gets done, and then build to fit.

David Allen Wind is the founder of Millwright Logic, a subsidiary of DAW Partners LLC. He founded Guaranteed Home Mortgage Company in 1992 and has worked in lending, real estate and fintech operations since. Find him on LinkedIn at linkedin.com/in/davidwind, on X at @davidwind, and at davidallenwind.com.

By David Allen Wind, founder of Millwright Logic and DAW Partners.