What If Mortgage’s Flywheel Theory Is Bullshit?

Mortgage is obsessed w/ a flywheel: originate the loan, keep the servicing, recapture the borrower & do it all over again.

But Karen Postiglioni asks a pretty sagacious question in today’s Mortgage Scoop: What if owning the whole damn thing isn’t actually the best use of your capital? She digs into mortgage’s beloved servicing/origination “natural hedge,” the billions tied up in MSRs & why smaller IMBs may have more strategic options than they think.

Plus, I’ve got fresh details on layoffs at FICO, unreported cuts at several major mortgage firms, some intriguing Seller Guide changes at Fannie, UWM’s new cap markets guru & a bunch of people moves across the industry.

Most of that 👆 is for paid subscribers, so click here if you wanna get 10% off an annual subscription.

A quick note for subscribers: I’m closing on a condo Thursday & moving Friday, so the next edition may land this weekend. I’ll also have a list of parties, events & things to do — & definitely not do — in Chicago for MBA Annual.

By Karen Postiglioni

For years, the mortgage industry’s biggest players have been building or buying every turn of the flywheel.

Originate the loan → Own the servicing → Keep the customer → Recapture the refinance → Originate a new loan → Create new servicing → Repeat

That’s the integrated model. It’s also the foundation of the ‘natural hedge’ between originations & servicing. In a simplified rate cycle, here’s what happens to each side:

And that’s the hedge. Rates rise, production gets harder & existing servicing becomes more valuable. 

Rates fall, borrowers refinance & that servicing value comes under pressure. Originations are supposed to take the baton back, capture the next loan, create a new MSR & go around again.

You can see why the heavies like Rocket & Pennymac have built scale on both sides. Servicing brings payment interactions, borrower data &, theoretically, a better shot at the next transaction. If you have the capital, why wouldn’t you want the whole flywheel?

The answer is b/c the capital has to earn its keep.

The hedge has real value. MBA found that 78% of lenders in its ‘25 sample were profitable w/ servicing included; w/o servicing profits, that figure would have been 64%. But that still doesn’t tell us what return they earned on the capital committed to MSRs.

And MSRs eat up a lot of capital. They also come w/ financing costs, hedging, liquidity needs & servicing obligations. So obviously servicing helps offset a lousy original market. But lenders should be thinking more along the lines of whether owning both actually gives you the best return on capital over a full cycle.

And what, exactly, are lenders buying w/ that capital? Servicing income & the hedge, yep. But part of the strategic pitch has always been access to the customer too.

Those pieces ☝️don’t necessarily have to come together, especially today.

In Q2 2026, Rocket reported selling MSRs associated w/ $53B in unpaid principal balance for $795M, while retaining subservicing & recapture services on nearly 80% of what it sold. It got paid for the asset w/o walking away from much of the customer-facing work.

This example doesn’t tell us if Rocket has resolved its own ROIC question yet. Rocket’s arrangement obviously also doesn’t mean a smaller IMB can sell servicing & expect to keep the same payment interactions or recapture rights. It does show that owning the MSR & maintaining access to the customer are separate decisions. A smaller lender has to find its own way to stay relevant.

A smaller IMB has choices though. It can look for an MSR buyer that won’t compete for the next loan, or negotiate who may solicit those borrowers after the sale. The contract matters. Some deals even restrict the seller’s ability to pursue a refinance. So a lender can’t sell the asset & assume it kept the customer opportunity. It has to protect and price that opportunity in the deal, then do the work to maintain the relationship itself.

AI may push that separation further in two ways. Let’s talk about that!

First, capacity. Historically, being ready for the next wave meant carrying extra people through a slow market or scrambling to hire & train them when volume returned. A lender automating more of its workflow may be able to hold increasingly more standby capacity in technology instead of payroll. It’ll still need people, but production capacity may no longer have to rise & fall so closely with headcount.

Second, the relationship. Selling servicing can mean losing valuable borrower contact & data. It doesn’t have to mean disappearing from the customer’s life.

An IMB w/ permissioned contact, useful data & actual reasons to stay relevant can identify likely financing needs & put a person in front of the customer when it matters.

That takes work, yes. Totally. But AI could make that more practical across a large book of former customers than ever before. It doesn’t manufacture trust or hand an originator the servicer’s payment data. The lender still has to maintain & earn its own relationship.

But that may be a better investment than deploying capital on a multibillion-dollar MSR book just to get another shot at the borrower. 

Here’s the distinction: a lender that sells its MSRs gives up the MSR hedge. It may also free up capital to pursue the next transaction another way. The hedge is valuable only if the return it produces beats what that capital could earn elsewhere, after accounting for the risk.

That’s the opening for smaller IMBs. They don’t have to become Rocket or PennyMac to compete for the next loan. They can invest in customer engagement and a production platform that responds when demand returns, then compare those returns w/ what they could have earned owning servicing.

Some will find that MSR ownership wins, while others may find they wanted the customer opportunity more than they wanted the asset.

So before another lender decides it needs a servicing portfolio to complete its flywheel, I’d ask: Are we chasing servicing because it produces the best return on invested capital, or because that’s how the game used to be played?

Fannie Mae is loosening the leash a bit. The GSE is ditching its prescribed methods for verbal verifications of employment, essentially telling lenders to use a method that’s “reasonable, verifiable, and appropriate.” Timing rules still apply of course.

There is one new hoop. Starting Dec. 2, lenders will need a verbal VOE when a borrower qualifies using Schedule K-1 income from a business they own less than 25% of.

Fannie also loosened documentation rules for non-contingent job offers, moved one income-duration test to the application date & expanded use of Attorney Opinion Letters, including for Texas 50(a)(6) loans.

Our freshest reporting, including layoff news nobody else has reported yet, is behind the paywall. If you want the scoops before they get recycled into everyone else’s newsletter later this month, click here & become a paid subscriber.