Essay

America's largest builder owns the mortgage, the title company, and the price. One in five buyers still walk. They walked in the boom too.

The dominant proptech pitch of the last five years says the transaction is the thing to fix. Shave the friction, speed the paperwork, and deals move. D.R. Horton is yet the cleanest test of that pitch anyone will ever run, as it is the seller who owns nearly every lever a fix could pull. It sets its own price and its own construction schedule, there is no second seller haggling over repairs, and it runs the mortgage, the title work, and the insurance in-house. It has been America’s largest homebuilder by volume since 2002. And one purchase contract in five still dies before closing, through every market it has operated in.

Start with what the company reported on July 21. The fiscal Q3 2026 cancellation rate (cancelled sales orders divided by gross sales orders, the company’s own definition) came in at 20 percent, up from 17 percent a year earlier and 16 percent the quarter before. Net sales orders held flat year over year at 23,084 homes, closings rose 4 percent to 23,983, and the company cut its full-year revenue guidance to $32.5 to $33.0 billion from $33.5 to $34.5 billion. Executive Chairman David Auld expects sales incentives to “remain elevated during the fourth quarter.” Moreover, on the earnings call, the company said its average rate buydown actually eased to 1.6 percent from 1.7 percent, and CEO Paul Romanowski described the pullback as a decision to “hold margin a little more than push into the units.” This is a builder trimming the incentive dial and accepting the consequences, and its cancellation rate is a rounding error from its long-run average. The quarter is notably unremarkable.

How normal 20 percent is. CFO Bill Wheat said it on the call, verbatim: “within our normal historical range.” He is right. The rate ran 22 percent in fiscal 2017 and again in 2018, the healthiest housing market in a generation. It printed 19 percent in the fall of 2021, when mortgages were near 3 percent and buyers were camping outside sales offices. It spiked to 32 percent in the 2022 rate shock, then settled back. A decade of prints from the most frictionless seller in American housing, and the number has never left the high teens except to go higher. The floor does not respond to the variable proptech keeps tuning. Whatever is killing these deals, it is indifferent to how smooth the paperwork is.

What actually kills the signed deal is arithmetic about the buyer. Wheat again, when an analyst asked why buyers cancel: “qualification is still largely the biggest reason for cans.” Follow what that means. The buyer signs. Weeks pass (Horton’s median start-to-close time improved three weeks year over year, so this is a fast operation). Rates drift, the payment math moves, and underwriting says no. Here is the part that should end the friction debate: the builder owns the lender. DHI Mortgage has every incentive a lender can have to get to yes, and qualification still kills more deals than anything else, because qualification is a fact about the buyer’s balance sheet, and the buyer is the one input the seller does not own.

Now the strongest defense of the transaction pitch, which lives inside that same quote. Qualification failure is the financing leg of the transaction, and a whole proptech category aims exactly there: underwrite the buyer fully upfront, backstop the deal with cash. Those products are real and they catch some of these deals. But look at how they work: they swap a stronger balance sheet in for the buyer’s. That is a route around the buyer’s financing, and the category’s existence concedes the thesis: when the transaction’s own leg fails, the winning product replaces the leg instead of lubricating it. Horton already operates the strongest version of this machine. Its buyers in backlog carried a 4.9 percent rate against a market around 6.5, 160 basis points of payment relief, the most powerful lever money can pull on a mortgage. The floor held anyway.

And say the honest version of the other side too: the integrated stack is working. Closings grew 4 percent in a quarter when orders went sideways. The stack converts more of a bad market than anyone without one, and it is a fine business. What it has never done, across a decade of owning every lever, is push the share of contracts that close much past four in five. If friction were what killed deals, a decade of frictionlessness is where it would have shown up.

The proptech deck reads the closing table as a queue to be sped up. The builder’s own ledger reads it as weather. You do not de-risk a transaction-heavy business by attacking the transaction. You route around it. The builder can move the price, the schedule, even the rate. It cannot requalify the buyer.

Sources: