The 10-K warranty line item that challenges builder scale claims

In Texas, before you buy a ranch, you walk the fence line, check the water and look at the roads. And if somebody tells you everything is perfect, you may just look a little harder for something that isn’t.

Investors in public homebuilders might well do the same.

Every public builder tells Wall Street how many homes it closed. Revenue. Backlog. Average selling price. Gross margin. Sales pace. Land owned versus controlled. And, of course, scale.

But another number is buried much deeper in the 10-K: How much money does the builder expect to spend fixing the homes it just sold?

That number exists, yet almost nobody talks about it. I pulled the warranty disclosures for 12 of the largest publicly traded U.S. homebuilders I’ve been following throughout this Chasing Scale series.

This time, I wasn’t looking for the biggest. I wanted to know what each builder was betting against its own work. But what I found raised a bigger question. This is not really a story about warranty reserves.

It is a story about whether America’s largest homebuilders can prove that scale makes them better operators, not merely bigger companies.

Every builder is already making the bet

When a builder closes a home, it estimates how much it will eventually spend addressing warranty issues, based largely on historical claims experience and assumptions. Think about what that means.

Management has already looked back at what has gone wrong with its homes and estimated what is likely to go wrong with the homes it is selling today.

Every 10-K includes some form of warranty reconciliation: beginning reserve, new accruals, claims paid, adjustments to earlier estimates, and ending reserve. But one useful number almost nobody gives you is warranty accrual per home closed.

So, I calculated it. Toll Brothers is reserving roughly $3,800 per home. LGI Homes is reserving roughly $384. Nearly ten to one. That ought to make you curious. But let’s be clear: Warranty accrual is not a quality score. These companies aren’t building identical houses. A Toll Brothers home and an LGI home are not the same product, at the same price point, with the same complexity. Comparing their raw warranty dollars as though they were the same would be like comparing the maintenance costs of a Ferrari and a King Ranch pickup and declaring one better built.

Context matters

Source: Company FY2025 Form 10-K filings and warranty-liability reconciliations filed with the SEC; calculations by the author. Warranty accrual per home calculated as current-year warranty accrual divided by FY2025 homes closed. We omitted Hovnanian Enterprises and Taylor Morrison because their disclosed warranty/construction-defect reserves are not sufficiently comparable.

Change the denominator

Instead of asking only how many dollars each builder reserves, I asked, “How much of the gross profit from a closing does a builder reserve for future warranty costs?”

The story changes.

Toll goes from looking almost worst on warranty dollars per home to among the best when warranty accrual is compared with gross profit. Lennar moves in the other direction. Its roughly $3,035 reserve per home equates to about 4.4% of gross profit, the highest percentage in this comparison.

Neither metric tells us who builds the better house. They answer different questions. Change the denominator, and the story can change.

Where the bet meets reality

Every warranty reserve is ultimately a prediction. Management is saying: Based on what we know today, this should be enough. Fine. Then let’s see whether they were right. These disclosures include adjustments for homes closed in prior periods.

The original reserve is management making the bet. The adjustment is where the bet meets reality.

PulteGroup caught my attention. Its upward adjustments rose from about $4.1 million in 2023 to $6.5 million in 2024 and to $14.3 million in 2025. One year can be noise. Three consecutive years moving in that direction warrant a question.

Toll Brothers reported approximately $10.3 million in higher accruals for homes closed in prior years and increased expected recoveries from insurers and suppliers by approximately $44 million.

M/I Homes disclosed something even more useful: an approximately $11.2 million charge tied to attic-ventilation defects in two Florida communities. A particular problem. A particular geography. That’s what a useful early-warning disclosure looks like.

Then I followed the cash.

Cash payments against warranty liabilities show what companies actually paid for during the year. However, there’s an important limitation: cash paid in 2025 doesn’t perfectly match homes closed in 2025. Some of those checks relate to homes delivered in prior years.

But now we have three windows into the same operating system.

1. A warranty accrual reflects what management expects.
2. A prior-year adjustment indicates whether management changed its mind.
3. Cash payments tell us what the company actually wrote checks for.

Estimate → Revision  → Cash

Together, they tell us considerably more than the reserve alone does. 

How many problems were there?

Then I went looking for the number I assumed would settle the question.

How many warranty claims did each builder have?

Not dollars. Claims. That is where I hit the wall. I looked through 10-K legal proceedings, earnings calls, state contractor and attorney-general information, litigation records, and consumer-complaint data. I couldn’t find it consistently. Not across the public builders. Not under a common definition. Not in a form that allows a credible apples-to-apples comparison. We can find BBB complaints, lawsuits, and individual construction-defect cases. But those aren’t the same thing, and they aren’t consistently reported.

Wall Street can tell you almost exactly how many houses a builder sold. It still cannot tell you how many times the builder had to return to fix one.

The data is not missing

Let’s not confuse undisclosed with unknown. Large builders know how many claims are open, where they occurred, how long they have been open, their cost, and whether the same problems recur.

The data is not missing. It’s inside the company. That is where my question shifted. Instead of asking why investors don’t have the information, the better question is: What information should everyone agree to measure?

Same facts, different scoreboard

Before anyone assumes I am accusing builders of manipulating the numbers, I am not. I suspect something much more ordinary happened. As an Aggie, I understand how that can happen unintentionally. One accounting department developed a reserve methodology. One warranty organization defined a claim. One customer-care department developed its service clock. Another company acquired builders and inherited their systems.

Practices became processes, processes became policies, and policies became culture. Nobody had to be doing anything wrong. But now try comparing 12 companies.

Here’s an Aggie demonstration. In all-time NFL player production, Texas has produced more players than Texas A&M. Fine. FIDO. Forget It. Drive On.

Now, let the Aggie choose the measurement period. In the 2026 NFL Draft, Texas A&M had 10 players selected. Texas had six. Suddenly, A&M wins. And if that doesn’t work, I’ll change the metric. The Fightin’ Texas Aggie Band has never lost a halftime. Absurd? Of course.

But that’s the point.

The facts don’t have to be wrong for the conclusion to change. Change the measurement period, denominator, or unit of analysis, and you can produce a different answer from perfectly legitimate data.

The problem is not necessarily that anyone is hiding the football. Everyone brought a different scoreboard.

Agree on the ruler

The industry does not need another ranking. It needs a common ruler. What constitutes a warranty claim? When does the clock start and stop? What constitutes a repeat or reopened claim? Then measure claims per 1,000 closings and resolution within 7, 14, 30, 60 and 90 days.

Measure at three levels: Company, division and community.

Suppose a builder reports 30 claims per 1,000 closings. What if most divisions operate at 15, while one operates at 150? What if one community accounts for half of that division’s claims?

We have learned something now.

A company-wide problem may point to product design, specifications, purchasing or systems. A division problem may point to leadership, supervision, trades or geography. A community problem may point to a subcontractor, product, plan, soil condition or installation practice.

Corporate size tells us how large the company became. Consistency across divisions and communities tells us whether it actually scaled its operating system.

So, let’s imagine we can build a ruler. We don’t need all 12 builders.

We need two of the 12 to start.

Without asking their permission, I will volunteer M/I Homes and D.R. Horton. Not because I think either has a quality problem. Quite the opposite. They are serious homebuilders operating at different scales. Start with the definitions. Agree on the ruler. Then invite three more builders. Something interesting might happen: They might learn from one another.

Pricing, land strategy, purchasing negotiations, and competitive playbooks stay out. But surely the industry can agree on what a warranty claim is without giving away how to build or sell a house. Better measurement could make everybody better. Competition does not disappear. Competition improves when everybody agrees on where the goal line is.

One catch: The referee should not own a team. That is where an independent third party can help. I have no financial interest in any of these builders. I bring a preferred method, not a preferred winner: Agree on the definitions before seeing the results. Then let the facts fall where they may.

I will volunteer to start the conversation. The builders build the scoreboard. I just want to get them on the same field.

Does scale actually scale?

We are back where our Chasing Scale series started. Think about what an 80,000-home builder has. Thousands of HVAC installations. Roofs. Foundations. Kitchens. And thousands upon thousands of opportunities to learn what fails, why it fails, and how to prevent it from happening again.

In theory, scaling should create a learning machine, not just purchasing power, cheaper capital, or the ability to spread corporate overhead across more closings.

Knowledge.

But experience creates an advantage only if the organization can capture, understand, and apply it. The most important scaling question is not, “How many homes did you build?”

It is, “Does the knowledge travel?”

Does what you learn in Phoenix make Dallas better? Does a problem discovered in one community get caught before it appears in 20 others? Does a recurring warranty issue prompt a change in specifications, installation practices, or purchasing decisions across the company? Do the strongest divisions pull the weakest toward them? Or are 40 divisions effectively operating as 40 different homebuilders under the same ticker symbol?

That’s the difference between growing and scaling.

If warranty claims per 1,000 closings decline, resolution times improve, and repeat claims fall as a builder grows, that’s evidence worth studying. If a problem discovered in one market disappears in the others, that’s evidence too.

That’s what operating at scale should look like. Not perfection. Learning.

Problems will happen. Houses are complicated. They are built outdoors by thousands of people, using thousands of components, across diverse climates, soils, municipalities and trade bases. The question isn’t whether a large builder will ever have a problem.

The question is whether being large makes it better at preventing the next one.

If the largest builders are merely producing proportionally more problems as they build more homes, that isn’t operating leverage. That’s just making mistakes faster. And if scale really does confer a learning advantage, we should see it in the data.

Walk the property

Which brings me back to something I learned a long time ago about land. In Texas, when buying a ranch, you do not stand at the gate admiring the sign. You walk the property.

Ironically, I did not learn that from some old Texas rancher. I learned it as a young man at KB Home from Albert Praw, Executive Vice President of Real Estate and Business Development and, in my mind, the undisputed champion of land. Albert was a mentor, but there was nothing gentle about his teaching. He pushed you hard, questioned everything, and could dismantle a poorly thought-out land deal faster than you could finish presenting it.

You learned quickly: Know your deal, or do not walk into the room.

Albert was also a recovering attorney from Beverly Hills, about as far from the stereotypical Texas landman as possible.

Whenever I brought him a deal, his first question was, “Why are we so lucky to see this deal first?” His second was, “Have you personally walked the property?”

One question challenged the story you were being sold. The second forced you to see the facts for yourself. Turns out Beverly Hills and College Station have more in common than you might think.

You must agree on the nomenclature.

And that’s the lesson here. Closings tell us how big a builder has become. Margins tell us how much money it made. But if we want to know whether scale made the organization better, we must walk the property. The data is there. Let us agree on what to call it, measure it the same way, and see what it tells us. Scale is an input, not an outcome. Do not just show us how big you’ve gotten.

Show us what you have learned.