Skill, Portfolio, or Physics

Howdy Folks,

Last week I ended the relaunch issue with a promise. Two national final-mile providers, same category, same year, growth rates twenty points apart:

• J.B. Hunt Final Mile Services: $824 million, down 9.5%
• Ryder E-commerce and Last Mile: $983 million, up 9.1%

Those are 2025 full-year gross revenue figures, reported by Eric Kulisch at FreightWaves on July 23, 2026, sourced to Armstrong & Associates. I said I had opinions about which of those numbers was skill and which was portfolio.

I did. Then I went and pulled the actual quarterly filings, and the honest answer is that I was asking too small a question.

Here's the thesis, up front: the spread between those two companies is mostly portfolio. The thing worth your attention is what happened underneath both of them. In Q2 2026, every national big-and-bulky provider raised the price of a stop, and every one of them lost margin doing it. That is not a competitive story. That is a cost curve moving faster than anyone's pricing power.

Let me show you.

1. The quarter, from the filings

Three providers, one quarter, primary sources only. J.B. Hunt and RXO both break out final mile. Ryder does not, so its last-mile business sits inside Supply Chain Solutions along with automotive and everything else, which is a caveat I will come back to.

Sources: J.B. Hunt Q2 2026 earnings release, July 15, 2026 (segment tables). RXO 2Q 2026 press release, August 6, 2026 (revenue disaggregated by service offering, and the gross margin reconciliation). Ryder Q2 2026 results, released July 2026 (SCS segment table).

Look down the last column. Every single one.

One company shrank on purpose and lost margin. One company grew share and lost margin. One company grew revenue and lost margin. If execution were the variable, the three of them would not be moving in the same direction.

2. Everybody raised the price of a stop

This is the part that surprised me, and it is the reason the growth-rate comparison is a trap.

J.B. Hunt's stops fell 13.7% in Q2. Revenue fell only 6%. Both of those numbers are in the same filing, which means you can divide one by the other and get the number the company did not print:

Revenue and stops are printed in J.B. Hunt's Q2 2025 and Q2 2026 earnings releases. Revenue per stop and income per stop are my arithmetic on those figures, not company-reported metrics.

Revenue per stop is up 9.0% year over year and up 7.3% over two years. That is the "improve revenue quality" initiative working exactly as designed. They fired accounts, they repriced the ones they kept, and the price of a J.B. Hunt stop went up.

Operating income per stop went from $18.00 to $6.45. Down 64% in two years.

They charged nine percent more per stop and kept a third as much of it.

RXO ran the opposite play and landed in the same place. Last Mile revenue grew 9.2% to $344 million on stop growth of 3%, which is roughly 6% more revenue per stop. Then look at the gross margin reconciliation in the same press release: complementary services revenue rose from $457M to $488M, and complementary services gross margin dollars went from $104M to $103M.

Thirty-one million dollars of additional revenue produced one million dollars less gross margin.

That is not a company failing to execute. RXO gained share in a down market and got FreightWaves and CargoNet awards for carrier vetting in the same quarter. They did the job. The job just costs more than it used to.

3. The number nobody puts on a slide

So where is the money going? It is not fuel, and it is not wage inflation showing up all at once. It is density, and there is exactly one line in these filings that shows it.

J.B. Hunt reports average trucks in the Final Mile segment. Divide stops by trucks and you get productivity per truck per quarter:

Down 5.2% in a year. Down 10.1% in two.

And notice they were not asleep at the wheel. They pulled 175 trucks out of the fleet over two years, a 12.7% reduction, actively chasing the volume down. Stops still fell faster than trucks did. Density got worse anyway.

This is the whole ballgame in residential big and bulky, and it is the thing that separates our category from parcel. A parcel network absorbs a demand drop by putting fewer packages on the same truck. The truck was going down that street regardless. Our trucks are not going down that street regardless. A two-person crew, a fourteen-foot box, an appointment window, and a sectional that has to go up a staircase: that route only works if the next stop is close to the last one. Take 14% of the stops out of a market and you do not get a smaller version of the same route. You get the same route with more windshield time between the money.

Cost per stop in this business is a density function. Density is a function of how many people in a given ZIP code bought a couch this month. And that number is set by the housing market, which nobody in this industry controls.

4. Which is why the housing number matters more than the market-share number

The demand side has not recovered, and it is not about to.

Existing-home sales ran at a seasonally adjusted annual rate of 4.06 million in July 2026, down 1.7% from June, per the National Association of Realtors release on August 11. That is up a whole 0.7% from a year ago. The median existing-home price hit $434,100, the 37th straight month of year-over-year increases.

NAR chief economist Lawrence Yun described this as home sales being "remarkably stable." He is not wrong. That is the problem. Stable is the worst possible outcome for a category whose unit economics need turnover, and stable is what we have had for three years.

Expensive houses that nobody is trading. Armstrong & Associates put the turnover rate at 28 homes per 1,000 in 2025, against 44 per 1,000 in both 2021 and 2019. That is a 38% collapse in the single event that most reliably causes a human being to buy a sofa, a mattress, a washer, and a grill in the same ninety days.

Armstrong sizes big and bulky final mile at $10.6 billion, growing to $12.3 billion by 2027. That is 5.1% annually, against 10.6% over the prior eight years. Category gross margins slid from 28.9% in 2022 to 27.5% in 2025.

So the category is growing at half its historical rate, margins are compressing, and route density is deteriorating faster than providers can pull trucks out. Against that, a nine-point spread between two providers' revenue growth is noise.

5. So: skill or portfolio?

Here is my actual answer, now that I have looked.

Portfolio explains the top line. J.B. Hunt's decline is substantially self-inflicted and they say so in plain language: revenue fell on "known business losses given our ongoing efforts to improve revenue quality and profitability across various accounts." You cannot call that losing to Ryder. It is a company walking away from freight it decided it did not want, and the revenue-per-stop number says the walking worked.

Ryder's growth has a portfolio story too, in the other direction. Their last mile lives inside Supply Chain Solutions, and that segment's earnings fell 7% on lower automotive results, offset partly by what Ryder calls "the optimization of omnichannel retail network." Automotive dragged them down, retail network work held them up. Neither of those is a statement about how well they deliver a treadmill.

Skill explains almost none of the spread, and everything about who survives it. Because the margin line is identical across all three, and margin is where execution actually shows up. RXO grew stops 3% in a market where flat was the plan, and still gave back 170 basis points. That tells you the cost curve beat a genuinely good operator.

Physics explains the margin. Density fell, cost per stop rose, and pricing caught maybe half of it. Everything else is commentary.

Which brings me to the uncomfortable part for anyone selling in this category. When density is the binding constraint, the winner is not whoever has the best app or the nicest crews. It is whoever has the most stops per square mile in a given market. That is a structural advantage, it compounds, and it is very hard to buy your way into.

I wrote in Gray Glove that the service-level menu in this category is mostly a pricing fiction dressed up as a product. Two years later the filings agree: when you can charge 9% more per stop and still lose two thirds of the profit on it, your problem was never the service tier.

6. What I would actually do with this

Not advice, just what I would be arguing about in a Q4 planning meeting.

• Stop benchmarking cost per stop nationally. It is a meaningless average across markets with wildly different density. Benchmark it by market, against your own stop count in that market, and watch the direction more than the level.
• If you are a shipper, your rate is a density bet your provider made about you. When your volume in a metro drops, the cost of serving you goes up whether or not your rate does. The provider absorbing that quietly is the one who repricing you next year. Ask which of your markets you are underwater in before they do.
• Injection and shared routing stop being optional. At 719 stops per truck per quarter, roughly eleven a day, the only lever left that does not require more demand is putting somebody else's freight on your truck.
• Watch stops per truck, not revenue growth. It is the one metric in this whole category that cannot be flattered by mix, price, or an acquisition. J.B. Hunt is the only one of the three that publishes enough to calculate it, which is either an accident of disclosure or the most useful thing in their release.

I have been beating the curbside drum for a while now, and this is the version of that argument with a P&L attached. In a 10% growth market, bad density is a rounding error. In a 5% market it is the whole margin.

7. I could be wrong because...

Standing feature. Here is where this take falls apart.

• The Q2 number is doing a lot of work, and the half-year number disagrees with it. J.B. Hunt's Final Mile operating income for the first six months of 2026 was $12.72M against $12.67M in the same period last year. That is up four tenths of a percent. I built a decline narrative on a quarter that the half-year flattens out. If Q3 comes in flat, my cost-curve story is really a tough-comp story.
• Ryder does not belong in this table and I put it there anyway. Supply Chain Solutions is mostly not last mile. Attributing that segment's margin move to big-and-bulky density when the company explicitly blames automotive is me fitting the data to the thesis. I flagged it, but flagging it does not fix it.
• Revenue per stop is a mix number, not a price number. If J.B. Hunt shed cheap curbside accounts and kept the white-glove ones, revenue per stop rises with no one raising a single rate. That is a completely innocent explanation for my most-cited figure, and I cannot rule it out from the filings.
• Three companies is not an industry. The national providers are a minority of this market. Most big-and-bulky delivery in America is done by regional carriers and agent networks that publish nothing, and their density math may look nothing like this.
• I sell in this category. "Density is destiny and the incumbents cannot buy their way out of it" is a very convenient thesis for a regional operator. Discount accordingly, same as last week.

What's next

Two weeks of this and I owe you the other side of the rotation. Next week is AI at Rocket, the pillar where my own company is the subject instead of the example: what we actually run, what it costs per month, and the two things we built and then threw in the trash. Including the one that worked perfectly and that we still killed.

See ya next week.

Gabe

If you run routes in this category: is your stops-per-truck number down over two years, or is J.B. Hunt's problem specific to their book? Reply and tell me. I will run the best answers in a future edition, credited or anonymous, your call.