Four flat years, and twenty-six percent on capital

On the close of Friday 11 September 2026, A. O. Smith traded at $57.39. Sperio values a business from its filed accounts and a discount rate, giving no credit of any kind to future growth, and arrives at $122.77.

The market is paying 47 cents for each dollar the accounts support without growth. That is the fifteenth widest gap in the index this week, and it is the one the Sunday ranking put first. As last week, explaining why is most of the article — and this week the explanation includes a change to the ranking itself.

Why this one, and not the widest discount on the list

The widest discount in the index this week belongs, again, to a fertiliser company: net income down 85%, and a verdict that flipped to SELL in late August. Behind it sit a laboratory landlord and a building-products distributor whose net income has fallen 84%. Ranking by discount answers "what looks cheap", and it answers it reliably with whichever company is deteriorating fastest, because a collapsing denominator makes any price look like a bargain.

So the ranking applies two hard gates first. Does the business durably earn more on its capital than the capital costs — return on invested capital clearing ten percent in four fifths of the readable fiscal years and in each of the last three. And are both lines going the right way — revenue and operating income each compounding upward over the full readable history. Of the 456 companies the board has valued, 105 clear both.

Last week's piece said the survivors were then ordered by an expected rate of return. Two days after it was published, that ordering was retired, and the reason belongs in print. The expected-return ordering cannot be tested against the past: its probability term is the board's conviction score, a model output, and replaying it would score today's prompts against a history they were written knowing. Ordering by cheapness instead — the full-cycle operating multiple, EV to the median operating profit of up to fifteen years — can be tested, and was, on the two-line screen alone. Over sixteen half-yearly rebalances from the end of 2017, the cheapest name that cleared both lines returned a median of +22.8% at one year and beat its own universe 62% of the time. That is about eight independent observations, on a universe where the companies that disappeared are only partly accounted for: an upper bound and a small sample, not a promise. It is also the only ordering in this project with any evidence behind it, and last week's article already said which instrument the evidence favoured. It is now the one in use.

Two things sit on top of it that have not been measured: the franchise gate, and the flags that block a name from being published. This week the measured screen's own first name is a homebuilder at 9.2 times that the board rates SELL. The gate and the flags exist for exactly that disagreement, and they are the part of the recipe without evidence. What has been tested is the direction of the ordering, not the whole machine.

On that ordering A. O. Smith is the seventh cheapest of the 105 survivors, at 15.4 times full-cycle EV/EBIT. Each of the six names ahead of it carries a flag that blocks publication: four are companies the board rates HOLD, one has net income down 3% with earnings per share held up by buybacks, and one has net income down 52%. A. O. Smith is the cheapest survivor the board also argues for — a BUY, at a conviction of 50, which is the board's middle of the road and is reported as such.

What the twenty-six percent is

Return on invested capital, across the readable record: about 26%. Return on equity in the trailing twelve months is 27.1%, gross margin 38.6%, debt 0.37 times equity. And the line that matters most for a valuation built on cash: in fiscal 2025 the company reported $546.2m of net income and $546.0m of free cash flow. What it earned, it collected.

The accounts, five fiscal years, as filed:

Read across the last four columns and the picture is a straight line. Revenue between $3.75bn and $3.85bn. Operating income between $659m and $757m, on a margin between 17.6% and 19.7%. Net income within twenty million dollars of $550m in three of the four years. A water heater fails and is replaced whether or not the economy is growing, and the company itself describes most of its North American volume as replacement demand; the consolidated accounts look exactly like that description.

Over the four fiscal years the ranking's window covers, the share count fell 9%. Measured from fiscal 2021, before the year that does not fit, earnings per share are up 27%, net income up 12%, and the share count down about 12%. More than half of the per-share growth is arithmetic. Last week's subject was the opposite case — a treasury that did almost nothing, so that what reached the shareholder was what the business earned. Here the business earned the same amount and the treasury did the rest. Neither is a fault. They are different things, and a reader is owed the distinction.

The number that does not fit

Fiscal 2022. Operating income was $659m; net income was $236m. Most of the gap between them is a single item below the operating line — the settlement charge from terminating the company's pension plan — and it has not recurred.

The ranking's four-year window starts on that year. Read naively, its output says earnings per share rose 154% over the window, net income 132%. Both numbers are in this week's run. Neither is what happened. Measured from the year before the charge, earnings per share are up 27%; measured from the year after it, they are up 4%. A one-year comparison would have printed the same mirage on this company that it printed on last month's health insurer, whose 83% earnings rebound was a recovery from a collapsed year and not a progression. This is the second time in a month that the screen's own arithmetic has produced a headline number that had to be taken away from it, and publishing that is part of the point: the screen is a filter, not a thesis. It removes traps. It does not manufacture the proof of the opposite.

The downside, priced

A single valuation cannot be argued with. Three can.

The first line is the one that decides how much of a position a professional would take, and it does the thing a trough case is not supposed to do: it sits below the price. At $45.59 against the $57.39 the market is paying, a buyer who is wrong about everything except the trough is down about a fifth, not protected. That is disclosed on the row rather than filtered out of it. It is the only flag this name carries, and it is the reason the credited figure on the third line should be read as the frame's arithmetic and not as a target.

The rate sweep says the same thing from the other side. At today's cost of capital — 9.95%, built on the 4.95% ten-year Treasury as FRED reported it on 11 September — 105 companies clear both gates and five of them are worth more than their price even in the trough case. Raise the cost of capital by a hundred basis points and 90 still clear the gates but one keeps that cushion; at two hundred, none. A. O. Smith is not among the five today.

What the price is saying, then, is not that the business is worth nothing without growth. It is that the flat line will bend downward. There is some support for that in the most recent figures: the 16.0 trailing multiple on the company's own page implies about $3.59 of earnings per share over the last twelve months, some 7% below fiscal 2025's $3.85, and the shares sit 28% below their 52-week high and under their 200-day average. The market has been selling this company steadily, not in one shock. The accounts to fiscal 2025 show four flat years; the trailing twelve months show the first small dip. The whole question is which of those two is the trend.

What the list looks like around it

The twenty cheapest survivors this week include two home-improvement retailers, two railroads and two defence contractors. A screen ranks companies one at a time and has no idea that two of its answers are the same bet. Anyone reading the list as twenty independent ideas is reading it wrong, and a water-heater maker that sells through one of the first pair is not independent of them either.

The case against, in the board's own words

Sperio runs a devil's advocate whose only job is to attack the thesis. It is reproduced without softening, because a recommendation that publishes only its bull case is one that cannot be retracted honestly later:

The company faces cyclical headwinds from slowing GDP growth and reliance on construction markets. Furthermore, a pivot toward inorganic growth and recent impairment charges signal potential erosion of capital allocation discipline.

Both halves land, and the second one lands harder than it looks. A business whose own sales have been flat for four years and which starts buying other businesses is doing the one thing a flat, high-return business should be most careful about: the 26% is earned on capital already in the ground, and it says nothing about the return on capital paid out for someone else's. The impairment charges are the accounts' way of reporting that some of that money has already been marked down. They sit below the operating line, where fiscal 2022's charge sat — but that one came from closing a pension plan, and these come from a choice that can be repeated.

The accounts add their own case, and it is the one this piece takes most seriously. A flat business at a 26% return on capital is worth roughly twice its price on a no-growth valuation, and roughly four fifths of its price at the trough. The distance between those two numbers is the distance between "this line stays flat" and "this line is the top of a cycle", and nothing in the filed figures can settle that. The trailing twelve months lean the wrong way. The buyback flatters the per-share line. And the widest number in our own output for this company is one we had to correct before printing.

Why now: nothing has happened

There is no catalyst. The verdict has not moved, the accounts have not changed, and the gap has been open for a while. Saying so is more honest than manufacturing a reason.

One date is worth watching. The next results are scheduled for 27 October 2026, with consensus at $0.90 a share. For context and not as a prediction: the last four quarters came in against consensus at +3%, +7%, −10% and +12%. That is not a company clearing its guidance by a rounding error every quarter; it is one whose quarters land ten percent either side of what the analysts expected, in both directions. The number to read on that day is not the beat or the miss. It is whether four quarters at that pace still add up to something near $3.60 — the first small dip — or whether the flat line the annual accounts have shown for four years has started to bend.

Last week's pick, scored

Last Sunday's piece said that this one would report whether Paychex still clears the gates, and score it out loud.

It clears them: same accounts, revenue up 17%, operating income up 14% in the fiscal year just filed. Its shares closed Friday at $115.80 against the $121.71 of the previous Friday, down 4.9% in a week — a number that says nothing about the thesis and is reported because it was promised. What has changed is its place in the ranking. Under the ordering retired on 8 September it was first. Under the one that replaced it, at 33 times full-cycle EV/EBIT, it is not among the twenty names the run prints. Last week's article already said that the only measured instrument in this project ranked it ninety-ninth on the cheap-and-growing screen, and that the honest thing was to say which instrument had evidence behind it. The ranking now agrees with the evidence. Paychex's numbers did not move; the instrument did.

What this is, and is not

Every figure here comes from accounts filed with the SEC, from a board valuation dated 6 September 2026, and from a ranking run dated 12 September 2026. The price used is the close of Friday 11 September 2026. The cost of capital is built from FRED observations as of 11 September 2026. The valuation gives no credit to growth beyond what the franchise record earns, and the ranking that produced this name orders the survivors of two gates by their full-cycle operating multiple — a way of comparing them to each other, tested on a small sample, not a forecast of what any of them will do.

The board that produced those figures will run on any company you name. What it will not do is decide for you — nothing here is investment advice.