The Double Trigger: Why Every Multibagger Is Two Bets, Not One

I spent years being right about companies and wrong about returns.

I’d find a business compounding earnings at 20%, buy it, hold it five years, and end up with a stock that had roughly doubled. Which is fine. Which is nothing. Which is what happens when you understand half the machine.

The half I was missing is hiding inside a piece of arithmetic so simple that everyone skips past it:

Price = Earnings × Multiple.

That’s the entire identity. Every stock price on earth is the product of two numbers — what the company earns, and what the market agrees to pay for each unit of those earnings.

Which means a stock can only rise two ways. Earnings grow. Or the multiple expands.

And here is the part that reorganized how I look at everything:

Those two numbers multiply. They do not add.

A company earns $2.00 a share and trades at 10×. Price: $20.

Scenario A — earnings double. EPS $4.00, multiple stays 10×. Price $40. You made 100%. Good year.

Scenario B — the multiple doubles. EPS stays $2.00, multiple goes to 20×. Price $40. You made 100%. Also good. Also — note this — the business did nothing.

Scenario C — both. EPS $4.00 at 20×. Price $80. You made 300%.

Read that again. Doubling one number gets you a double. Doubling both gets you a quadruple.

The gap between a good investment and a career-defining one is almost never “earnings came in a little better than I modeled.”

It’s that the second trigger fired.

Here is the same arithmetic, run backwards on real stocks. The split is the whole point: how much of the return came from the business, and how much came from the market changing its mind.

                     EPS (TTM)          MULTIPLE          ROUGH        WHAT ACTUALLY
                     then → now         then → now       OUTCOME        HAPPENED
──────────────────────────────────────────────────────────────────────────────────────
COMFORT SYSTEMS      $3.46 → $34.70     11.8× → ~54×      ~45×         BOTH TRIGGERS
(FIX)                    ×10.0             ×4.6         $40 → $1,867     FIRED
Jun 2020 → 2026
POWELL INDUSTRIES    $1.30 → $14.98     ~30× → ~21×       ~10×         ONE TRIGGER.
(POWL)                   ×11.5             ×0.7         $40 → $327    Multiple went
FY2022 → FY2025                                                        DOWN. Still won.
UNITEDHEALTH         ~$27 → ~$16        ~24× → ~18×       −50%+        BOTH TRIGGERS,
(UNH)                    ×0.6              ×0.75        $600 → ~$260    IN REVERSE
2024 → 2025
VERTIV               —                  ~69× → ~81×        —          Priced for
(VRT)                                                                  perfection.
today                                                                  No slack left.
──────────────────────────────────────────────────────────────────────────────────────

Figures are TTM, approximate, and drawn from public filings and market data. They are meant to show orders of magnitude, not to be precise to the decimal — which is itself a lesson I’ll come back to.

Four names. Four different lessons. This table is worth more than any ticker I could hand you.

Comfort Systems is a Houston mechanical and electrical contractor. Not glamorous. In mid-2020 it traded at 11.8× earnings, with EPS of $3.46 — priced exactly as you’d price a cyclical construction firm standing on what the market assumed was a peak.

Then two things happened at once. The earnings went up tenfold. And the multiple went up more than fourfold, because the company’s backlog reached $12.45 billion by March 2026, nearly double the prior year, driven by data-center and AI infrastructure demand — and the market stopped filing it under “contractor” and started filing it under “AI infrastructure.”

10× on earnings. 4.6× on the multiple. Multiplied: roughly 45×.

Note what did not happen. Nobody needed to predict a fortyfold move. Two entirely explicable moves — better earnings, a reclassification — multiplied into one that looks insane.

Powell Industries makes electrical switchgear. Same buildout, same demand.

Earnings went from around $1.30 to $14.98 — over eleven times. The stock went from roughly $40 to $327. A superb outcome.

But look at the multiple. It went from about 30× to about 21×. It contracted.

Powell’s earnings did all the work, and the market took some of it back. The market kept classifying Powell as a cyclical — kept discounting the peak — even as the peak refused to arrive. This is a one-trigger stock. It made you rich anyway. It just made you a fraction of what it would have if the second trigger had fired too.

(As I write this, POWL’s TTM P/E has run to the mid-50s. The second trigger may be firing right now, years late. Which is its own lesson: the market gets there eventually. Eventually is not an entry point.)

This is the one to sit with.

UnitedHealth was a compounder. It historically traded around 24× earnings on consistent double-digit growth. Then the medical care ratio blew out, 2025 EPS guidance was cut from roughly $30 to around $16, and the multiple collapsed to about 18×.

Earnings fell around 40%. The multiple fell around 25%.

Neither of those, alone, is a catastrophe. Multiplied, they took the stock from over $600 to the $250–320 range.

The two effects compounded: lower earnings multiplied by a lower multiple, crushing the stock from both sides.

That is the double trigger in reverse, and it is why this framework is not an optimism machine. It’s an identity. It runs both ways with equal force, and the drawdown is always worse than your model — because your model almost certainly flexed the earnings and held the multiple constant.

Vertiv is the same theme, the same demand, an excellent business. It trades around 80× trailing earnings.

Ask the only question that matters: which trigger is still loaded?

The earnings can grow. Fine. But at 80×, the multiple is not a coiled spring — it is a stretched one. You are being paid for exactly one of the two triggers, while carrying the risk of both.

That’s not a short thesis. It’s an asymmetry thesis. And it’s the difference between owning FIX in 2020 and owning it today.

Now go do this on your own portfolio. Take your best holding. Find the EPS and the P/E at your entry, and today. Split the return.

Most people find something they did not want to know: that their “great business” was carried almost entirely by multiple expansion. The business didn’t improve. The market got more excited. That’s a rented return — and rent comes due, because a multiple that expanded on sentiment contracts on sentiment, without a single dollar of earnings changing.

This is structural, not psychological — which is why it persists.

The entire apparatus of professional finance exists to forecast the first number. That is what a sell-side model is: a machine for producing next year’s EPS. Analysts are graded on it. Earnings season exists for it. Thousands of extremely capable people, with better data than you, compete on that number every single day.

You will not beat them there. Neither will I.

The multiple, meanwhile, gets treated as a residual — the cell you paste into the bottom of the DCF and label “terminal multiple,” which is the polite term for the number I needed for my price target to look sensible.

But the multiple is not a residual. It is a verdict: the market’s standing judgment on how durable, how predictable, and how strategically positioned those earnings are.

And verdicts get overturned. Comfort Systems didn’t go from 11.8× to 54× because it got better at installing ductwork. It went there because the market changed its mind about what kind of company it was looking at.

Earnings tell you what a company is doing. The multiple tells you what the market believes it is. A multibagger is what happens when both beliefs are wrong at once — and both get corrected.

There is no army modeling why the multiple should be different. There is no consensus estimate for “what will this company be classified as in three years.”

That is not a gap in coverage. That is an unoccupied position.

The multiple is a verdict. So to forecast it, you stop asking what will earnings be — the crowded question — and start asking: why is this thing valued the way it is, and what would break that?

A stock trades at 8× for a reason. Usually a good one. In my working grid there are six such reasons, and each has an expiry date.

Here is one of them, entire. Not a teaser — the condition, the release event, and the tell. Run it this week. If it works, you’ll know what the other five are worth.

The compression. A market will not pay a high multiple for earnings it cannot predict. Project-based businesses, contractors, capital-equipment makers — anything where revenue arrives in unpredictable slabs — get filed under cyclical, and cyclicals get discounted for the peak they’re assumed to be standing on. It doesn’t matter how good the current year is. The market assumes it’s the top.

This is precisely why Comfort Systems traded at 11.8× in 2020, and why Powell traded at 21× while its earnings were compounding elevenfold.

The release event. Revenue quietly stops being lumpy — and becomes contracted, recurring, or backlog-covered.Multi-year framework agreements. A book-to-bill sustained above 1 for several consecutive quarters. Revenue visibility extending well beyond the next twelve months. The earnings haven’t become larger. They’ve become predictable — and predictability is exactly what a multiple pays for.

The tell — and this is the part that matters. The rerating does not happen when the backlog is announced. It happens two to four quarters later, once the backlog has been converted — once the market has watched the company deliver against it and concluded the visibility is real rather than promotional.

The market does not reprice a promise. It reprices a demonstrated promise.

Which means the entry window is not the announcement. The entry window is the boredom that follows it — the two or three quarters where the news is priced in, the story has gone quiet, and the conversion hasn’t yet been proven.

That is the whole condition. Go look for it.

I’ve just handed you a mechanism that makes 45× look like arithmetic. So let me be direct about what I have not handed you, because the gap between the two is where people lose money.

The identity runs backwards. UNH is the proof. Same machine, same compounding, reversed — and it took over 50% off a stock that thousands of people held precisely because it was safe. That is not a footnote. That is half the machine, and it is the half that gets skipped.

“Cheap and it should be less cheap” is not a thesis. It’s a hope. Any screen on earth can find you a low P/E with growing earnings — which is why any screen on earth is worthless. A low multiple is almost never held down by onething. It’s held down by a stack. Release the top condition and the stock does not move, because the one underneath is still load-bearing.

Powell is the visible version of this: earnings up eleven-fold, and the multiple fell. The condition holding it down was never released — it was simply overwhelmed. That worked out. Usually it doesn’t.

I know this because I bought that exact setup — a stock where I had correctly identified a compression and correctly predicted its release, and where the multiple didn’t move at all. And then, having diagnosed the loss as bad luck rather than bad method, I bought it again.

That second purchase is the most expensive tuition I have ever paid. It produced the single most useful filter I own, and it is now the first thing I run on any candidate — before the earnings, before the story, before the price.

It’s on the other side of this line, with the rest of the working system.

The free half gave you the identity, the decomposition, and one of the six conditions. Here is the system.

The Rerating Map — all 6 conditions. Each with: the compression, the specific release event, the tell that separates a real release from a false one, and — critically — the observed lag between event and rerating. It is never the announcement date. Knowing the lag is the difference between a 40% entry and a 4,000% one.

The Binding-Condition Test. The filter that cost me two positions to learn. How to determine, before you buy, whether the compression you’ve found is the one actually holding the multiple down — or whether there are two more underneath it, still load-bearing. This kills roughly 70% of what looks like a coiled spring. It is the reason I no longer own value traps.

The two positions that taught it to me. Named, with entry, exit, the loss, and the exact reasoning error — walked through slowly, because the error is subtle and it is the one you are most likely to repeat. Most letters show you their winners. This one shows you the invoice.

The Decomposition Grid — a working file, not a concept. A downloadable table that splits any return into its two components, with the four methodological traps already handled: trailing vs forward P/E, near-zero-EPS years that blow the ratio to infinity (Powell’s TTM P/E once printed above 1,300× — a number that means nothing and would ruin any naive screen), non-recurring charges, and share-count changes. Run it on your own portfolio tonight.

The Reclassification Test. How to tell whether a company is genuinely changing categories — as FIX did — or merely having a good year, as Powell was assumed to be having for three straight years. There is a concrete way to test this before the market notices. It is the single highest-value item in the grid.

The downside arithmetic, in full. What the double trigger does in reverse, why the loss is always worse than your model says, and the position-sizing rule that follows. UNH is the case study, walked through step by step.

Where the triggers are loaded right now. Not tickers as gifts. A walk through the specific places in this market where both conditions are simultaneously present — and three places where they look present and are an illusion. Vertiv is the free preview of that logic. The other three are not free.