Valuing a company that is shrinking on purpose

UPS guided 2026 revenue to roughly $89.7 billion — about flat — while planning to shed another million Amazon pieces a day. Revenue had already fallen from a $100.3 billion peak in 2022 to $88.7 billion in 2025, and adjusted operating margin had compressed from 13.8% to 9.8%. Amazon alone was 10.6% of 2025 revenue, almost all of it in U.S. Domestic.

The standard way to value a company like this is to pick a growth rate and a margin, discount the cash flows, and read off a price. For UPS in 2026, that produces a clean answer to the wrong question.

The wrong question: how fast will UPS grow?

If you index UPS to parcel demand, the story reads as decline. But 2025 volume fell mostly because UPS chose to exit low-yield business, not because demand disappeared. The first quarter of 2026 made the trade visible: U.S. Domestic average daily volume fell 7.4% while revenue per piece rose 6.5%.

So the real question is narrower and harder: can UPS replace departing Amazon density with higher-yield volume faster than its fixed costs deleverage? Everything in the valuation follows from how you answer that.

Model the decision, not the index

I rebuilt U.S. Domestic revenue from its mechanics instead of a growth line:

• Revenue = operating days × average daily volume × revenue per piece.
• Volume = starting volume − the Amazon glide-down + SMB, B2B, and enterprise replacement + a macro demand effect.
• Revenue per piece = base rate increases + mix + package characteristics + fuel surcharge + premium product share.

That structure makes the core trade explicit: Amazon volume at roughly $8–9 a piece leaving, SMB, B2B, and healthcare volume at $12–16 a piece arriving. It also forces honesty about segments. Supply Chain Solutions, for example, can’t be forecast from its own history — a divestiture and a string of healthcare acquisitions sit inside it — so I split healthcare logistics from the legacy business.

The cost-out is a floor, not a ceiling

UPS delivered about $3.5 billion of savings in 2025 and committed to another $3.0 billion in 2026. It closed 93 U.S. facilities in 2025, took out roughly 27 million labor hours, and announced up to 30,000 operational position reductions and 24 more closures for 2026. Yet U.S. Domestic adjusted cost per piece still rose to $13.35 in the first quarter of 2026, from $12.19 a year earlier — volume was leaving faster than the network could shrink.

That is why I built margin from a cost bridge up rather than assuming a margin line down. Switch the drivers off below and see what each one is carrying.

The cost program mostly offsets what the Amazon exit and labor take away. Margin only expands — to about 10% by 2030 — if premium mix and automation land as well. Cost-out is table stakes; the upside has to be earned elsewhere.

I estimated each driver independently, and together they overshoot the guided path by about 120 basis points. Rather than bury that, the bridge shows it as an explicit execution-and-timing haircut. Switch it off and you get roughly the margin of the upside scenario — which is exactly the point: the upside case is simply the one where every driver delivers.

Four scenarios, one judgment call

A single DCF would hide the thing that matters most: execution. So the model runs four scenarios, each with its own margin path and its own cost of capital, re-levered by scenario using Hamada. WACC runs from 7.6% in the upside case to 9.4% in the weak-replacement case, because poor execution widens credit spreads and raises the equity risk premium at the same time — a double hit. In the sensitivity analysis, a one-point move in WACC swings value by about $28 a share, more than any operating driver.

Here are the four cases. The probabilities are mine; move them and see where the value lands.

With my weights, UPS is worth about $102 a share, against a market price near $111 at the time. The gap is small enough to call UPS roughly fairly valued on the guided path. The more interesting output is what the market price implies: holding my view of the weaker cases constant, ~$111 needs the automation-and-quality case to carry close to 30% probability — about double my 15%. That is the bet embedded in the stock.

What I would watch

Three indicators decide whether the thesis holds in 2026:

1. Replacement volume — SMB, B2B, and healthcare volume against the Amazon loss. The upside case needs at least 70% replacement by year-end.
2. U.S. Domestic revenue per piece — up 6.5% in the first quarter; it needs to stay above 3%.
3. Cost-out realized — about $600 million in the first quarter against a $3.0 billion target; a $2.4 billion run-rate by year-end is the floor.

If replacement falls below 50%, revenue per piece growth drops under 3%, or cost-out misses $2.4 billion, the weak-replacement case becomes the base case.

What building it taught me

A model should show management choices, not spreadsheet outputs. The useful version of this model isn’t the one with the most tabs; it’s the one where a reader can see which decisions move value and by how much.

Every assumption needs a home and a source. The final build has 26 tabs and 22 integrity checks, and the major assumptions are tied to named sources — company filings, earnings materials, and market data. That discipline is what lets a range be defended rather than just asserted.

The range is the answer. My conviction is high on direction — the quality-over-volume strategy is the right one — and medium on magnitude, because the upside depends on execution in the second half of 2026 that no model can promise.

Outside-in analysis using public data only (UPS 10-K and 2025–2026 earnings materials, SEC EDGAR, FRED, Damodaran). Not investment advice.