United Parcel Service UPS

Reference table

1. What the business is, and what it is deliberately giving up

United Parcel Service operates through U.S. Domestic Package, International Package and Supply Chain Solutions, with approximately 460,000 employees, chief executive Carol Tomé, and $41,994M of net property, plant and equipment on a $73,090M balance sheet.

The strategic fact that governs everything else is in the 10-K, stated without euphemism:

> "In 2025, one customer, Amazon.com, Inc. and its affiliates, represented approximately 10.6% of our consolidated revenues, substantially all of which was within our U.S. Domestic Package segment. As previously disclosed, our strategy involves reducing volumes from this customer by more than 50% by June 2026 from 2024 levels."

Amazon was 11.8% of revenue in each of 2023 and 2024 and 10.6% in 2025, and 13.6% of accounts receivable. The reduction completes in the month before the quarter just reported, which makes the September 2026 print the first clean look at the post-Amazon revenue base.

Three other structural changes, all from the filings:

• The USPS arrangement. "In December 2025, we entered into an agreement with the United States Postal Service ('USPS') to support final-mile delivery for a portion of our Ground Saver and Mail Innovations volumes starting in 2026." In Q1 2026 this raised facility and transportation costs by $341M through "higher fees paid to the USPS" while lowering direct labour by $423M — and it is the reason cost per piece rose 9.7% in the quarter, because transition costs and "excess operational staffing" ran alongside the handoff.
• Healthcare. UPS completed the acquisitions of Frigo-Trans and Biotech & Pharma Logistics and Andlauer Healthcare Group ("AHG") in 2025, both reported within SCS. "In 2025, our global healthcare portfolio generated more than $11 billion in revenue." In Q1 2026, healthcare logistics revenue rose $188 million year on year "primarily due to our fourth quarter 2025 acquisition of AHG." Vendor cf_a records −$1,968M of acquisitionsNet in FY2025, consistent.
• Network Reconfiguration and Efficiency Reimagined. "In the first quarter of 2026, we closed 23 leased and owned buildings, 22 of which have been permanently closed as of March 31, 2026. We have identified 27 additional buildings for closure in 2026... In the first three months of 2026, we achieved approximately $600 million of program cost savings, and expect to achieve approximately $3 billion in full year-over-year cost savings from this initiative in 2026." Cost to achieve: "between $1.3 and $1.5 billion in cost during 2026, primarily related to employee separation benefits and third-party consulting fees, of which $1.2 billion will be related to the Driver Choice Program." The programmes "are expected to conclude by 2027."

Note what that last disclosure does. It was filed on 2026-05-06 and it pre-announces the exact charge that produced the June quarter's GAAP/adjusted divergence. A $1.2 billion pre-tax Driver Choice charge is roughly $0.95 billion after tax at the trailing 22.07% effective rate, or approximately $1.12 per share on 851 million shares — against an observed GAAP-to-adjusted gap of $1.05 per share in the quarter. The arithmetic ties. This is the rare case where a large adjustment is fully explained by a prior filing, and it should be credited as such.

2. The June 2026 quarter — read both lines

The Q2 2026 10-Q is not in this archive (the most recent quarterly filing is the March 2026 10-Q, filed 2026-05-06). Quarterly figures below therefore come from the vendor's inc_q and earn_cal blocks, cross-checked against each other, and are labelled as such. The 8-K of 2026-07-28 confirms the release date and that Exhibits 99.1 and 99.2 carry the results, but the exhibit text was not extracted.

Three readings, and they do not agree with each other, which is the point.

On the revenue line the inflection is real. $22,834M is the highest quarterly revenue since Q4 2024 and the first year-on-year growth in the series. It is not organic parcel volume — AHG contributed, revenue per piece rose (International +10.7% in Q1 on "revenue quality efforts"), and the USPS handoff keeps the customer billing while moving the delivery cost. But growth of 7.6% in the quarter in which Amazon volume is supposed to fall by half from 2024 levels is a materially better outcome than a shrinking-network narrative would predict.

On the GAAP line the quarter is the worst in the series. Operating income of $930M is a 49% decline and a 4.07% margin, against 8.58% a year earlier. On a full-year basis operating income has run $12,968M (2022), $9,371M (2023), $8,685M (2024), $7,867M (2025) — a 39% decline on an 11% revenue decline, which is textbook operating deleverage in a fixed-cost network.

On the adjusted line the quarter is a beat. $1.76 against $1.65, +6.7%, and up 13.5% year on year against Q2 2025's $1.55. Reconstructing the adjusted operating profit from the EPS gap — $1.05 per share × 851M shares = $894M after tax, ≈ $1,146M pre-tax at the 22.07% trailing rate — implies adjusted operating income of roughly $2,076M and an adjusted margin near 9.1%, against roughly 8.7% a year earlier. (Our arithmetic, not a company disclosure; the Q2 release exhibit is not in this archive.) So on the company's preferred basis the margin improved while on the reported basis it halved. Both are true; the difference is $1.2 billion of separation payments the company told you about in May.

The four-quarter beat pattern: $1.55 vs $1.56 (−0.6%), $1.74 vs $1.29 (+34.9%), $2.38 vs $2.20 (+8.2%), $1.07 vs $1.04 (+2.9%), $1.76 vs $1.65 (+6.7%). Four beats and one in-line in five quarters — consensus is running slightly behind the cost programme, which is the normal pattern when savings arrive faster than analysts model them.

3. Free cash flow and the dividend — the check that mattered, and it came back clean

This dive was directed to treat capex and free cash flow as guilty until proven innocent, because those two fields have been corrupt on nine capital-intensive names in this programme, always in the direction that flatters the yield. UPS is clean. We state the verification in full because a verified-clean result is a finding.

The 10-K, verbatim: "In addition to capital expenditures of $3.7 and $3.9 billion for 2025 and 2024, respectively, there were principal repayments of finance lease obligations of $133 and $136 million in these years." The vendor's $3,685M and $3,909M round to exactly those figures. Free cash flow is subtracted, not added. capexToDepreciationTTM is 0.8827 — below 1.0, which is unusual but corroborated by the company's own guidance of "approximately $3.0 billion for all of 2026" against $3,746M of D&A. UPS is genuinely under-investing relative to depreciation, on purpose, and the vendor reports it correctly.

Which makes the dividend arithmetic trustworthy, and it is the most important number on this page.

And the corroborating signal, in the company's own words: "In January 2023, the Board approved a share repurchase authorization for up to $5.0 billion... During 2025, we repurchased 8.6 million shares of class B common stock for $1.0 billion under this authorization. As of December 31, 2025, we had $1.3 billion available under our share repurchase authorization. We do not anticipate further share repurchases in 2026."

A company with $1.3 billion of authorisation remaining that tells shareholders in advance it will not use it is telling you where the cash is going. In 2025 UPS "returned $6.4 billion in cash to shareholders by completing $1.0 billion of share repurchases and paying $5.4 billion in dividends" against $4,765M of free cash flow. The gap was funded from the balance sheet: total debt rose from $25,652M to $32,290M across 2025, and net debt from $19,540M to $26,403M — a $6.9 billion increase in net debt in a year in which $6.4 billion was returned and $1,968M was spent on acquisitions.

This is not a solvency question. Net debt/EBITDA of 2.27x and interest coverage of 4.04x are comfortable. It is a question about whether the current distribution survives contact with the next two years, and the honest answer is that it depends entirely on the $3 billion cost programme landing. If FY2026 operating cash flow holds near $8.5 billion and capex falls to the guided $3.0 billion, free cash flow is roughly $5.5 billion against a $5.6 billion dividend — still short, but within a rounding error, and improving. That is the base case and it is tight enough that we will not underwrite an increase in the payout.

4. Balance sheet

The debt step is the item to notice: total debt +$6,638M and net debt +$6,863M in a single year, against $1,968M of acquisitions and $6.4 billion of shareholder returns. The 10-Q notes interest expense rose "primarily due to higher average outstanding debt balances." The 10-K reports $5.9 billion of cash, cash equivalents and marketable securities at year end and $5.8 billion at 2026-03-31, of which approximately $2.0 billion sits at foreign subsidiaries.

A lease double-count test was run and is inconclusive on this archive. totalDebt of $32,290M equals shortTermDebt $1,371M + longTermDebt $27,219M + capitalLeaseObligations $3,700M exactly. On T, TJX, GD and TT this pattern indicated finance leases counted twice — once inside long-term debt and once again as a separate line. The UPS 10-K's balance sheet tables were stripped in extraction (tables: false in the filings manifest), so the reported long-term debt and finance-lease figures cannot be read directly and the test cannot be closed either way. We say so rather than assuming. If the double-count is present, totalDebt is overstated by up to $3.7 billion (11.5%), which would take net debt/EBITDA from 2.27x to roughly 2.14x — material to the leverage description, immaterial to the verdict.

5. Segments — the vendor's blocks do not tie, and the filing cannot fix them

seg_prod and seg_geo both fail the sum test for FY2025 and are rejected.

$24.1 billion — 27.2% of the company's revenue — is missing from the FY2025 segment block. The FY2024 rows tie almost exactly ($60,376M + $17,960M + $12,734M = $91,070M against revenue of $90,894M, 100.2%), so this is a defect specific to the most recent year, i.e. exactly the year a reader would use. Note also that seg_geo is not a geography block at all: its FY2025 keys are "UNITED STATES", "International" and "Supply Chain & Freight" — a segment table wearing a geography label, with the same 72.8% shortfall.

The filing cannot supply the corrected figures. UPS's filings in this archive were extracted without table preservation (tables: false for all six documents), so the segment revenue tables in both the 10-K and the 10-Q are absent from the text. We therefore report segment structure and drivers from the filings' narrative and decline to publish segment revenue dollars for FY2025 at all, rather than using a block we have shown to be 27% short. This is a real gap in the file and it is stated rather than worked around.

What the narrative does supply, from the Q1 2026 10-Q:

• U.S. Domestic Package: operating profit fell $464M; operating margin −320bp to 3.6% (adjusted 4.0%); cost per piece +9.7% (adjusted +9.5%), driven by USPS fees, Ground Saver transition costs and excess staffing, third-party aircraft leases after the Q4 2025 MD-11 retirement, weather and contractual wage increases. Compensation and benefits fell $392M.
• International Package: operating profit fell $94M; margin −270bp to 12.0% (adjusted 12.1%). Domestic average daily volume −6.6%, export −5.5%, both attributed to "revenue quality efforts and global trade policy changes in 2025, including de minimis exclusions." Revenue per piece +10.7% (domestic +15.9%, export +9.1%).
• Supply Chain Solutions: operating profit rose $159M; margin 8.1%. Forwarding revenue −$70M, Logistics −$163M on Mail Innovations declines, offset by +$188M in healthcare logistics from AHG and +$57M from Roadie and Happy Returns.

The shape of that is worth naming: the two package segments lost 270-320 basis points of margin while the smallest segment gained. Supply Chain Solutions, and specifically healthcare, is the only part of UPS currently improving.

6. Valuation — priced in or room?

At $109.095 (market cap $92.68B, ~849.5M shares):

Estimate coverage is good on FY2026 and FY2027 — 18 analysts on EPS in each — and thin beyond: 6 on FY2028 and 4 on FY2029. The FY2029 row is excluded from every conclusion in this dive; the FY2028 row is used only as a cross-check. The FY2027 EPS range is $7.207 to $8.401, a 16.6% spread, which is wide for a mature industrial and reflects genuine disagreement about whether the cost programme lands.

est.ebitdaAvg and est.ebitAvg are REJECTED — fixed-ratio fabrication, all eight years. ebitdaAvg is exactly 16.017% of revenueAvg in every single year from FY2023 through FY2030, and ebitAvg is exactly 12.391% in every year. Test against reality: FY2025 actual EBITDA per inc_a was $11,927M on $88,661M of revenue — a 13.45% margin — against the estimate block's $14,125M (16.02%), which is 18.4% too high. The block is a formula, not a forecast. All forward valuation in this dive uses epsAvg only. This is the thirteenth consecutive name on which this defect has appeared.

Peer set is unhelpful. The vendor peers — EMR, FDX, GD, HWM, ITW, JCI, MMM, NOC, TDG, WM — are a market-capitalisation cohort, not a business cohort. Only FedEx is a comparable. No peer multiple comparison is drawn.

6a. What today's price assumes (the inversion)

At $109.095 — 15.2x FY2026 consensus, 13.6x FY2027, 6.01% dividend yield — the price embeds:

• EPS of $7.18 in 2026 and $8.03 in 2027. (Consensus; 18 analysts each.) The first half delivered $1.07 + $1.76 = $2.83 on an adjusted basis, so FY2026 requires $4.35 in the second half — with Q3 consensus at $1.67, that leaves roughly $2.68 for Q4, against Q4 2025's actual $2.38. A 12.6% year-on-year Q4 improvement is required, and Q4 is the quarter with the most operating leverage in either direction.
• The ~$3 billion 2026 cost-savings target is achieved. (Company guidance, half-weighted per house rule.) Q1 delivered ~$600M. The consensus EPS path is essentially this number; without it, FY2026 does not reach $7.18 on a revenue base growing 3%.
• The dividend is maintained at $6.56 and the market keeps capitalising it at ~6%. (Our derivation.) This is the most fragile assumption in the price. Not because a cut is likely in the next four quarters — the balance sheet can fund the gap — but because the yield is doing a large part of the valuation work, the payout is 91% of forward earnings and above trailing free cash flow, and the buyback has already been sacrificed to defend it. At a 5.5% yield the stock is $119; at 7% it is $94. The entire street-consensus premium is a 50-basis-point move in the required yield.
• The Amazon glidedown completes in June 2026 with the stranded costs removed. (Company disclosure + our derivation.) Volume down >50% from 2024 levels; the 10-K warns that "in the event we are not able to successfully make appropriate adjustments or control related costs, our profitability could be materially impacted." The September quarter is the first period in which no further Amazon reduction is scheduled, so it is the first clean read.
• The market keeps paying roughly 13-15x forward earnings. (Our number.) At 12x FY2027E the stock is $96; at 16x it is $128. UPS traded above 20x in 2021-22 on peak earnings and below 12x in 2025 at the trough; 13.6x is mid-range and there is no obvious reason for it to move.

6b. The return bridge (why the multiple moves)

Expected return over the next twelve months decomposes as: EPS growth (+11.8% FY2026, +7.5% FY2027) + multiple drift (roughly HELD; our base of $112 is 13.95x FY2027E against 13.6x today, +2.5%) + dividend yield (+6.01%) ≈ +14% to +16% on the consensus path — of which more than 40% is the dividend.

Our base assumes the multiple holds. It does not assume expansion, and that is the reason the fair value lands where it does. There is a legitimate re-rating argument — a network that has finished shrinking, with a growing healthcare vertical and $3 billion less cost, ought to earn more than 13.6x. There is an equally legitimate de-rating argument — a company whose operating income has fallen 39% in three years and whose dividend consumes all of its cash ought to earn less. We resolve it by holding the multiple and letting the earnings do the work, which produces a fair value 2.7% above spot, which is a Hold.

Note what this makes the return: mostly dividend, some earnings growth, and nothing from re-rating. That is a bond-like return profile with equity-like risk to the coupon. It is a legitimate thing to own; it is not a Buy.

6c. Variant perception (where we differ, what would surprise)

• On the level we differ from the street by 5.9% and from spot by 2.7%, which is to say we have essentially no variant perception and we say so. Consensus target is $119 with 20 holds against 21 buys and 4 sells and a vendor consensus label of Hold. Our disagreement with the street is a modest one about the dividend's claim on cash flow, not about the earnings path.
• Where we do differ is on how the GAAP/adjusted gap should be weighted. The June quarter's $1.05 per share of adjustment is 148% of GAAP earnings, and adjustments of $1.3-1.5 billion are guided for the full year. We give the company the credit it has earned — the Q1 filing pre-announced the $1.2 billion Driver Choice charge and the arithmetic ties to the penny-ish — but we note that these programmes have now run continuously since Transformation 2.0, through Fit to Serve, into Network Reconfiguration and Efficiency Reimagined, and are guided to "conclude by 2027." (Company language, half-weighted.) Watchable number: whether FY2027 guidance contains any transformation-cost exclusion at all. A fourth consecutive programme would mean these are operating costs, not one-offs, and the correct earnings number is the GAAP one — which would put the stock on 20x, not 13.6x.
• We think the knowledge base's single positive claim deserves a hearing and we are not underwriting it. jordi_visser (2026-07-26, conviction 48, skill 2.0) argues that low-margin, labour-heavy firms are the underpriced AI winners because a 5% expense cut on a thin margin is a 20-25% EPS jump, naming UPS first. On a cost base near $81 billion, a 5% cut is $4 billion — larger than the entire current cost programme, and roughly $3.70 per share pre-tax. The claim is dated, falsifiable and from the highest-skill voice in the lane. It is also conviction 48, unsupported by anything in the filings, and describes a mechanism UPS has never claimed. We record it as the strongest available bull case and we do not put it in the base.
• Positive surprise that would force a re-rate: a September quarter with U.S. Domestic Package margin back above 7% and cost per piece falling year on year, confirming that the Amazon stranded costs came out on schedule; or the healthcare portfolio given separate revenue and margin disclosure, which would let the market value $11 billion of cold-chain logistics at something other than parcel multiples.
• Negative surprise that would break the thesis: a dividend cut, or any language softening the commitment to it; a goodwill impairment in Global Freight Forwarding ($877M) or Healthcare Logistics and Distribution ($738M), both flagged in the 10-Q as having "a limited excess of fair value above carrying value"; a downward revision to the ~$3 billion savings target; or a fourth transformation programme announced for 2027.

Synthos fair values

All three anchors are multiples of the FY2027 consensus EPS distribution (mean $8.027, low $7.207, high $8.401, 18 analysts), each cross-checked against the dividend-yield frame.

• Bear ~$88 — 12.2x the FY2027 consensus LOW of $7.207, and a 7.45% dividend yield on a maintained $6.56 (or roughly a 6.0% yield on a payout cut to $5.28). Cross-check: 6.6% above the 52-week low of $82.58; 12.3x FY2026E. The scenario: the cost programme underdelivers, Amazon stranded costs persist past June 2026, U.S. Domestic margin stays below 5%, and the dividend is cut to restore coverage. −19.3%.
• Base ~$112 — 13.95x the FY2027 consensus MEAN of $8.027, and a 5.86% yield on a maintained dividend. Cross-check: 15.6x FY2026E; 13.0x FY2028E. The scenario: ~$3 billion of savings arrives, FY2026 lands near $7.18, the dividend is held and just about covered as capex falls to $3.0 billion, and the multiple does not move. +2.7%.
• Bull ~$135 — 16.1x the FY2027 consensus HIGH of $8.401, and a 4.86% yield. Cross-check: 12.5x FY2029E; 12.5% above the 52-week high of $120.00. The scenario: the post-Amazon network runs at structurally better margins than consensus models, healthcare logistics is disclosed separately and re-rated, cost per piece falls year on year, and the buyback resumes in 2027. +23.7%.

Base is 2.7% above spot; asymmetry 1.23:1 (19.3% down, 23.7% up), plus a 6.01% dividend of contested durability. A base case inside 5% of spot with payoff barely above 1:1 is the definition of Hold, and we are not going to dress it as anything else.

7. Knowledge base — four claims, three bearish, one interesting

Raw hits: 24. After a case-sensitive entity re-run: 4. Used: 4. Discarded: 20.

The sweep ran entity terms UPS, United Parcel Service and United Parcel, plus free text on parcel, last-mile, package delivery and FedEx, across the full distilled claim store. A case-sensitive run was mandatory here — lowercase "ups" is one of the most common substrings in English, and a case-insensitive entity match is unusable. It was also, unusually, unnecessary: there is no homograph collision on UPS in this store. The 20 discards are all free-text hits — SpaceX as "the railroad of space" with last-mile logistics, sidewalk robots covering Uber's delivery TAM, and eighteen others across lex_fridman (4), all_in (2), real_vision (2), raoul_pal (2), dylan_patel (2) and nine more channels — none of which names the company. They are thematic noise, not entity noise, and they are discarded without further analysis.

The four surviving claims, verbatim:

> 2025-07-29 · BEARISH · conviction 80 · horizon: thesis · entities: UPS · channel: compound_and_friends · no named speaker · skill 1.0

> "UPS is a falling knife in the cleanest downtrend you'll ever see, down five years running; nobody's long it except on fundamentals — avoid."

> 2025-05-08 · BEARISH · conviction 65 · horizon: fact · entities: UPS, Penske · channel: forward_guidance · no named speaker · skill 1.0

> "Logistics layoffs already starting — UPS announced 20,000 cuts, Penske also — as containerized China volume (20% of trucking) collapses."

> 2025-11-05 · BEARISH · conviction 70 · horizon: thesis · entities: Amazon, UPS · channel: jordi_visser · no named speaker · skill 2.0

> "Labor market weakness is underestimated; companies will cut workers and replace them with digital/AI employees — ~70% odds of December cut."

> 2026-07-26 · BULLISH · conviction 48 · horizon: thesis · entities: UPS, CSCO, JBL, CNC, WMT · channel: jordi_visser · speaker_role: independent · skill 2.0

> "Low-margin, labor-heavy firms become big AI winners — a 5% expense cut yields ~20-25% EPS jump; agents deployed as infrastructure unlock margin benefit not yet priced."

What this lane is.

Three of the four are bearish and the highest-conviction of them has been wrong. compound_and_friends, 2025-07-29, conviction 80: "the cleanest downtrend you'll ever see... avoid." Twelve months later UPS has returned +29.1% against SPY's +24.3%. The claim was correct about the multi-year shape — maximum drawdown from peak is −53.0% — and wrong about the following year. We grade it as wrong on its own horizon and report it rather than quietly dropping it.

The forward_guidance claim (2025-05-08) is tagged horizon: fact and it is one — UPS did announce approximately 20,000 operational reductions in 2025, and the 10-Q's Network Reconfiguration disclosure is the same programme under its own name. A fact claim that has been confirmed by subsequent filings adds no forward information but does validate the store's sourcing.

The fourth claim is the only one that says something the price does not already know. Jordi Visser, 2026-07-26 — nine days before this dive — at conviction 48, argues that labour-heavy, thin-margin firms are the underpriced beneficiaries of agentic AI, and lists UPS first. The mechanism is arithmetically sound: UPS's cost base is roughly $81 billion against a 5.08% net margin, so a 5% cost reduction is roughly $4 billion pre-tax, about $3.70 per share, against $7.18 of forward EPS. That is the 20-25% EPS jump the claim describes, and it would be roughly $50 of value at a held multiple. Note the same speaker's 2025-11-05 claim is bearish on the same mechanism from the labour side — that firms will cut workers for digital ones — so the two claims are consistent rather than contradictory: bad for employment, good for the employer's margin. UPS is one of the largest private employers in the United States, which places it on both sides of that trade.

Attribution note: all four claims carry speaker: null except the 2026-07-26 claim, which carries speaker_role: independent. All are channel-attributed — jordi_visser at skill 2.0 supplies two of the four, compound_and_friends and forward_guidance one each at skill 1.0. jordi_visser at 50% of the used lane is below the one-third concentration threshold's practical concern only because the two claims point in opposite directions; we note the concentration and do not adjust for it. No claim is a management voice.

Conclusion. Breadth 3 channels, claim count 4, net conviction negative-low. The knowledge base is net bearish on UPS at low conviction, on claims that are mostly about labour-market conditions rather than about the company, with one recent, dated, falsifiable positive claim from its highest-skill voice that we record as the strongest available bull case and decline to underwrite.

8. Data integrity — what we rejected and why

Six findings. UPS's vendor file is, by this programme's standards, unusually honest about cash flow and unusually broken about segments.

1. Capex and free cash flow — CHECKED AND CLEAN. This is the finding, and it is a positive one. The vendor reports FY2025 capex of −$3,685M and FY2024 of −$3,909M; the 10-K states "capital expenditures of $3.7 and $3.9 billion for 2025 and 2024, respectively." Free cash flow is computed as operating cash flow MINUS capex ($8,450M − $3,685M = $4,765M), not plus. capexToDepreciationTTM of 0.8827 is below 1.0 — the signature that has flagged corruption elsewhere — but it is corroborated by the company's own guidance of "approximately $3.0 billion for all of 2026" against $3,746M of D&A. UPS is deliberately spending less than depreciation. The 5.92% free-cash-flow yield stands as reported and no correction is applied.

2. seg_prod and seg_geo do not sum for FY2025 — REJECTED. The three FY2025 lines total $64,517M against $88,661M of revenue, 72.8% coverage — $24.1 billion missing. FY2024 ties at 100.2%, so the defect is specific to the most recent year. seg_geo is additionally mislabelled: its keys are "UNITED STATES", "International" and "Supply Chain & Freight" — a segment table under a geography heading. No segment revenue dollars for FY2025 are published in this dive.

3. The filings were extracted WITHOUT tables — stated, not worked around. All six UPS documents carry "tables": false in the filings manifest. This means the income statement, balance sheet, cash flow statement and segment tables in the 10-K and both 10-Qs are absent from the extracted text. Every filing-sourced figure in this dive comes from narrative prose (the capex sentence, the dividend approval, the repurchase disclosure, the Amazon percentages, the cost-savings targets, the building closures) and each is quoted rather than paraphrased. The consequences: the lease double-count test could not be closed; FY2025 segment revenues could not be recovered to replace the rejected vendor block; and the Q2 2026 statements are unavailable because the Q2 10-Q post-dates this archive. These are gaps in the file, and they are reported as gaps.

4. est.ebitdaAvg and est.ebitAvg carry the fixed-ratio fabrication signature — REJECTED. ebitdaAvg is exactly 16.017% of revenueAvg in all eight years (FY2023-FY2030); ebitAvg is exactly 12.391% in all eight. Tested against actuals: FY2025 reported EBITDA was $11,927M (13.45% of revenue) against the block's $14,125M (16.02%) — 18.4% too high. All forward valuation uses epsAvg. Thirteenth consecutive name.

5. returnOnEquityTTM is present in km_ttm (0.2907) and ABSENT from ratios_ttm — the recurring block asymmetry. We use the km_ttm figure of 29.07% and say so. Note that a 29% return on equity here is a function of a small equity base ($16,227M against $73,090M of assets, financial leverage 4.73x), not of exceptional returns — returnOnInvestedCapitalTTM is 8.87% and returnOnAssetsTTM is 6.41%, which are the more informative numbers.

6. Quote and technical blocks disagree on the 52-week band — minor, and tech is used. quote.yearHigh/yearLow report $122.41 / $82.00; tech.hi52/lo52 report $120.00 / $82.58. Discrepancies of 2.0% and 0.7%. We use tech throughout for internal consistency with the moving averages and RSI, and record the difference. Separately, quote.price is $109.095 while tech.last is $109.11 — a 1.5 cent difference; the quote price is used.

Enterprise value — checked, small unexplained residual, ACCEPTED. enterpriseValueTTM of $116.696B against a rebuild of market cap $92.676B + total debt $32.290B − cash $5.887B + minority interest $0.028B = $119.107B. Residual $2.41B, 2.0% — well inside the 15% withholding threshold, and most likely explained by a more recent balance sheet than the 2025-12-31 bal_a snapshot. EV/EBITDA of 11.0x is used.

Corporate actions — 8-Ks scanned, two governance events found, no transactions. The three 8-Ks in the archive are: 2026-05-13, recording that Kevin M. Warsh's resignation from the Board became automatically effective upon his confirmation as Chairman of the Board of Governors of the Federal Reserve System; 2026-06-15, recording Eva Boratto's resignation from the Board on her appointment as CFO of Cencora; and 2026-07-28, the Item 2.02 results filing whose Exhibits 99.1 and 99.2 were not extracted. No acquisition, disposal, financing, segment realignment or authorisation change appears in the 8-K record. The AHG and Frigo-Trans acquisitions predate this window and are disclosed in the 10-K. This is one of the minority of names in recent batches with no undisclosed corporate action — recorded as a clean result.

Non-equity tripwire — checked and passed. UPS Class B common stock, NYSE-listed, per the 8-K cover pages (which also register Senior Notes due 2028 and 1.500% Senior Notes due 2032 — those are the listed debt securities, not this instrument). Beta 1.03; price $109.095; volume 3.07M shares (~$335M of turnover); 52-week band $82.58-$120.00, a 45% range. This is common equity. Note the dual-class structure: Class A shares (super-voting, family- and employee-held) are not the traded security.

9. Technicals

• Price $109.095, sitting exactly on the 50-day moving average of $109.105 — a 0.01% gap, which is as close to the line as a stock gets. +5.4% above the 200-day average of $103.55, with the 50-day above the 200-day.
• −9.1% from the 52-week high of $120.00; +32.1% above the low of $82.58. Position in the annual range: 71st percentile.
• RSI 44.1 — neutral, tilted soft, despite a +2.08% day. MACD −0.70 — negative, i.e. short-term momentum is still below the trend.
• Maximum drawdown from peak: −53.0%, the deepest reading in this batch and the number that gives the compound_and_friends claim its context. UPS peaked above $230 in 2022.
• Relative performance: 3-month +13.3% vs SPY +7.6% and QQQ +7.7%; 6-month −1.1% vs SPY +11.1%; 12-month +29.1% vs SPY +24.3% and QQQ +30.8%. The twelve-month figure is the one that surprises: UPS has beaten the S&P over the last year. The six-month figure shows why it does not feel that way — the gain was made in a burst and has since gone sideways for half a year.
• Sentiment: 2 strong buy, 19 buy, 20 hold, 4 sell across 45 analysts; vendor consensus label Hold. Target consensus $119 (+9.1%), median $120, high $132 (+21.0%), low $76 (−30.3%). The 74% spread between the high and low targets is the widest disagreement in this batch and is an accurate representation of the situation: the cost programme either lands or it does not.

Today's move and what it does to the entry

UPS closed 2026-08-04 at $109.095, up 2.08% or $2.225 from $106.87, opening at $107.125 and trading $105.66 to $109.42 — closing at the day's high on 3.07M shares. No company-specific filing is dated 2026-08-04; the last event was the 2026-07-28 earnings release, seven days earlier.

The honest read: the entry is neutral in every sense. Exactly on the 50-day, RSI at 44, MACD negative, fair value 2.7% above spot. There is no discount to buy and no extension to sell. For an existing holder the position is unchanged; for a new one, the 6.01% yield is the entire proposition and its coverage is the open question.

10. Insiders — eight transactions, all mechanical, zero signal

Every transaction in the file is dated 2026-05-15 and every one is mechanical. Three executives had restricted stock units vest (M-Exempt, price $0) and each had shares withheld at $100.78 to cover the tax (F-InKind). There is not one open-market purchase and not one open-market sale. Total shares disposed: 3,798, all to the taxing authority.

This is the signal-free case and it is reported as such. Unlike COF, where eight of eight were open-market sales, UPS's file contains no discretionary transaction at all. The absence of buying at a 6% yield and a 13.6x multiple is mildly notable; the absence of selling is mildly reassuring; neither is worth a basis point of the valuation. No transaction by chief executive Carol Tomé or by the chief financial officer appears in the window.

11. Verdict, kill-criteria and flip conditions

Hold.

The arithmetic that produces it. At $109.095, UPS trades at 15.2x the FY2026 consensus of $7.18 and 13.6x FY2027's $8.03, with a 6.01% dividend yield. Our base fair value of $112 is 2.7% above spot — inside the band where this programme declines to have a view — with asymmetry of 1.23:1 ($88 bear, $135 bull). The street is at $119 with a Hold consensus label and twenty holds against twenty-one buys. Nobody has an edge here, including us, and the correct response to that is to say so.

What is genuinely good. The cost programme is real, disclosed, quantified and already delivering — ~$600M in Q1 against a ~$3 billion 2026 target, 23 buildings closed with 27 identified, capex down from $5,158M in 2023 to a guided ~$3.0 billion. The June quarter grew revenue 7.6% in the month the Amazon glidedown completed, which is better than the shrinking-network story predicted. Healthcare logistics — more than $11 billion of 2025 revenue after AHG and Frigo-Trans — is the one line improving on merit, and it drove SCS operating profit up $159M while both package segments lost 270-320 basis points of margin. And the cash-flow data, the field most often corrupt in this feed, is verified correct.

What keeps it at Hold.

First, the dividend consumes all of the cash. $6.56 per share is 91% of FY2026 consensus EPS, 122% of trailing GAAP EPS and ~102% of trailing free cash flow; FY2025 paid $5.4 billion against $4,765M of free cash flow. Net debt rose $6.86 billion in 2025. And the company has told shareholders in writing that it "do[es] not anticipate further share repurchases in 2026" with $1.3 billion of authorisation unused. That is what protecting a dividend looks like from the inside.

Second, the GAAP and adjusted numbers tell opposite stories and have for four consecutive programmes. Q2 2026 GAAP operating income −49%; adjusted EPS +13.5%. The $1.05 per share gap is fully explained by the pre-announced $1.2 billion Driver Choice Program, and we credit that. But Transformation 2.0 ran into Fit to Serve which ran into Network Reconfiguration which ran into Efficiency Reimagined, and $1.3-1.5 billion more is guided for 2026. At some point a permanent adjustment is a cost.

Third, the growth is not volume. International domestic volume −6.6%, export −5.5%; revenue per piece +10.7%. Revenue has fallen in three of four years and operating income in all four. The consensus path to $8.03 in FY2027 is a margin path, and margins in the two package segments are currently going the wrong way.

Fourth, the knowledge base is net negative — three bearish claims to one bullish, and the bullish one is conviction 48 about a mechanism the company has never claimed.

Pre-registered KILL criteria — what would take this to Avoid:

• A dividend cut, or any Board language qualifying the commitment to $1.64 per quarter. The yield is the thesis for most holders; without it this is a shrinking industrial on 13.6x.
• The ~$3 billion 2026 cost-savings target revised down, or Q3 cumulative savings tracking below ~$1.8 billion.
• U.S. Domestic Package operating margin below 4% in the September quarter, i.e. no recovery after the Amazon glidedown completed in June.
• A goodwill impairment in Global Freight Forwarding ($877M) or Healthcare Logistics and Distribution ($738M), both explicitly flagged in the 10-Q.
• A fourth transformation programme announced for 2027-28, which would establish that the adjustments are structural.
• Cost per piece still rising year on year in the December quarter — it rose 9.7% in Q1 2026.

Pre-registered UPGRADE conditions — what would take this to Buy — Tactical:

• Free cash flow covering the dividend with room, i.e. FY2026 free cash flow above roughly $6.2 billion, which at ~$3.0 billion of capex requires operating cash flow near $9.2 billion.
• The buyback resuming, which is the single cleanest statement the Board can make about coverage.
• U.S. Domestic Package margin back above 7% with cost per piece falling year on year.
• Separate disclosure of the healthcare portfolio's revenue and margin. More than $11 billion of cold-chain logistics valued at parcel multiples is the clearest mispricing candidate in this file, and it is currently invisible.
• A price below roughly $95 on non-fundamental weakness, which at a maintained dividend is a 6.9% yield and would restore the asymmetry to better than 2:1.

Where UPS fits in the Synthos Framework Portfolio. No position. This is an income holding masquerading as an industrial, and the income is not covered. If the buyback resumes or free cash flow covers the dividend with a margin, it becomes a legitimate 1.5-2% yield sleeve position at these prices; until then there is no reason to own it in preference to a bond. Logged as a tracked Synthos call (Hold) as of 2026-08-04 at $109.095, with the fair-value anchors, kill criteria and upgrade conditions all gradeable.

Single biggest risk: the dividend is not covered by cash and the buyback has already been sacrificed to defend it. UPS paid $5.4 billion of dividends against $4,765M of free cash flow in FY2025, funded the gap with a $6.86 billion increase in net debt, and has told shareholders it will not repurchase stock in 2026 despite $1.3 billion of remaining authorisation. At $6.56 the payout is 91% of forward consensus earnings. The business can fund this for several years; the question is whether it should, and the answer depends entirely on a $3 billion cost programme with one quarter of evidence behind it. A cut would remove roughly 40% of the expected return and would repriced the shares against a different buyer base entirely.

Provenance & disclosures

• Traceability: 4 knowledge-base claims name United Parcel Service (raw hits 24, surviving a case-sensitive entity re-run 4, used 4, discarded 20; breadth 3 channels, net conviction negative-low). Three are bearish: compound_and_friends 2025-07-29 conviction 80 ("a falling knife in the cleanest downtrend you'll ever see... avoid" — graded WRONG on its own horizon; UPS returned +29.1% over the following twelve months against SPY's +24.3%); forward_guidance 2025-05-08 conviction 65, horizon fact (the 20,000-person layoff announcement — confirmed by subsequent filings); jordi_visser 2025-11-05 conviction 70, skill 2.0 (labour-market weakness underestimated, workers replaced by digital labour). One is bullish and is the most recent: jordi_visser 2026-07-26, conviction 48, skill 2.0, speaker_role: independent — "Low-margin, labor-heavy firms become big AI winners — a 5% expense cut yields ~20-25% EPS jump" — naming UPS first among five. Recorded as the strongest available bull case and deliberately excluded from the base valuation. The 20 discards are free-text thematic hits (SpaceX, sidewalk robots, Uber's delivery TAM) across lex_fridman, all_in, real_vision, raoul_pal, dylan_patel and nine other channels, none naming the company; there is no homograph collision on UPS in this store, which is itself worth recording given that lowercase "ups" is a common English substring — the case-sensitive discipline was applied and found nothing to remove. All four claims carry speaker: null except the last; all are channel-attributed; none is a management voice. jordi_visser supplies 2 of 4 used claims, in opposite directions.
• Data as-of: fundamentals — annual statements through 2025-12-31 and narrative segment, cost-programme and liquidity disclosures through 2026-03-31, from the 10-K filed 2026-02-17 and the 10-Q filed 2026-05-06; the June 2026 quarter is vendor-sourced only (inc_q and earn_cal), because the Q2 2026 10-Q post-dates this filing archive and the 2026-07-28 8-K's Exhibits 99.1 and 99.2 were not extracted · estimates 2026-08-04 · prices 2026-08-04, quote timestamp 1785873599 = 2026-08-04T19:59:59Z ($109.095, +2.08%; 50-DMA $109.105; 200-DMA $103.55; RSI 44.1; MACD −0.70) · knowledge-base claims 2026-08-04. UPS's fiscal year is the calendar year. All figures come from the Synthos vendor data file for UPS or from the SEC filings in the UPS archive; no figure comes from memory, recall or external retrieval.
• Filing archive contents and a limitation that must be stated: 10-K filed 2026-02-17 (fiscal 2025); 10-Q filed 2025-11-05 (September 2025 quarter); 10-Q filed 2026-05-06 (March 2026 quarter — the source of the cost-programme, building-closure, Driver Choice and segment-driver disclosures); 8-K filed 2026-05-13 (Kevin M. Warsh's Board resignation on confirmation as Chairman of the Federal Reserve); 8-K filed 2026-06-15 (Eva Boratto's Board resignation on appointment as CFO of Cencora); 8-K filed 2026-07-28 (Item 2.02 second-quarter results; exhibits not extracted). ALL SIX DOCUMENTS WERE EXTRACTED WITHOUT TABLE PRESERVATION ("tables": false). Consequently the income statement, balance sheet, cash flow and segment tables are absent from the text, every filing-sourced figure in this dive is quoted from narrative prose, the finance-lease double-count test on totalDebt could not be closed, and the rejected FY2025 segment block could not be replaced from the filing. These are gaps in the file and are reported as gaps rather than worked around.
• Where the filings CONFIRMED the vendor — the directed test, and it passed: capital expenditure and free cash flow are correct. The vendor's FY2025 capex of −$3,685M and FY2024 of −$3,909M match the 10-K's "capital expenditures of $3.7 and $3.9 billion for 2025 and 2024, respectively"; free cash flow is computed by subtracting capex ($8,450M − $3,685M = $4,765M); and the sub-1.0 capexToDepreciationTTM of 0.8827, a corruption signature elsewhere, is corroborated by the company's own guidance of "approximately $3.0 billion for all of 2026" against $3,746M of D&A. The 5.92% free-cash-flow yield stands as reported and no correction is applied — and it is this verified figure that establishes the dividend is not covered. Also confirmed: the $1.64 quarterly dividend approved 2026-01-27; the $1.0 billion of 2025 repurchases and the statement that no further repurchases are anticipated in 2026; Amazon at 10.6%/11.8%/11.8% of consolidated revenue for 2025/2024/2023; and the $1.2 billion Driver Choice Program that reconciles the June quarter's GAAP-to-adjusted gap to within a few cents.
• Where the vendor was REJECTED: seg_prod and seg_geo for FY2025, whose three lines sum to $64,517M against $88,661M of revenue — 72.8% coverage, $24.1 billion missing — while FY2024 ties at 100.2%; seg_geo is additionally mislabelled, carrying segment names under a geography heading. est.ebitdaAvg and est.ebitAvg, which are exactly 16.017% and exactly 12.391% of revenueAvg in all eight years and overstate FY2025's actual 13.45% EBITDA margin by 18.4% — all forward valuation uses epsAvg only, and the FY2029 row (4 analysts) is excluded from every conclusion. returnOnEquityTTM (0.2907) appears in km_ttm and is absent from ratios_ttm; the km_ttm figure is the one used, with the note that 29% on 4.73x financial leverage is less impressive than 8.87% return on invested capital suggests. quote.yearHigh/yearLow ($122.41/$82.00) disagree with tech.hi52/lo52 ($120.00/$82.58) by 2.0% and 0.7%; tech is used throughout.
• Corporate-action scan — RUN AND CLEAN. All three 8-Ks were read in full. They contain two Board resignations (Warsh, Boratto) and one results announcement. No acquisition, disposal, financing, ATM, segment realignment, authorisation change or debt assumption appears in the 8-K record, and the two 2025 acquisitions (Frigo-Trans, Andlauer Healthcare Group) are properly reflected in the vendor's cf_a acquisitionsNet of −$1,968M and in the goodwill step from $4,300M to $5,837M. UPS is one of the minority of recent names with no corporate action missing from the payload.
• Estimate coverage: 18 analysts on both FY2026 and FY2027 EPS — the anchor for all three fair values, with an FY2027 range of $7.207 to $8.401 (16.6%); 14 on FY2028 revenue but only 6 on FY2028 EPS; 4 on FY2029 EPS, excluded from every conclusion.
• Peer note: the vendor peer set — EMR, FDX, GD, HWM, ITW, JCI, MMM, NOC, TDG, WM — is a market-capitalisation cohort spanning aerospace, defence, HVAC and waste. Only FedEx is a business comparable. No peer-multiple comparison is drawn.
• Fair-value caveat: the $88 / $112 / $135 anchors are multiples of the FY2027 consensus EPS distribution — 12.2x the low of $7.207, 13.95x the mean of $8.027, and 16.1x the high of $8.401 — each cross-checked against the implied dividend yield on a maintained $6.56 payout (7.45%, 5.86%, 4.86%). Stated arithmetic, not a discounted cash flow. The base assumes the multiple holds; it assumes neither re-rating nor de-rating, and it assumes the dividend is maintained.
• Basis note: consensus estimates are on adjusted diluted EPS, which excludes transformation strategy costs (guided at $1.3-1.5 billion for 2026, of which $1.2 billion is the Driver Choice Program), goodwill and asset impairments, and income-tax valuation-allowance reversals. The trailing multiple of 20.3x is computed on GAAP earnings and the forward multiples of 15.2x and 13.6x on adjusted earnings; they are not comparable to each other and the gap between them is the entire adjustment question set out in Section 6c. The implied adjusted operating margin of ~9.1% for Q2 2026 is our arithmetic from the disclosed EPS gap, not a company figure.
• Accessibility note: no information in this dive is conveyed by colour. All emphasis is carried by bold text, table structure and explicit labelling.
• Not investment advice. Independent research, educational and informational only, never personalised. No recommendation to buy, sell or hold any security is made to any person.
• Version: 2026-08-04-full.