Capital One Financial COF

Reference table

1. What the business is, and the two acquisitions that reshaped it

Capital One describes itself in the July 2026 earnings release as "a leading technology-based financial services company with $484.3 billion in deposits and $673.8 billion in total assets as of June 30, 2026... a premier global payments provider and diversified financial institution, delivering a broad suite of products and consumer lifestyle and shopping experiences through its Credit Card, Consumer Banking including its Global Payment Network, and Commercial Banking lines of business. As the only major U.S. bank to migrate entirely to the public cloud..."

Chairman, founder and chief executive Richard D. Fairbank; approximately 76,300 employees.

Three reportable segments, from the June 2026 10-Q ($M):

Read that and Capital One is a credit-card company with two attachments. Credit Card is 74.2% of net revenue and Domestic Card alone is 70.1%. Commercial Banking, the only segment that is not consumer credit, is 5.4% and shrinking 9%. Consumer Banking's 26% growth is the interesting line: its non-interest income rose 97% year on year to $778M, which is where the Discover and PULSE network economics show up.

The Discover merger — 18 May 2025

From the 10-K: "On May 18, 2025... Discover Financial Services ('Discover') merged into Capital One and Discover Bank merged into CONA." The auditors' report and management's own internal-control assessment both state that Discover "constituted approximately 21% of total assets and 20% of total net revenue as of and for the year ended December 31, 2025" and was excluded from the 2025 internal-control assessment — a standard first-year exclusion, and a reminder that the integration is fourteen months old.

The accounting consequence dominates the FY2025 income statement and must be understood before any trailing number is used. From the 10-K: "Our provision for credit losses increased by $8.9 billion to $20.7 billion in 2025 as compared to 2024 primarily driven by the initial allowance for credit losses of $8.8 billion for non-PCD loans acquired in the Transaction." And: "Our allowance for credit losses increased by $7.2 billion to $23.4 billion as of December 31, 2025... Our allowance coverage ratio increased by 20 bps to 5.16%."

That $8.8 billion is a day-one requirement, not a credit event. Current expected credit loss accounting forces an acquirer to establish a full lifetime reserve against acquired non-purchased-credit-deteriorated loans through the income statement, even though the purchase price already reflected their expected losses. It ran the June 2025 quarter to a $4,277M net loss and a −32.99% return on tangible common equity, and it cut FY2025 GAAP net income to $2,453M against $4,750M in FY2024 — on 37% higher revenue. Every trailing earnings figure for Capital One is distorted by it, and every comparison in this dive says so.

The Brex acquisition — 7 April 2026, and it is nowhere in the payload

From the 10-Q:

> "On April 7, 2026, the Company completed its previously announced acquisition of Brex Inc.... Brex offers businesses solutions to issue corporate cards, automate expense management and make secure, real-time payments. The Brex acquisition enhances the Company's offerings in the business payments marketplace. The total consideration paid to Brex shareholders for the acquisition was approximately $4.5 billion and included $2.6 billion of cash consideration and 10.6 million shares of common stock... with a fair value of $1.9 billion. Immediately following the completion of the Brex acquisition, the Company paid off Brex's outstanding debt of $1.1 billion."

Approximately $5.6 billion of total outlay including the debt repayment, and 10.6 million shares issued — 1.7% of the current count. Brex results are folded into Domestic Card from the second quarter of 2026 with no separate revenue disclosure, alongside Capital One's legacy corporate-card product. Integration expenses were $96M in the quarter, $0.12 per share.

Nothing in the vendor payload reflects any of this. bal_a ends at 2025-12-31, cf_a ends at FY2025, acquisitionsNet for FY2025 reads +$16,465M (a positive number, i.e. cash acquired in the Discover merger, not cash paid), and there is no FY2026 row anywhere. This is the "whole transaction absent" defect class, and Capital One has two of them.

The knowledge base has something to say about Brex and it is not encouraging — see Section 7.

2. The vendor's revenue line is a gross-up the company does not publish

Capital One's published top line is "total net revenue" — net interest income plus non-interest income. The vendor reports gross interest income plus non-interest income, before deducting interest expense.

The 10-K states FY2025 total net revenue of $53,434M, against $39,112M in FY2024 and $36,787M in FY2023. The vendor's inc_a reports FY2025 revenue of $69,252M.

The reconciliation is exact: $69,252M less the vendor's own interestExpense of $15,818M = $53,434M, to the dollar. So the vendor is not wrong about the components; it is publishing a basis Capital One does not use, and it is 29.6% too high.

Every ratio computed against the vendor's revenue line is therefore understated by roughly 30%, including priceToSalesRatioTTM of 1.73x (correct figure on published revenue: approximately 2.24x), netProfitMarginTTM of 13.4% (correct: approximately 17.5%) and operatingProfitMarginTTM of 21.0%. This is the SCHW gross-up defect class, and it is confirmed here by exact arithmetic against the filing. Every revenue and margin figure in this dive uses the published total-net-revenue basis.

The trailing quarterly series is on the same gross-up basis — inc_q reports June-quarter revenue of $19,861M against the company's published $15,850M — and is used in this dive only for net income, where it matches.

3. The earnings inflection, quarter by quarter

This is the table that carries the thesis, and it needs the accounting context of Section 1 to read correctly.

The June 2026 quarter against the June 2025 quarter is the cleanest available statement of what the merger did:

Four observations.

First, the credit line is the surprise and it runs the right way. The 10-Q attributes the Domestic Card improvement to "favorable observed credit performance" — a 49 basis point fall in the loss rate on a $275.4 billion loan book. The quarter included a $662 million reserve release against net charge-offs of $3.6 billion. In a period when several independent voices in the knowledge base have been warning about consumer-credit deterioration, the largest US card issuer's actual charge-off rate went down. That is the most important fact in this dive and it is filing-verified.

Second, expenses grew nearly as fast as revenue. Total non-interest expense +29% against revenue +27%, with marketing +23% and operating expense +31%. The efficiency ratio of 57.05% is respectable for a card issuer but not improving, and the operating-expense growth includes both integrations. Positive operating leverage from the merger has not yet appeared and is a large part of what the FY2027 consensus needs.

Third, the adjusted-to-GAAP bridge is disclosed cleanly and is finite. June-quarter GAAP EPS of $4.73 becomes adjusted EPS of $5.81 through three items the release itemises: acquisition amortisation $494M pretax ($0.60), Discover integration expenses $298M ($0.36) and Brex integration expenses $96M ($0.12). Two of the three are transitional and will fall away; the amortisation runs longer. A 22.8% adjustment is large in absolute terms but its composition is transparent, which is more than can be said for several other names in this batch.

Fourth, the beat pattern is erratic. $5.95 against $4.49 (+32.5%), then $3.86 against $4.14 (−6.8%), then $4.42 against $4.50 (−1.8%), then $5.81 against $4.79 (+21.3%). Two large beats and two small misses in four quarters. Consensus is visibly struggling to model a company mid-integration, and that is a reason to weight the reported credit and capital metrics above the earnings headline.

4. Balance sheet, capital and the capital return

All figures from the 10-Q at 2026-06-30, filing-verified:

The capital position is solid and the trend is deliberate. CET1 of 13.7% is well above requirement and 60 basis points lower than a year ago — capital being returned rather than eroded. Deposits of $484.3B fund 71.9% of $673.8B of assets, which is the single most important structural feature of this business: a card book funded by insured retail deposits rather than wholesale markets does not experience a funding run in the way a monoline lender does.

The buyback is running at a pace that changes the arithmetic of owning the stock. Treasury stock at cost went from $10,146M to $14,269M across the six months to 2026-06-30 — $4,123M of stock repurchased in half a year, 3.0% of the current market capitalisation. Treasury shares rose 17.18 million. Shares outstanding nonetheless fell only from 625.1M to 613.5M (−1.9%), because 10.6 million shares were issued as part of the Brex consideration.

Annualised, that repurchase pace is roughly $8.2 billion against a $136.1 billion market capitalisation — approximately 6.0% per year. For comparison, FY2025's full-year repurchase was $4,599M and FY2024's was $734M. The pace has increased roughly elevenfold in two years, which is what a bank does when it has excess capital and believes its own stock is cheap.

The dividend. The vendor reports dividendPerShareTTM of $3.00, a 1.35% yield, and lastDividend of 3. The filing suggests the forward run-rate is higher: the statement of changes in stockholders' equity shows a common dividend of approximately $505M in the March 2026 quarter against roughly 620 million shares — approximately $0.81 per share per quarter, or $3.24 annualised, a 1.46% forward yield. We use the filing-implied forward figure and flag the vendor's as trailing.

Total shareholder yield is therefore approximately 6.0% buyback plus 1.46% dividend ≈ 7.4%. At 9.3x FY2027 consensus earnings that is not a rounding item; it is roughly half the expected return.

Cash flow is not a useful lens for a bank and we do not use it as one. The vendor reports FY2025 operating cash flow of $27,718M and free cash flow of $26,140M, a freeCashFlowYieldTTM of 23.65%. For a lender, operating cash flow includes changes in loan balances and deposits and bears no relation to distributable earnings. freeCashFlowYieldTTM, incomeQualityTTM (2.99), evToFreeCashFlowTTM and priceToFreeCashFlowRatioTTM are all rejected as lenses for this name. The relevant measures are return on tangible common equity (18.04%), the CET1 ratio (13.7%) and tangible book value growth (+5.9% year on year).

5. Credit — the only variable that matters

Capital One is a spread lender with a 74.2% concentration in credit cards. Its earnings are total net revenue less non-interest expense less credit losses, and the third term has by far the widest distribution.

The current position, all from the 10-Q:

Two things run in opposite directions and both should be reported.

Card credit is improving materially. The 10-Q: "The net charge-off rate decreased by 49 bps to 4.71% in the second quarter of 2026 compared to the second quarter of 2025 and decreased by 72 bps to 4.88% in the first six months of 2026... primarily due to favorable observed credit performance." A 49 basis point improvement on a $275 billion book, with a $662 million reserve release on top, is a genuine and material positive, and it is the opposite of what a deteriorating consumer would produce.

Consumer Banking credit is deteriorating. "The net charge-off rate increased by 18 bps to 1.48% in the second quarter of 2026... and increased by 14 bps to 1.59% in the first six months." That segment is predominantly auto lending, and subprime auto is precisely the stress point that five independent knowledge-base sources have been flagging since 2023 (Section 7). At 1.48% on a much smaller book the dollar impact is modest — but the direction is the one the bears named, in the sub-portfolio they named.

The sensitivity, stated plainly. Domestic Card loans are the majority of a $275.4 billion book. A 200 basis point rise in the card charge-off rate is roughly $4-5 billion of additional annual pre-tax provision — against an FY2027 consensus that implies roughly $14-15 billion of pre-tax income. That is not a tail scenario; it is roughly the distance between the current 4.71% and the levels this portfolio reached in 2009 and briefly in 2020. The allowance at 5.16% coverage is designed to absorb the first tranche, and the deposit-funded balance sheet means Capital One survives it comfortably — but the equity does not survive it at 9.3x forward earnings without a large drawdown.

This is why the verdict is Tactical. The valuation is attractive conditional on the card loss rate staying near current levels, and the kill criteria in Section 11 are all card-credit metrics for exactly that reason.

6. Valuation — priced in or room?

At $221.895 (market cap $136.13B, 613,484,836 shares):

Estimate coverage is good on the near years — 14 analysts on both FY2026 and FY2027 EPS — and thin beyond. FY2028 rests on 9 analysts and FY2029 on 3, and the FY2029 row is excluded from every conclusion. The FY2027 range is $22.693 to $24.898, a 9.7% spread — wide enough to be informative and narrow enough to anchor on.

est.ebitdaAvg and est.ebitAvg are rejected outright and would be meaningless for a bank even if they were internally consistent. EBITDA is not a concept that applies to a lender whose principal cost is interest. The rows also carry the fabrication signature: in every forward year ebitdaAvg is exactly 27.85% of revenueAvg and ebitAvg is exactly 19.98%. And netIncomeAvg exceeds ebitAvg in FY2027 ($12,881M against $13,545M — narrowly not, but by FY2029 $21,202M against $15,712M, which is impossible). All forward valuation uses epsAvg.

Enterprise value is rejected as a concept here. The vendor prints enterpriseValueTTM of $175.4B and netDebtToEBITDATTM of 1.79x. For a deposit-funded bank, "net debt" is not a meaningful quantity — deposits are the raw material, not leverage — and the vendor's own returnOnInvestedCapitalTTM of −3.87% on investedCapitalTTM of −$334 billion demonstrates that its capital framework simply does not accommodate this balance sheet. All enterprise-value-based metrics in this file are discarded.

Peer context is thin but not useless. The vendor peer set — Bank of America, Robinhood, Progressive, Santander, S&P Global, TD, UBS — contains no card issuer at all, which for the largest card-led bank in the United States is a significant gap. No peer multiple comparison is drawn. The relevant comparison set (American Express, Synchrony, Bread, and formerly Discover itself) is absent from this file entirely.

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

At $221.895 — 10.9x FY2026 consensus adjusted EPS, 9.3x FY2027 and 2.11x tangible book — the price embeds:

• Adjusted EPS reaches $20.44 in 2026 and $23.88 in 2027. (Consensus; 14 analysts each.) The first six months of 2026 delivered $10.23, so the FY2026 figure needs $10.21 in the second half — essentially a repeat, which the September guide of $5.37 is consistent with. FY2027's $23.88 requires 16.8% growth on top, from operating leverage, falling integration costs and a shrinking share count.
• The Domestic Card net charge-off rate stays near 4.71%. (Our derivation; the filing supplies the rate.) This is the most fragile assumption in the price, not because deterioration is likely — the trend is favourable and the rate fell 49 basis points year on year — but because it is the assumption with the widest distribution and the largest coefficient. A move to 6.5% is roughly $4-5 billion of annual pre-tax earnings, most of the consensus.
• The buyback continues at roughly the current pace. (Our number, from the $4,123M of H1 repurchase.) At approximately 6% of market capitalisation annually, the buyback contributes roughly 6 percentage points to per-share growth. If CET1 pressure or a credit turn stops it, roughly a third of the expected return disappears without any change in earnings.
• The market keeps paying roughly 9-11x forward earnings and roughly 2.1x tangible book. (Our number.) At 8x FY2027E the stock is $191; at 12x it is $287. The multiple band is narrower than for most names in this batch because bank multiples are anchored by book value, which is why the tangible-book cross-check matters.
• Neither integration produces a negative surprise. (Our number.) Discover's internal controls were excluded from the 2025 assessment; Brex has been owned for four months and is reported inside Domestic Card with no separate disclosure. Two simultaneous integrations at a combined cost of roughly $14 billion of purchase consideration is a lot of unobservable machinery.

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

Expected return over the next twelve months decomposes as: adjusted EPS growth (+16.8%, from FY2026E $20.436 to FY2027E $23.880) + multiple drift (roughly HELD, from 10.9x on the forward year to about 10.5x on the then-forward year, −4%) + shareholder yield (+7.4%) ≈ +15% to +20%.

Our base assumes the multiple holds rather than expands, and that is a deliberate refusal to underwrite the strategic case. There is a legitimate argument that owning a payments network should move Capital One's multiple toward American Express territory over time. We are not paying for it. The base case is a card issuer earning 18% on tangible equity, valued at roughly 10.5x forward earnings and just over 2x tangible book, returning 7% a year to shareholders. The re-rating, if it comes, is free optionality in the bull case rather than a requirement of the base.

Note what this makes the return: roughly half earnings growth, roughly half shareholder yield, and essentially none from the multiple. That is an unusually robust decomposition and is the strongest structural argument for the verdict — a return that does not require the crowd to change its mind about anything.

If the multiple compressed to 8x FY2027E the price would be $191 (−13.9%). If it expanded to 12x, $287 (+29.3%). The bull case at $325 is 13.1x the FY2027 consensus high — a level that requires both an earnings beat and the network re-rating.

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

• On the level we have NO variant perception, and we say so plainly. Our base of $255 is within 0.4% of the street's $256, and 36 of 57 analysts are at buy. We are not seeing anything the market is missing about the earnings path. What we are doing is taking a position on the shape of the return — earnings growth plus a very large shareholder yield, with no multiple expansion required — rather than on a differentiated forecast. That is a legitimate basis for a Tactical position and an illegitimate basis for a Core one, which is exactly how it is graded here.
• We differ from the vendor data on three material points, all documented in Section 8. Revenue is reported on a gross-up basis 29.6% above the company's published total net revenue; the $4.5 billion Brex acquisition is absent from every field; and the enterprise-value and return-on-invested-capital framework produces figures (a −$334 billion invested capital base) that are not merely wrong but structurally inapplicable. Anyone screening this name on the vendor block is looking at a different company.
• We think the credit improvement is being under-weighted relative to the macro narrative. The knowledge base carries five independent channels warning about consumer-credit deterioration between 2023 and 2026. The actual Domestic Card charge-off rate at the largest US card issuer fell 49 basis points year on year in the most recent quarter, with a $662 million reserve release. Watchable number: the Domestic Card net charge-off rate in the 2026-10-20 print. A third consecutive year-on-year improvement makes the bear narrative harder to sustain; a reversal above 5.2% validates it and is a kill criterion.
• We think the Brex acquisition deserves more scepticism than it is getting, and the knowledge base is the reason. Capital One paid $4.5 billion plus $1.1 billion of debt repayment on 7 April 2026 for a corporate-card and expense-management platform. Four days earlier, a knowledge-base claim (empire, 2026-04-03, conviction 70) stated that in that market 'Ramp [is] leaving Brex behind'. That is one independent voice, channel-attributed, and it is not dispositive — but it is the only outside read on the asset in our possession and it is negative. Brex has no separate revenue disclosure and is folded into Domestic Card.
• Positive surprise that would force a re-rate: the Global Payment Network being given its own disclosure with volume and margin economics, which would begin the American Express comparison in earnest; or a September quarter with the efficiency ratio below 55% and Domestic Card charge-offs below 4.5%, which would deliver the operating leverage the FY2027 consensus needs and the credit improvement the bears say cannot last.
• Negative surprise that would break the thesis: Domestic Card charge-offs above 5.2%, i.e. back to where they were a year ago; a reserve build rather than a release; CET1 below 13% with the buyback suspended; a Brex impairment; or any disclosure that the Discover integration is running behind. Any of these hits a stock that has risen 16.8% in three months.

Synthos fair values

All three anchors are multiples of the FY2027 consensus adjusted EPS distribution (mean $23.880, low $22.693, high $24.898, 14 analysts), each cross-checked against tangible book value of $105.21 per share.

• Bear ~$180 — 7.9x the FY2027 consensus LOW of $22.693, and approximately 1.7x current tangible book. Cross-check: 8.8x FY2026E; 2.2% above the 52-week low of $176.10. The scenario: the card credit cycle turns, charge-offs move toward 6.5%, reserve releases become builds, the buyback is suspended to defend CET1, and the market applies a trough multiple to trough earnings. −18.9%.
• Base ~$255 — 10.7x the FY2027 consensus MEAN of $23.880, and approximately 2.4x current tangible book (roughly 2.1x tangible book as it stands at the end of 2027 on current accretion). Cross-check: 12.5x FY2026E; 9.2x FY2028E. Sensitivity, stated openly: 9.5x gives $227 — essentially spot — and 12x gives $287. The whole answer lives in a 9.5-12x band on one estimate row. The scenario: credit holds near current levels, the integrations complete on schedule, operating leverage appears, the buyback continues at roughly 6% a year, and the multiple does not move. +14.9%.
• Bull ~$325 — 13.1x the FY2027 consensus HIGH of $24.898, approximately 3.1x current tangible book. Cross-check: 11.7x FY2028E; 26.0% above the 52-week high. The scenario: card credit improves further, the Global Payment Network produces disclosable economics, and the market begins valuing Capital One partly as a network rather than wholly as a lender. +46.5%.

Base is 14.9% above spot; asymmetry roughly 2.46:1 to the upside (18.9% down, 46.5% up), before a shareholder yield of approximately 7.4%. That combination — a base case above 14%, a payoff ratio near 2.5:1, and a return that requires no multiple expansion — is what the Buy — Tactical tier exists for. What keeps it out of the Core tier is set out in Section 11.

7. Knowledge base — sixty-four hits, one name-level claim, and it is bearish

Raw hits: 64. After a case-sensitive entity re-run: 5. Entity matches: 4. Name-level claims on Capital One: 1, and it is BEARISH. Discarded: 59.

The primary sweep ran entity terms COF, Capital One, Capital One Financial, Discover Financial and Brex, plus free text on capital one, subprime, credit card and consumer credit, across all 51,928 distilled claims. It returned 64 hits across 27 channels — the largest raw lane in this batch and almost entirely noise for the purpose it was run.

The discard, and why it is a discard rather than a lane. Fifty-nine of the 64 hits are generic macroeconomic claims about consumer credit that never name Capital One, caught by the free-text terms. They span arthur_hayes (6), forward_guidance (5), niall_ferguson (5), luke_gromen (4), eurodollar_university (4) and twenty-two other channels, and cover topics from 2008 mortgage causation to blockchain rails to bitcoin. A case-sensitive entity re-run collapses the lane to five hits, which is the correct discipline for a three-letter ticker and is the same remedy applied to APP elsewhere in this batch.

The five surviving claims, verbatim:

> 2023-01-19 · BEARISH · conviction 72 · horizon: thesis · entities: DFS, COF · channel: warren_pies · no named speaker · skill 1.0

> "Short card issuers that carry balance-sheet/credit risk — Discover weak after results, Capital One worse impacted — as consumer credit deteriorates."

> 2025-05-09 · bearish · conviction 60 · horizon: principle · no entities · channel: all_in · no named speaker

> "Subprime lenders' price-to-book (credit acceptance vs Capital One spread) hitting highs historically portends a liquidity crisis — blinking yellow."

> 2026-04-03 · bullish · conviction 70 · horizon: thesis · entities: Ramp, Brex, Rain · channel: empire · no named speaker

> "Winners are running away fast — Ramp leaving Brex behind, Rain's month-on-month growth still accelerating at scale; concentrate capital into accelerating leaders."

> 2026-03-17 · bullish · conviction 70 · horizon: thesis · entities: Ramp, Brex · channel: invest_like_the_best · no named speaker

> "Enterprise software that controls the underlying money (not just software) wins; Ramp/Brex prove you must control the dollar."

> 2024-08-08 · bullish · conviction 55 · horizon: thesis · entities: Brex, Coinbase · channel: no_priors · no named speaker

> "Bullish on crypto; Brex is building in the space, but core value is durable money-movement rails regardless of crypto."

What this lane is, stated without inflation.

There is exactly one name-level claim on Capital One in 51,928 distilled claims, and it is a short recommendation. Dated 2023-01-19, conviction 72, it argues for shorting card issuers with balance-sheet credit risk as consumer credit deteriorates, naming Discover and Capital One. It has been decisively wrong. Consumer credit did not deteriorate to the degree the claim required; Capital One went on to acquire Discover rather than be damaged alongside it; and the stock is materially higher. We report it and grade it as wrong rather than quietly excluding it, because a knowledge base that only surfaces claims that aged well is not a knowledge base.

Three of the five claims are about Brex — and that is the most useful thing in this lane. None of the three was written about Capital One; all three predate or barely postdate the acquisition; and the most recent, dated 2026-04-03 — four days before Capital One closed a $4.5 billion purchase — says of that market that 'Ramp [is] leaving Brex behind'. The other two are structurally positive about the category ("you must control the dollar", "durable money-movement rails"). Net: one independent outside read on the competitive position of an asset Capital One just bought, and it is unfavourable; two positive reads on the category it sits in. Capital One provides no separate revenue disclosure for Brex, so there is nothing in the filings to test it against.

And the discarded macro lane should not be dismissed even though it is discarded. It is not homograph noise in the way the "Cash App" collisions elsewhere in this batch are — it is thematically relevant material that simply fails the entity test. Multiple independent channels warn on consumer-credit deterioration between 2023 and 2026: forward_guidance ("subprime auto delinquencies, K-shaped consumer... point to wheels coming off", 2025-11-14), luke_gromen ("rising AI-driven joblessness will spread a consumer credit problem", 2026-03-11), lyn_alden ("lower-income consumers are stressed with subprime auto defaults", 2025-10-27), andreas_steno ("subprime auto default rates are starting to surge", 2023-11-19) and eurodollar_university. Against them, compound_and_friends (2026-01-30) argues "Low-grade consumer stress in subprime/auto is real but the lower 50% matters less to GDP, so it's a political story, not yet a market story" and cites Ally Financial near a 52-week high as forward-looking evidence.

That disagreement is genuinely unresolved and it bears directly on the only variable that determines Capital One's earnings. It is not name-level conviction and it does not enter the conviction rating. It is why this is a Tactical position with credit-based kill criteria rather than a Core one.

Attribution note: all 64 hits carry speaker: null and are attributed only to a channel — the weakest sourcing the 4-lane policy admits, and it applies to the entire lane including the one bearish name-level claim. No claim carries a management speaker_role. No concentration sensitivity test is meaningful across five claims from five different channels.

Conclusion. Breadth 5, claim count 1, net conviction negative-low. The Synthos knowledge base is mildly negative on Capital One, on the basis of a three-and-a-half-year-old short call that has been wrong, and mildly negative on the asset Capital One most recently bought. Neither is strong. Both are stated.

8. Data integrity — what we rejected and why

Eight findings. Capital One's vendor file is among the least usable in this programme: the revenue line is on a basis the company does not publish, an entire $4.5 billion acquisition is missing, and the capital framework is structurally inapplicable to a bank. Against that, the share count and tangible book value are both correct — including the book-versus-tangible-book test this dive was specifically asked to run.

1. revenue is a gross-up the company does not publish — REJECTED and rebuilt. The vendor reports FY2025 revenue of $69,252M; the 10-K states total net revenue of $53,434M. The reconciliation is exact: $69,252M less the vendor's own interestExpense of $15,818M = $53,434M. The vendor reports gross interest income plus non-interest income; Capital One publishes net interest income plus non-interest income, and every analyst estimate in the file is struck on the latter. Consequential rejections: priceToSalesRatioTTM 1.732x (correct ≈ 2.24x), netProfitMarginTTM 13.38% (correct ≈ 17.5%), operatingProfitMarginTTM 21.02%, ebitdaMarginTTM 27.96%. The same gross-up runs through inc_q — June-quarter revenue of $19,861M against the published $15,850M. All revenue and margin figures in this dive use the published basis.

2. The $4.5 billion Brex acquisition is absent from every vendor field. Completed 7 April 2026: $2.6B cash plus 10.6 million shares worth $1.9B, plus $1.1B of Brex debt repaid immediately after closing. bal_a ends at 2025-12-31, cf_a ends at FY2025, and there is no FY2026 row anywhere in the payload. The 10.6 million shares issued are, however, reflected in the current share count. This is the "whole transaction absent" defect class.

3. acquisitionsNet for FY2025 reads +$16,465M — a POSITIVE number — and is misleading. In a year in which Capital One acquired Discover, the acquisitions line shows a $16.5 billion cash inflow. That is cash acquired in a stock-for-stock merger, not consideration paid, and a reader treating it as acquisition spend would conclude Capital One was a net seller of businesses in 2025. Not used.

4. Enterprise value, invested capital and all derived metrics are structurally inapplicable to a deposit-funded bank — REJECTED as a class. enterpriseValueTTM of $175.4B treats $484 billion of customer deposits as if they were absent and $51 billion of borrowings as if they were leverage. netDebtToEBITDATTM of 1.79x, evToSalesTTM, evToEBITDATTM, interestCoverageRatioTTM of 0.99x and freeCashFlowYieldTTM of 23.65% are all meaningless here. The clearest demonstration is the vendor's own investedCapitalTTM of −$334,028M and returnOnInvestedCapitalTTM of −3.87%: a negative capital base producing a negative return for a company that earned 18.04% on tangible common equity in the most recent quarter. All of these are discarded; the dive uses CET1, ROTCE and tangible book value instead.

5. seg_prod stops being a segment table after FY2023 — REJECTED for FY2024 and FY2025. The FY2023 and earlier entries carry the genuine segment structure (Credit Card $25,669M, Consumer Banking $9,302M, Commercial Banking $3,520M, Other −$1,704M). The FY2024 and FY2025 entries are instead the revenue-from-contracts-with-customers fee disclosure — "Interchange Fees, Contracts" $6,443M, "Service Charges And Other Customer Fees" $857M, "Other Contract Revenue" $762M — summing to $8,062M, which is not revenue on any basis. So the payload has no segment data covering either the Discover or the Brex period. All segment figures in this dive come from the 10-Q.

6. seg_geo stops at FY2023 and contains a corrupt row. The most recent entry is FY2023 (United States $35,400M, International $1,387M). The FY2020 row reports International revenue of $12,341M against $1,376M in FY2022 and $1,400M in FY2019 — an eightfold spike with no corresponding event, almost certainly a mis-mapped domestic line. The block is not used.

7. bookValuePerShareTTM does not deduct preferred equity — minor overstatement. The vendor reports $183.389, which is total stockholders' equity divided by shares. Capital One has 4,980,700 preferred shares outstanding, so common equity is lower and book value per common share is roughly $177, an overstatement of approximately 3.4%. shareholdersEquityPerShareTTM is identical to bookValuePerShareTTM, confirming no preferred deduction is applied.

8. est.ebitdaAvg and est.ebitAvg carry the fixed-ratio fabrication signature AND are conceptually void — REJECTED. In every forward year ebitdaAvg is exactly 27.85% of revenueAvg and ebitAvg is exactly 19.98%. netIncomeAvg exceeds ebitAvg by FY2029 ($21,202M against $15,712M), which is impossible. Separately, the FY2022 and FY2023 rows report ebitAvg of $10.6 million and $11.7 million — three orders of magnitude wrong — while epsAvg for those years ($18.61 and $12.89) is broadly sensible. All forward valuation uses epsAvg only. The FY2029 row rests on 3 analysts and is excluded.

The book-versus-tangible-book test — RUN, and PASSED. This dive was specifically directed to check whether book and tangible book are reported as identical, the defect found on GS, AXP and MS. They are not. The vendor reports bookValuePerShareTTM of $183.389 and tangibleBookValuePerShareTTM of $106.124 — a $77.27 difference reflecting $28,509M of goodwill and $16,578M of intangibles. The filing's own figure is $105.21 per share at 2026-06-30, so the vendor is within 0.9%. We use the filing's $105.21, and record the vendor's tangible book as substantially correct.

Also correct and worth recording: the share count — the 10-Q states 613,484,836 shares outstanding at 2026-06-30, against 613,485,000 implied by market capitalisation divided by price, an exact match; effectiveTaxRateTTM of 21.9%, consistent with the quarterly provisions; dividendPerShareTTM of $3.00 as a trailing figure (the forward run-rate implied by the filing is nearer $3.24); and quote.yearHigh/yearLow ($259.64/$174.24) against tech.hi52/lo52 ($257.94/$176.10) — a 0.7% and 1.1% discrepancy; we use tech for internal consistency.

Vendor composite rating — accepted with a caveat. The rating block gives A- / 4 overall, the highest in this batch, with 5 out of 5 on discounted cash flow. The discounted-cash-flow sub-score is built on the free-cash-flow figure rejected in finding 4 and should carry no weight. The remainder is unobjectionable but adds nothing this dive has not derived directly.

Non-equity tripwire — checked and passed. COF is common stock, $0.01 par value, NYSE-listed, per the 10-Q balance sheet (1,000,000,000 shares authorised, 714,105,102 issued, 613,484,836 outstanding). Price of $221.895 is not par-like; beta is 1.022; the dividend is variable; volume was 3.14M shares (~$696M of turnover); the 52-week band of $176.10 to $257.94 is a 46% range. Note the separate 4,980,700 preferred shares, which are NOT this security. This is common equity.

9. Technicals

• Price $221.895. −14.0% from the 52-week high of $257.94; +26.0% above the 52-week low of $176.10. Position within the annual range: 56th percentile.
• Both moving averages are below the price and the shorter one is above the longer: +11.9% above a 50-day average of $198.37; +7.0% above a 200-day average of $207.45; the 50-day sits below the 200-day but is rising fast toward it. This is the only name in this batch trading above both averages.
• RSI 64.0 — strong, not overbought. MACD +4.42 — the most positive reading in this batch.
• Maximum drawdown from peak over the trailing year: −14.0%, the shallowest in this batch.
• Relative performance, with the shape that matters: 3-month +16.8% against SPY +7.6% and QQQ +7.7% — outperforming by 9 points; 6-month −0.8% against SPY +11.1%; 12-month +7.0% against SPY +24.3% and QQQ +30.8% — a 17-point deficit. The underperformance is a year old and the last quarter has reversed it decisively. That is consistent with the earnings inflection documented in Section 3: the stock lagged through the Discover accounting distortion and has re-rated since it cleared.
• Sentiment: 36 buy, 17 hold, 4 sell across 57 analysts — the widest coverage in this batch; consensus target $256 (+15.4%), median $255, high $300 (+35.2%), low $229 — 3.2% ABOVE spot. The lowest analyst target in the file is above the current price, which is unusual and means no covering analyst currently sees downside. We treat that as a caution rather than a comfort: unanimity about the absence of downside is itself a risk marker in a credit-cycle business.

Today's move and what it does to the entry

COF closed 2026-08-04 at $221.895, up 1.94% or $4.215 from $217.68. It opened at $219.33, traded $217.45 to $222.92, and closed within 0.5% of the day high on 3.14M shares. No company-specific filing is dated 2026-08-04; the last event was the 2026-07-21 earnings release, fourteen days earlier, and the last insider filing was 2026-08-03.

The honest read: this is a stock in an established short-term uptrend, and the entry is therefore NOT a discount. Above both moving averages, RSI at 64, up 16.8% in three months. We are not buying a drawdown here; we are buying a re-rating that has already started, on the argument that at 9.3x FY2027 consensus with a 7.4% shareholder yield it has a long way to run before it is expensive.

That distinction matters for sizing and is stated explicitly. A Tactical position taken into strength should be smaller and should carry tighter kill criteria than one taken into weakness. The pre-registered kills in Section 11 are all credit metrics precisely because the technical setup gives no warning of a credit turn — charge-off rates lead stock prices in this business by roughly two quarters, and the chart will look fine until it does not.

10. Insiders — eight transactions, all of them open-market sales

Total: 21,639 shares, approximately $4.5 million, across four executive officers.

The reading, and it is the only genuinely negative item in the fundamental file. Every single transaction is coded S-Sale — an open-market disposal. There is not one F-InKind tax withholding, which is the mechanical, signal-free transaction type that dominates most insider files in this programme, and there is not one purchase.

Three qualifications, all of which reduce but do not eliminate the signal.

The amounts are small relative to the holdings. Ravi Raghu sold 9,726 shares across three tranches and retains 26,328 — a 27% reduction, the largest in the table. Matthew Cooper sold 7,000 across two dates and retains 90,194, a 7.2% reduction. Kaitlin Haggerty sold 1,426 and retains 49,181, 2.8%. Timothy Golden, the chief accounting officer, sold 3,487 and retains 7,429 — a 32% reduction and the smallest remaining position in the table.

The pattern is consistent with programmed selling. The Cooper sales are both exactly 3,500 shares, two months apart; the Raghu sales are three tranches on a single day at rising prices, which is a single order executing.

But three of the eight — and both of the largest reductions — were executed on 29 and 31 July, eight and ten days after a quarter that beat consensus adjusted earnings by 21.3%, at prices of $209 to $211. Selling into your own beat is not disqualifying and it is not evidence of anything specific. It is, however, the opposite of what a management team convinced the stock is at 9.3x a durable earnings stream would do, and it is reported as such.

What the file does not contain: any transaction by Richard Fairbank, the founder and chief executive, or by the chief financial officer. No purchase, no sale. On a name where the founder's alignment is a significant part of the qualitative case, the absence of both is neutral and uninformative.

11. Verdict, kill-criteria and flip conditions

Buy — Tactical.

This is the first Buy in this batch and we are going to be explicit about why the tier was reached rather than defaulted past.

The arithmetic. At $221.895 Capital One trades at 10.9x the FY2026 consensus of $20.44, 9.3x FY2027's $23.88 and 2.11x filing-verified tangible book value of $105.21, having just earned 18.04% on average tangible common equity in the June quarter. Our base fair value of $255 is 14.9% above spot with 2.46:1 asymmetry, and the company is returning approximately 7.4% a year to shareholders — $4,123M of stock repurchased in the six months to June 2026 alone, 3.0% of the market capitalisation, plus a dividend running near $3.24. The expected return decomposes as roughly half earnings growth and roughly half shareholder yield, with NO multiple expansion required. A return that does not depend on the crowd changing its mind is a materially better-quality return than one that does.

The trajectory. Total net revenue +27% year on year; net interest margin 8.01% and rising; efficiency ratio 57.05%; Domestic Card net charge-offs DOWN 49 basis points to 4.71%; a $662M reserve release; return on tangible common equity from −32.99% to +18.04% in twelve months as the $8.8 billion day-one Discover provision cleared. A 13.7% CET1 ratio, $484.3B of deposits funding 71.9% of $673.8B of assets, and tangible book value per share up 5.9%. And the strongest technical configuration in this batch: above both moving averages, RSI 64, MACD +4.42.

Why Tactical and not Core, stated as four specific reservations rather than as hedging.

First, we have no variant perception. Our base of $255 is within 0.4% of the street's $256, with 36 of 57 analysts at buy. We are not seeing anything the market is missing; we are taking a position on a favourable return shape at a modest price. That is precisely the distinction between the two Buy tiers.

Second, the knowledge base is negative. The only name-level claim on Capital One in 51,928 is a short recommendation from 2023 — wrong so far, but the only one there is — and the only outside read on the $4.5 billion asset Capital One bought four months ago says a competitor is pulling ahead.

Third, the macro overlay is genuinely unresolved. Five independent channels have warned on consumer-credit deterioration through 2025-2026. They have been wrong at Capital One specifically — card charge-offs fell 49 basis points — but the sub-portfolio they named is the one that deteriorated: Consumer Banking charge-offs rose 18 basis points to 1.48%.

Fourth, every insider transaction in the file is a sale, including two of the largest reductions executed eight and ten days after the beat.

And one more, which is not a reservation but a discipline: the entry is into strength, not weakness. The stock is up 16.8% in three months. That argues for a smaller initial position and a scaled build, not a full weight at once.

Pre-registered KILL criteria — what would take this to Watch or Avoid. Every one is a credit metric, because credit is the only variable that determines this outcome:

• Domestic Card net charge-off rate above 5.2% in any quarter — back to where it was a year ago, and the end of the improvement that underwrites the base case.
• A reserve BUILD replacing the release, i.e. provision exceeding net charge-offs, against the $23.4 billion allowance and 5.16% coverage.
• Consumer Banking net charge-offs above roughly 1.8%, extending the 18-basis-point deterioration already visible in the segment the knowledge-base bears named.
• CET1 below 13.0% with the buyback slowed or suspended. The buyback is roughly half the expected return; losing it changes the arithmetic more than losing a quarter of earnings would.
• Any disclosure that the Discover integration is running behind, or a material weakness in internal controls now that Discover is inside the assessment scope.
• A Brex impairment, or the disclosure of Brex revenue at a level that makes $4.5 billion plus $1.1 billion of debt repayment look expensive.
• A price break below the 200-day moving average of $207.45 on company-specific news rather than market beta.

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

• The Global Payment Network given separate disclosure with volume, revenue and margin economics. That is the single fact that would convert Capital One from a lender that owns a network into a network that also lends, and it is worth a multiple, not just an estimate.
• Two more quarters of Domestic Card charge-offs below 4.75% with reserve releases, which would settle the credit question the knowledge base's macro lane raises and remove the principal reason this is Tactical.
• An efficiency ratio below 55%, demonstrating the merger operating leverage the FY2027 consensus requires but has not yet shown.
• Insider buying, or at minimum the cessation of the current selling pattern.
• Name-level knowledge-base conviction turning positive. At present the only claim is a stale short call.

Where COF fits in the Synthos Framework Portfolio. The financials sleeve, entering at 1.5% today with a 3% target scaled on either a pullback toward the 200-day at $207 or a second consecutive quarter of improving card credit. Sizing note, and it is the operative discipline: this is a levered bet on the US consumer credit cycle, and the correct response to a 2.46:1 payoff ratio with an unresolved macro overlay is a position that can be doubled on confirmation, not one that has to be defended on a surprise. Logged as a tracked Synthos call (Buy — Tactical) as of 2026-08-04 at $221.895, with the fair-value anchors, the kill criteria and the upgrade conditions all gradeable.

Single biggest risk: the US consumer credit cycle. Credit Card is 74.2% of total net revenue and Domestic Card charges off at 4.71% on a book that is the majority of $275.4 billion of loans. A 200 basis point rise in that rate is roughly $4-5 billion of additional annual pre-tax provision, against an FY2027 consensus implying roughly $14-15 billion of pre-tax income. That is not a tail scenario — it is roughly the distance between today's rate and where this portfolio has been twice in the last two decades. The 5.16% allowance coverage and the 13.7% CET1 ratio mean Capital One survives it; the equity at 9.3x forward earnings does not survive it without a 30-40% drawdown. The knowledge base's macro lane says that turn is coming and has said so since 2023; the company's own most recent disclosure says the loss rate fell 49 basis points. We are siding with the disclosure over the narrative, and the kill criteria above are what we will change our minds on.

Provenance & disclosures

• Traceability: 1 name-level knowledge-base claim names Capital One out of 51,928 distilled claims, and it is BEARISH (raw hits 64, surviving a case-sensitive entity re-run 5, entity matches 4, name-level 1, discarded 59; breadth 5, net conviction negative-low). The claim: warren_pies, 2023-01-19, bearish, conviction 72, no named speaker — "Short card issuers that carry balance-sheet/credit risk — Discover weak after results, Capital One worse impacted — as consumer credit deteriorates." It has been decisively wrong (Capital One acquired Discover and the stock is materially higher) and is reported and graded as such rather than excluded. Three of the five surviving claims concern Brex, the corporate-card fintech Capital One acquired for $4.5 billion on 2026-04-07: empire 2026-04-03 (conviction 70, "Ramp leaving Brex behind" — four days before the deal closed, and the only outside read we hold on the acquired asset), invest_like_the_best 2026-03-17 (conviction 70, positive on the category), and no_priors 2024-08-08 (conviction 55). The fifth is all_in 2025-05-09, a text-only sector claim using Capital One as a reference point in a price-to-book spread against Credit Acceptance. The 59 discards are NOT homograph noise in the way the "Cash App" collisions on APP are — they are thematically relevant macro claims on consumer credit that fail the entity test, drawn from 27 channels including forward_guidance, luke_gromen, lyn_alden, andreas_steno and eurodollar_university (bearish on consumer-credit deterioration, 2023-2026) against compound_and_friends (2026-01-30, "not yet a market story"). That disagreement bears directly on the only variable determining this company's earnings and is an explicit input to the Tactical rather than Core designation, though it enters no conviction score. All 64 hits carry speaker: null and are channel-attributed — the weakest sourcing the 4-lane policy admits. No claim carries a management speaker_role. No concentration test is meaningful across five claims from five channels. All quotes are verbatim from the stored claim text.
• Data as-of: fundamentals — balance sheet, segment results, credit metrics and capital ratios through 2026-06-30, all filing-verified from the 10-Q filed 2026-07-28; second-quarter results, adjusting items and business metrics from the 8-K filed 2026-07-21, which carries the full earnings release as Exhibit 99.1; annual figures for FY2025, FY2024 and FY2023 from the 10-K filed 2026-02-19 · estimates 2026-08-04 · prices 2026-08-04, quote timestamp 1785873598 = 2026-08-04T19:59:58Z ($221.895, +1.94%; 50-DMA $198.37; 200-DMA $207.45; RSI 64.0; MACD +4.42) · knowledge-base claims 2026-08-04. Capital One's fiscal year is the calendar year. All figures come from the Synthos vendor data file for COF or from the SEC filings in the COF archive; no figure comes from memory, recall or external retrieval.
• Filing archive contents: 10-K filed 2026-02-19 (fiscal 2025 — the source of the $8.8 billion day-one allowance disclosure and the total-net-revenue series); 10-Q filed 2026-05-07 (March 2026 quarter); 10-Q filed 2026-07-28 (June 2026 quarter — the source of the Brex purchase terms, segment results, credit metrics and capital ratios); 8-K filed 2026-06-15 (no extractable item content); 8-K filed 2026-07-21, which carries the complete second-quarter earnings release with the adjusting-items bridge, net interest margin, efficiency ratio and business-metric detail; 8-K filed 2026-07-21 (second filing, no extractable item content). All carry preserved [TABLE] statement data.
• Where the filings contradicted or corrected the vendor (detailed in Section 8): the revenue basis — the vendor reports FY2025 revenue of $69,252M where the 10-K states total net revenue of $53,434M, a 29.6% gross-up reconciling exactly through the vendor's own interest-expense line, with the same defect running through inc_q (June quarter $19,861M vendor against $15,850M published) and invalidating priceToSalesRatioTTM, netProfitMarginTTM and every margin ratio; the $4.5 billion Brex acquisition of 2026-04-07, absent from every vendor field including bal_a (which ends 2025-12-31) and cf_a; acquisitionsNet of +$16,465M for FY2025, a positive number in the year of the Discover merger, representing cash acquired rather than consideration paid; the entire enterprise-value and invested-capital framework, structurally inapplicable to a deposit-funded bank and demonstrated void by the vendor's own investedCapitalTTM of −$334,028M producing a −3.87% return on invested capital for a company earning 18.04% on tangible common equity; seg_prod, which ceases to be a segment table after FY2023 and reports fee-revenue disclosures for FY2024 and FY2025, leaving the payload with no segment data covering either acquisition; seg_geo, which stops at FY2023 and contains a corrupt FY2020 international line of $12,341M against $1,376M in FY2022; bookValuePerShareTTM, which does not deduct the 4,980,700 preferred shares and overstates common book value by roughly 3.4%; and est.ebitdaAvg/ebitAvg, rejected for a fixed-ratio signature (exactly 27.85% and 19.98% of revenue in every forward year), for netIncomeAvg exceeding ebitAvg by FY2029, and for FY2022/FY2023 ebitAvg values three orders of magnitude wrong. Where vendor and filing AGREED — recorded, and one of them was a directed test: the book-versus-tangible-book test PASSED — the vendor reports bookValuePerShareTTM $183.389 against tangibleBookValuePerShareTTM $106.124, properly distinct, against the filing's $105.21 at 2026-06-30, a 0.9% difference (we use the filing); the share count is an exact match (10-Q 613,484,836 outstanding at 2026-06-30 against 613,485,000 implied by market capitalisation ÷ price); effectiveTaxRateTTM of 21.9% reconciles to the quarterly provisions; and quote.yearHigh/yearLow ($259.64/$174.24) sit within 1.1% of tech.hi52/lo52 ($257.94/$176.10), with tech used throughout for internal consistency.
• Basis note: analyst estimates are on adjusted diluted EPS, which excludes acquisition amortisation, Discover integration expenses and Brex integration expenses. The June 2026 bridge is disclosed cleanly in the earnings release: GAAP $4.73 plus acquisition amortisation $0.60 ($494M pretax), Discover integration $0.36 ($298M) and Brex integration $0.12 ($96M) equals adjusted $5.81 — a 22.8% adjustment, of which two of three components are transitional. Forward multiples in this dive use consensus adjusted EPS; the trailing multiple of 13.1x is computed on the vendor's trailing per-share figure and is itself distorted by the $8.8 billion day-one Discover provision in the FY2025 base. Total net revenue throughout means net interest income plus non-interest income, as Capital One publishes it, never the vendor's gross-up.
• Estimate coverage: 14 analysts on both FY2026 and FY2027 adjusted EPS — the anchor for all three fair values, with an FY2027 range of $22.693 to $24.898 (9.7%); 9 on FY2028; 3 on FY2029, which is excluded from every conclusion.
• Peer note: the vendor peer set — Bank of America, Robinhood, Progressive, Santander, S&P Global, TD, UBS — contains no credit-card issuer at all, which for the largest card-led bank in the United States is a material gap; the relevant comparison set (American Express, Synchrony, Bread, and formerly Discover) is absent from this file. No peer-multiple comparison is drawn.
• Fair-value caveat: the $180 / $255 / $325 anchors are multiples of the FY2027 consensus adjusted EPS distribution — 7.9x the low of $22.693, 10.7x the mean of $23.880, and 13.1x the high of $24.898 — each cross-checked against the filing's tangible book value of $105.21 per share (1.7x, 2.4x and 3.1x respectively). Stated arithmetic, not a discounted cash flow. The base is sensitivity-disclosed: 9.5x gives $227, essentially spot; 12x gives $287. The base case assumes NO multiple expansion; the strategic re-rating from owning a payments network is free optionality in the bull case, not a requirement of the base.
• Bank-metric note: cash-flow-based measures (freeCashFlowYieldTTM 23.65%, incomeQualityTTM 2.99, price-to-free-cash-flow 8.5x) are not used anywhere in this dive, because for a lender operating cash flow includes changes in loan and deposit balances and bears no relation to distributable earnings. The measures used are CET1 (13.7%), return on average tangible common equity (18.04%), tangible book value per share ($105.21, +5.9% year on year), net interest margin (8.01%), efficiency ratio (57.05%), net charge-off rates (3.23% total, 4.71% Domestic Card) and allowance coverage (5.16%) — all filing-sourced.
• Timing: second-quarter 2026 results were released 2026-07-21, fourteen days before this dive, and beat the consensus adjusted EPS estimate by 21.3% ($5.81 against $4.79). The next print is 2026-10-20, 77 days away. The most recent insider filings are dated 2026-07-30 and 2026-08-03, one and four days before this dive, and both are sales. 2026-08-04 carried no company-specific filing.
• 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.