Fit

Fit

Outside the framework's universe (U2 not met); contested: P4a

Market capitalisation is $8.48 billion on the filed share count, $7.44 billion on the deterministic feature basis, against a universe line of $10 billion. Confidence: high — two model families agreed, the trial was order-stable, and load-bearing spreads were at most 0.15. No exclusion hit, no China sensitivity flag, no name-mask divergence. One criterion is contested: P4a.

The framework's own order of operations is what produced that answer: U2 not_met -> out_of_universe. Analysis continues below on every other pillar, because the miss leads the report rather than ending it.

Universe and exclusions

Here is the decisive point. The universe screen is not a valuation judgment and not a quality judgment; it is a size line, and Oscar is 15% to 25% underneath it depending on which share count is used.

Oscar's Class A common stock is registered under Section 12(b) and trades on the New York Stock Exchange as OSCR; the issuer is incorporated in Delaware, commission file 001-40154 [1]. U1 (listing) is met, four votes to nil. Nothing in the record touches the Chinese-ADR exclusion.

U2 (scale) is not met, four votes to nil, and the miss survives every share convention available in the filings.

Sources: derived — 262,388 thousand weighted-average basic shares from the deterministic feature file; 263,552 thousand Class A plus 35,591 thousand Class B outstanding at 31 March 2026 [2]; 265,530,268 Class A plus 35,591,356 Class B at 10 April 2026 [3]; all at the 27 July 2026 close of $28.34.

The arithmetic: 262.388 million weighted-average basic shares at $28.34 gives $7,436.1 million, 25.6% short of the line. The 299.143 million shares actually outstanding at 31 March 2026 give $8,477.7 million, 15.2% short [2]. The most generous count in the record — 329.751 million Q1 2026 weighted-average diluted shares — still reaches only $9,345.2 million, $654.8 million short. Reaching $10 billion on the filed count requires $33.43 a share, 18.0% above the 27 July 2026 close.

The strongest counter-fact sits inside the same treatment: the feature file's own basis understates the company by roughly $1.0 billion, because it uses the FY2025 weighted-average basic count rather than shares outstanding. Correcting it narrows the gap from 25.6% to 15.2% — and independent market-cap trackers reported $8.25 billion to $8.76 billion across June and mid-July 2026, consistent with the corrected middle case. The line is still not cleared on any basis, which is why all four jurors recorded the same verdict.

Exclusions: a clean screen. All four disqualifying checks and the sensitivity check were run against primary evidence and none hit.

• X1, auto OEM — not hit. Health insurance premium is 98.02% of FY2025 revenue ($11,469.9 million of $11,701.4 million), investment income 1.73%, other revenues 0.24% [4]. No automotive line item exists anywhere in the filing.
• X2, promotional CEO — not hit, on one prong only. The exclusion requires a repeated promise-versus-delivery gap and weak insider economic ownership. Prong one is evidenced and stated plainly: February 2025 guidance of $225–275 million of earnings from operations became a $396.4 million operating loss [5], and the 2027 target of a 5% operating margin, carried in prepared remarks on six consecutive calls and defended under questioning on 6 August 2025 [6], was absent from both subsequent calls without restatement or withdrawal. Prong two fails outright: the chief executive beneficially owned 11,925,092 Class A shares at 10 April 2026 against FY2025 total compensation of $1,149,308 with nil stock awards and nil option awards [3]. Without both prongs, the framework records no hit.
• X3, structural decline — not hit. Total revenue rose in every year from FY2021 to FY2025, reaching $11,701.4 million [4], and FY2026 is guided to $18.7–19.0 billion, up 61% at the midpoint [7]. Consecutive years of decline: zero, against a three-year disqualifier. The trial's temporary probability of 0.71 is far above the 0.35 strongly-permanent trigger.
• X4, consensus-saturated story — not hit. Oscar trades at 0.64x FY2025 revenue on the feature basis and 0.39x guided FY2026 revenue; the sell side carries no buy rating in eleven recommendations (3 outperform, 7 hold, 1 underperform) and a mean target of $24.20 against the $28.34 close. The counter-fact in the same breath: the chart shape does carry the darling signal — up 161.2% from the 30 March 2026 trough in 119 days — and Oscar is the second most expensive of seven managed-care names on price-to-sales.
• S1, China dependence — not flagged. China revenue $0 of $11,701.4 million; China assets $0 of $9,289.6 million; all 2,042,449 members at 31 December 2025 sit in 18 US states; zero occurrences of "China" or "Chinese" in the FY2025 Form 10-K.

Pattern match

This is the framework's third pattern — healthcare and insurance forecasting errors — and it fits the pattern's specific checks closely enough to name it. That is framing, not a verdict; the pillar results below are unchanged by it.

The pattern's first check is a one-year cost-trend or risk-adjustment misforecast rather than a franchise problem. Oscar's FY2025 swing was $646.4 million against its own February 2025 midpoint, and 85.5% of it is one line: risk-adjustment transfers stepped from 14.52% of direct and assumed premiums in 2024 to 18.45% in 2025, which cost $552.5 million on the 2025 premium base [8]. The volume line never broke: revenue guidance was raised at the same moment earnings were cut, from $11.2–11.3 billion to $12.0–12.2 billion [9].

The second check is the market taking the stock down roughly one-for-one with the guidance cut as if it were permanent. Here it went further than one-for-one: over the thirteen sessions from 30 June to 18 July 2025 the shares fell 37.4%, removing about $2.10 billion of market value against a $500 million midpoint guidance swing — 4.2 times the guided number and 3.3 times the realised full-year shortfall.

The third check is the repricing mechanism and its regulatory friction. Participation in each individual market is an annual election and premiums require state and federal approval [10]; Oscar refiled 2026 rates in states covering close to 99% of membership at an approximately 28% weighted average increase [11]. The ACA's minimum-MLR provision requires rebates when medical costs fall below the specified threshold, which caps underwriting margin by statute in both directions [12].

Where the pattern breaks is the entry moment. The precedent the pattern is built on involved buying the anchored price; at $28.34 the shares sit 21.8% above the $23.27 close the drawdown started from, and the first post-repricing quarter has already printed — a 70.5% medical loss ratio and $704 million of earnings from operations [13]. The full anatomy is in Dislocation and Clock.

The pillar ledger

Source: the run's deterministic fit tally; per-criterion verdicts, vote splits, trimmed-mean probabilities and spreads as recorded. Reference lines, not scores.

Year-10 gate (P1) — not met

Not met, four votes to nil, trimmed-mean probability 0.385 with a spread of 0.11. This is the framework's one binary criterion, and genuine doubt resolves downward by its own construction.

None of the five conviction sources applies in a form that would underwrite higher revenue and higher adjusted free cash flow in 2036. Approximately 98% of revenue comes from ACA-regulated plans and 93% of FY2025 premiums were earned directly from CMS, with 7% from members [10]; approximately 97% of FY2025 direct policy premiums were subsidised by advance premium tax credits, and the enhanced credits expired at the end of 2025 [14]. Market structure is not a monopoly, duopoly or stable oligopoly: the 10-K names four categories of competitor, including local Blue Cross plans [15], and every carrier re-elects its footprint annually [10]. Capital intensity is 1.01% of assets (property, equipment and capitalised software of $94.2 million against $9,289.6 million) [2], so there is no replacement-cost moat; statutory capital and surplus is a reserve, and it fell to approximately $1.0 billion at 31 December 2025 from $1.2 billion a year earlier while revenue grew 27.5% [16]. Operating history runs 14 years against the framework's 30-to-50-year anchor, with one profitable year in it and an accumulated deficit of $3,294.4 million [17].

The deciding arithmetic is a reversion sensitivity. Management-reported footprint share moved from approximately 10% in plan year 2020 [18] to 17% in 2025 and 30% in 2026 [19]. Scaling the FY2026 revenue guidance midpoint of $18.85 billion back to the 2025 share level (17/30 = 0.5667) gives $10.68 billion — below FY2025 actual revenue of $11.70 billion, and 43.3% below the guided level, with no operating failure at Oscar required. A 13-point single-year share move is not the signature of a protected position.

The strongest surviving counter-fact belongs in the same treatment: Oscar grew straight through the shock the doubt is built on. Membership reached 3.2 million at 31 March 2026, up 56% year on year [20], and the addressable market contracted only about 5% at open enrollment — 23 million lives against a record 24 million [21] — against management's own planning assumption of a 20% to 30% contraction [7]. Essentiality of the product is evidenced rather than asserted. What the resilience does not establish is durability of the subsidy that sizes the market. Full treatment in Durability and Business.

Consistency (P2) — not met

Not met, four votes to nil, no probability recorded. The deterministic test could not be run: fit_features.fcf_stability and fit_features.adjusted_fcf are both not_computable, so the series below was rebuilt from the filed cash-flow statements using the framework's definition unchanged.

Source: derived — adjusted FCF = operating cash flow less capital expenditure less stock-based compensation, with no acquisitions line disclosed; from the FY2025 Consolidated Statements of Cash Flows [22] and the FY2022 statements for the earlier years [23].

The deciding arithmetic: the two available rolling five-year averages are $100.2 million (FY2020–FY2024) and $259.8 million (FY2021–FY2025), a move of 159% between adjacent windows. Adjusted FCF changed sign in four of the seven years on file. The framework tolerates a negative episode every five to eight years as part of an insurer's cycle; Oscar has had one every other year. Only two windows exist at all, which is too few to characterise stability with confidence in either direction — that limitation cuts both ways and is recorded as a data gap.

The counter-fact, and it is a real one: the risk-adjustment payable build is partly genuine insurance float rather than a timing trick. Oscar paid $1,611.7 million of prior-year transfers during 2025 and still closed the year with a $2,587.7 million gross payable, and reported operating cash flow was positive and rising — $978.2 million in FY2024, $1,094.9 million in FY2025 [22]. The independent check applied to this claim revised one figure: stripping only the risk-adjustment payable build leaves four of six years negative, not five, with a cumulative -$817.3 million. The no-stable-base conclusion held. Full treatment in Yield and Durability.

Dislocation and yield (P3a, P3b, P3c, P3d)

P3a, identifiable event — met, four votes to nil. The trigger is dated twice. On 1 July 2025 Centene disclosed that first-look Wakely Marketplace data implied market morbidity far above its risk-adjustment assumptions and withdrew 2025 guidance; the ACA-exposed group repriced the next day, and Oscar closed down 18.7% on 78.0 million shares, 29.1 times its pre-peak median daily volume. Three weeks later, on 22 July 2025, Oscar filed an 8-K revising its own full-year outlook [24], to revenue of $12.0–12.2 billion and a loss from operations of $200–300 million against original guidance of $225–275 million of earnings [9]. The counter-fact: the fall is not one event's work. The first 21.2% leg, from 19 September to 5 November 2024 over 33 sessions at 1.2x normal volume, came with no company disclosure attached — precisely the drift the framework excludes as an entry moment.

P3b, capitulation — met, four votes to nil. The measured 20-session volume spike is 15.47x the trailing median (41.5 million shares against 2.68 million), against a reference line of 2x, with single sessions at 30.8x, 29.1x and 24.4x. The counter-fact sits in the same entry: the spike window ended 29 July 2025 at $13.84, already 40.5% below the peak, and the price fell a further 21.6% over the following eight months to the $10.85 low of 30 March 2026. Peak emotional selling did not mark the price low.

P3c, yield versus bar — not met, four votes to nil. Oscar classifies fortress on the framework's own rule (net debt of -$2,344.1 million: $430.1 million of long-term debt against $2,774.2 million of cash) [25], which selects the 8.5% reference line rather than the 10% default.

Source: derived — adjusted FCF from the filed cash-flow statements [22] over the $7,436.1 million market capitalisation of record; fit_features.adjusted_fcf_yield and fit_features.yield_baseline are both not_computable for this run.

The FY2025 reading of 13.06% clears the line by 456 basis points, and it is the number that does not survive inspection. In plain terms it is not earnings; it is working-capital float. The CMS risk-adjustment payable rose $1,029.4 million in the same year — 106% of the $970.8 million numerator — while the company lost $443.2 million; strip working capital entirely and adjusted FCF is -$488.2 million [22]. The three-year average is 6.07%, 243 basis points short of the fortress line and 393 short of the default; the five-year average is 3.49%. Normalised for the underwriting cycle at a 3.5% mid-cycle operating margin on the guided FY2026 revenue base, adjusted FCF is $586.3 million and the yield 7.88% — 62 basis points short, with a 4.27% to 11.50% band across a 2% to 5% margin. The jurors recorded the position two ways, at -243 basis points on the three-year average and -62 on the mid-cycle base case; both sit below the line. There is also no stable historical baseline to have jumped from: annual readings run -20.9%, +45.7%, -22.6%, +23.5% and +25.8% across FY2021–FY2025.

The counter-fact, which the independent check confirmed and which narrows the finding: the window is decisive. The adjacent FY2024–FY2025 average is $905.7 million, or 12.18%, above the line; the FY2021–FY2025 average is 3.49%, well below it. The three-year window is dominated by FY2023, when adjusted FCF was -$457.4 million at roughly half the current revenue scale.

P3d, forward path — met, four votes to nil, probability 0.565 with a spread of 0.02. The vendor consensus free-cash-flow line is not usable as an anchor: it reports no contributor count, does not reconcile to its own cash-from-operations less capex in any year, and missed the now-actual FY2025 by 43.7% ($596.3 million against $1,058.5 million reported). The reconciled net-income proxy gives 5.35% for 2026, 7.38% for 2027 and 9.94% for 2028 — clearing the 8.5% line in 2028 by 144 basis points on the market capitalisation of record, and by 22 basis points on the filed share count. The counter-fact: the repricing that would carry that path is already in force rather than forecast, with Q1 2026 delivering a 70.5% medical loss ratio and the lowest administrative expense ratio in company history at 15.2% [13]. Full treatment in Yield; the drawdown anatomy is in Dislocation.

Balance sheet and self-help (P4a, P4b, P4c)

P4a, outlast and allocation headroom — contested, two votes to two, split by model family. Both readings are set out in Contested and undetermined below.

P4b, repurchase engine — not met, four votes to nil, on the criterion's own hard fail. Oscar has never repurchased a share as a public company: Item 5 of the FY2025 10-K reports issuer purchases as "None" [26], no repurchase line appears in the FY2023, FY2024 or FY2025 financing activities [22], no authorisation appears in any of the FY2021–FY2025 10-Ks, and the words "repurchase" and "buyback" do not occur once across the twelve earnings calls in the corpus. The share count moved the other way: 250,488 thousand shares outstanding at 31 December 2024 became 297,689 thousand a year later, up 18.8%, with 33.1 million of that from fourth-quarter 2025 conversion of $270.0 million of notes at an approximately $8.32 conversion price [27]. fit_features.share_count_trend.rising is true, with a 55.1% five-year compound rate. The framework's levered exception does not apply either: FY2025 adjusted FCF of $970.8 million against roughly $8,533.8 million of market value is an 11.4% yield against the exception's approximately 25% requirement, and the second leg — a demonstrated multi-year share-count reduction — fails outright.

The counter-fact: the company is not indifferent to dilution. It spent $34.4 million on capped call transactions in September 2025 specifically to limit dilution from the 2030 notes, capped at $37.46 a share. That is dilution management, not retirement. The absurdity check does not fire either: at the current price it takes 8.8 years of reported adjusted FCF to retire the float, against the framework's roughly three-year reference for a price that cannot survive.

P4c, dividend cover — not applicable, four votes to nil. Oscar has never declared or paid a cash dividend [28], the yield is 0.0% against the framework's approximately 4% materiality threshold, and the February 2026 credit agreement's negative covenants restrict distributions on equity interests [29]. No dividend forms part of the return case. Full treatment in Self-Help.

Diagnosis (P5) — met

Met, four votes to nil, probability 0.71 — carried from the adversarial trial rather than re-derived by the jury. Three independent judges returned 0.71, 0.66 and 0.72; the mean is 0.697, the spread 0.06, and the ruling was not contested. Reading order barely moved it: temporary-first briefs averaged 0.71 and permanent-first briefs 0.69, a gap of 0.02.

The temporary reading rests on the shape of the damage. The FY2025 hit was a margin event on an intact revenue base: a $646.4 million one-year swing of which 85.5% traces to risk-adjustment transfers stepping 393 basis points as a share of direct and assumed premiums [8], against revenue that grew 28% to $11.7 billion in the same year [21] and is guided up 61% for FY2026 [7].

The strongest surviving counter-fact is that the gap the framework hunts has closed. The run's two-scenario discounted model, weighted at the trial's 0.71, returns $19.99 a share; at the 30 March 2026 trough close of $10.85 that was a $9.14 gap, and at the 27 July 2026 close of $28.34 the price sits 41.8% above it. The independent check weakened this claim rather than confirming it outright: at a 10% discount rate the same model returns $26.01 and at a 5% steady-state operating margin it returns $29.57, essentially the current price. The supportable statement is therefore narrower — the price now embeds management's pre-event plan, not that the plan is unachievable. Management's 5% margin target dates from the June 2024 investor day [30] and was still described as live in August 2025 [6]. Both cases are presented in full in Damage Math.

Eight of the trial's quote checks failed on page attribution: the cited substance was verified verbatim one to four pages away in every case, so the judges treated them as pagination errors rather than fabrication and the ruling stood.

Instrument context (I1) — not verifiable

Not verifiable, four votes to nil. The docket's only entry asserts sixteen listed expiries out to 21 January 2028 (17.8 months) with 130,498 contracts of open interest in the two expiries beyond twelve months, and 30-day implied volatility of 86.6% against the framework's reference lines of up to approximately 55 acceptable and 60 to 70 elevated. Every one of those facts comes from web sources dated 27 July 2026 with no corpus page behind them, so the criterion is recorded as not verifiable and nothing is established either way — neither that qualifying long-dated options exist nor that they do not. The watchlist_only flag is not raised, because the framework attaches it only to fitting or lean-fitting outcomes. Instrument facts, as facts only, are in Clock.

What a 3x-in-3-years would require

The tally records no re-rating arithmetic: "Re-rating math unavailable because the applicable bar or normalized adjusted FCF is missing." Both inputs are not_computable in the deterministic feature file — yield_baseline, balance_sheet_class and adjusted_fcf all return no value — so the framework's target test cannot be stated from the tally. What follows is the same test built from the surviving claims' figures instead, labelled as such.

The applicable line is 8.5%, selected by the fortress classification derived from the filed balance sheet [25]. Mid-cycle normalised adjusted FCF is $586.3 million. At the line, that supports $6,898 million of market value — $23.06 a share on the 299,143 thousand shares filed at 31 March 2026, or 18.6% below the 27 July 2026 close. On normalised cash flow the framework's own bar sits below today's price rather than above it.

Running the test the other way: a 3x from $28.34 is $85.02, or $25,433 million of market value on the filed count. At the 8.5% line that requires $2,162 million of normalised adjusted FCF — 3.7 times the mid-cycle figure, 2.2 times the float-inflated FY2025 reading, and 9.7% of the FY2028 consensus revenue base of $22,226.6 million against the 3.3% the reconciled consensus proxy implies. Roughly three times the conversion rate consensus carries.

What consensus would have to concede is a much smaller number than that, and it is worth stating precisely: to reach a 10% adjusted yield by 2028 needs $743.6 million on the market capitalisation of record, only $4.6 million above the reconciled proxy, or $847.8 million on the filed share count — about $108.7 million more, which implies roughly a 4.6% operating margin on the 2028 revenue estimate against consensus's 4.11% and management's approximately 5% target [30]. The distance to the yield bar is small. The distance to a 3x is not the same question, and the arithmetic above is why.

The base rates make the timing point. Oscar's own listed history contains thirteen drawdown episodes of 35% or deeper in 1,357 trading days, roughly one every five months, with a median depth of -42.1% and a median 70 days from peak to trough. Twelve of the thirteen regained their prior peak, taking a median 170 days from the trough.

Source: derived from daily closes, 3 March 2021 to 27 July 2026, using a 35% swing-reversal segmentation; the current episode's peak and trough match the deterministic capitulation gauge.

The four prior episodes of 50% or deeper took between 166 and 1,772 days to regain their pre-drawdown peak, a median of 702. The current episode took 43 days from the 30 March 2026 trough to reclaim the $23.27 peak, six days after the Q1 2026 print — far faster than anything in the name's own record. That is the base-rate context for a framework whose instruments run 18 months and more: the re-recognition this pattern normally waits for has already been delivered, and the price is 21.8% above the peak the drawdown started from. This is arithmetic against the framework's reference lines, not a recommendation. The catalyst calendar is in Clock.

Contested and undetermined

P4a is contested, two votes to two, and the split runs cleanly along model families. Both claude seats recorded met; both codex seats recorded not met; the name-masked seat recorded not met. The two readings share every fact and differ on what "comfortably outlast without pivoting" requires.

The met reading. Debt is $445.0 million of principal — $410.0 million of 2030 notes at 2.25% and $35.0 million of 2031 notes at 7.25% — with nothing contractually due before a $35.0 million holder put on 30 June 2027 and no maturity before 2030. Cash interest paid in FY2025 was $12.8 million against $5,461.6 million of consolidated cash and investments [25]. On those numbers a pivot to debt paydown cannot be forced on this company at the moment repurchases would matter most, which is what the criterion asks.

The not-met reading. The cash that makes Oscar solvent is not available for capital allocation. Only $414.2 million of cash and investments sat outside the Health Insurance Subsidiaries at 31 December 2025, of which $14.7 million was restricted, and subsidiary excess capital over the minimum risk-based capital requirement fell from $734 million at the end of 2024 to approximately $315 million at the end of 2025 [31]. Parent cash fell further, to $279 million by 31 March 2026. The February 2026 revolving facility pledges substantially all assets as collateral and its covenants restrict distributions on and repurchases of equity interests, subject to limitations and exceptions, while requiring at least $200.0 million of liquidity plus undrawn commitments with at least $100.0 million in unrestricted cash at the company and guarantors [29]. Net of that reserve the free envelope is approximately $299.5 million, about 3.5% of market capitalisation, against roughly $853 million to add ten points to earnings per share.

The counter-fact that cuts toward the met reading: by the Q1 2026 call insurance-subsidiary capital and surplus had risen to approximately $1.7 billion including $809 million of excess capital, and total cash and investments to approximately $8.1 billion, so the December 2025 squeeze looks like a trough rather than a trend.

Because the tally records P4a as contested, no reading here is promoted to a verdict. The gate that produced the overall answer is U2, and it is not affected either way.

Nothing was recorded as cannot-determine. No criterion carries that verdict, so no named missing datapoint gates the answer. Two criteria are absent for structural reasons rather than uncertainty: P4c is not applicable because there is no dividend, and I1 is not verifiable because no citable option-chain source sits in the corpus. Both are stated above.

One further split worth naming, because it looks like disagreement and is not. The tally records no cross-family agreement on X1, X2, X3, X4 and S1. The codex seats labelled these checks "not hit" and the claude seats "not met" — the same finding under two labels, with identical evidence cited on both sides. The masked seat matched on all five.

Provenance

Source: the run's deterministic fit tally and refutation ledger, as recorded.

Two sentences on what that means. The verdict was reached by four jurors reading the same evidence dockets independently, two from each of two model families, plus a fifth juror shown the same dockets with the company's name removed — and the masked juror reached the same gate conclusions, which is why the run carries no prior-driven-risk flag. Every claim the verdict turns on was handed to a skeptic that recomputed the arithmetic from the cited pages: three claims came back weakened and are reported here in their weakened form, four could not be verified at all and carry no weight, and none was refuted.

The falsifier ledger

These are the standing conditions that would change the read. The ledger carries seventeen entries; several are the same test nominated by more than one tab, and each is reproduced exactly as recorded.

Framework templates.

1. adjusted FCF or EBITDA declines where flat-or-better was underwritten
2. revenue declines for a third consecutive year
3. capital allocation pivots to debt paydown over repurchases
4. share count inflects upward
5. the industry repricing cycle fails to materialize where industry-wide mean reversion was underwritten

Name-specific, with thresholds and windows as recorded. Reproduced verbatim from the run's ledger, in the ledger's own wording.

Note on entry 11 of that block: the ledger recorded its source reference in raw internal form and it is reproduced unaltered. It resolves to the Q1 FY2026 Form 10-Q, page 37, which carries the eAPTC expiration, Program Integrity Rules and OBBBA discussion [32].

Three of the anchors those thresholds hang on are worth pinning to their pages: the approximately 20% risk-adjustment assumption and the 82.4% to 83.4% MLR guide are management's FY2026 guidance [33]; the 24% to 24.5% first-quarter accrual and the naming of the Wakely claims-based report as the determining data are from the Q1 2026 call [34]; and the $250 million to $450 million earnings-from-operations guide sits alongside the FY2026 revenue range [7].

Data gaps

The tally records sixty data-gap entries, most of them the same limitation reported independently by several tabs. Deduplicated, they are these.

Deterministic feature-file defects. fit_features.revenue_trajectory is built on the wrong income-statement line: it records FY2025 revenue of $28,593 thousand, which is Oscar's other-revenues line, not the $11,701,427 thousand of total revenue in the same filing [4]. The whole per-year series is unusable and every year in it is misaligned by roughly three orders of magnitude; the derived flags happen to survive the correction, since consecutive-decline-years of 0 and three-year-high-single-digit-decline of false are both still correct on the corrected series, but the inputs must not be quoted. fit_features.market_cap uses 262,388 thousand shares, the FY2025 weighted-average basic count from the earnings-per-share note, rather than shares outstanding — an understatement of 36.8 million shares against the 31 March 2026 balance sheet and of roughly 14% of market capitalisation, which flatters every yield computed on it; the U2 conclusion is unchanged under every convention, and this tab reports both bases. fit_features.adjusted_fcf, adjusted_fcf_yield, yield_baseline, fcf_stability, balance_sheet_class and float_retirement_years all return not_computable, because data/financials/cash_flow.json carries operating, investing and financing totals with no capex, stock-based compensation or acquisitions line. Every yield, stability and balance-sheet-class figure on this tab was rebuilt from the filed statements and is labelled a derivation, not a substitute. fcf_stability also returns an empty rolling five-year average, and even on the derived series only two adjacent windows exist — too few to characterise stability with confidence in either direction. float_retirement_years is not restated: on FY2025's float-inflated numerator the answer is 7.7 to 8.8 years and on the mid-cycle figure 12.7, but the FY2025 numerator is not repeatable.

Market, consensus and share data. No CapIQ consensus estimate rows exist in data/estimates (zero rows), so consensus positioning for X4 rests on the ratings distribution rather than on forward estimates; analyst counts, ratings and the mean target come from a market-data feed and dated web sources as at 28 July 2026, not from any filing. There is no consensus revision tape spanning the trigger event: the vendor momentum series reaches back only to 28 January 2026, so the FY2025 and FY2026 consensus levels immediately before and after the 22 July 2025 guidance cut cannot be read, and the numerator in the damage work is built on the company's own dated guidance record instead. The vendor free-cash-flow series is dominated by risk-adjustment payable build and settlement rather than owner earnings, so its implied forward yields of 46.2% for FY2026 and 22.6% for FY2028 are not underwritable and were not used; there is no direct adjusted-FCF consensus at all, and the FY2029 anchor rests on a single broker. Enterprise value is not separately derivable for an insurer of this structure, because consolidated cash and investments sit inside regulated subsidiaries against benefits payable and risk-adjustment payables; market capitalisation is used as the proxy throughout, with net parent debt of $152 million deducted in the discounted work.

Policy, calendar and market-share denominators. No 2027 weighted average rate increase has been disclosed as of 27 July 2026, so the forward leg of the repricing mechanism can be dated but not quantified. The City of Columbus v. Kennedy outcome and any renewal of the enhanced premium tax credits are undated, and the filing states the company cannot predict either; whether Congress reinstates the enhanced structure is not determinable from the corpus. No FY2026 fourth-quarter or full-year earnings date has been announced, so early February 2027 is the company's historical cadence rather than a scheduled event. No independent market-share data by state or rating area exists in the corpus: the 17%-to-30% footprint-share move is management-stated on the Q4 FY2025 call and uncorroborated by any filing, regulator document or third-party source [19]. Total individual-market enrolment appears only as management's characterisation of CMS data, so the denominator behind every national share calculation is second-hand, and the 2026 enrolment figures used in the durability work come from external published research. Oscar's 10-K names competitor categories rather than companies, so the peer set comes from the run's auto-generated screen plus each peer's own filings; Elevance's and UnitedHealth's individual-exchange membership are not disclosed in the sections read, so their share could not be quantified from primary sources — though Centene's 5.5 million Marketplace members across 29 states [35] and CVS's exit from every Public Exchange state Aetna operated in [36] are on the record.

Ownership, flow and instrument data. The short-interest feed returned zero rows, so the entire series used in the drawdown work is web-compiled FINRA data with no page citation. The Form 4 extract holds 365 transactions dated 2021 to 2023 and 2026 but none dated 2024 or 2025, so insider activity through the drawdown window cannot be established. Institutional holder-base changes are not resolvable: the ownership file records 13D and 13G filing dates with null holdings, and no 13F position history is staged. No index-deletion, fund-liquidation or block-trade disclosure appears anywhere in the corpus for the window, so forced selling can be neither evidenced nor ruled out. Implied volatility and option-chain facts come from web sources dated 27 July 2026 and carry no PDF page, which is why I1 is recorded not verifiable.

Documents absent from the corpus. Oscar's 22 July 2025 preliminary-results press release (Exhibit 99.1 to the 8-K) is not indexed — only the 8-K body [24] — so the revised guidance figures are cited from the Q2 FY2025 call, where management restates them verbatim. Centene's 1 July 2025 guidance-withdrawal release, the primary trigger document, is likewise absent and is sourced from Centene's own call. The 2026 credit agreement exhibit is parsed as page images with no extractable text, so the restricted-payments basket size and the exact leverage definitions could not be read; the covenant terms here come from the 10-K risk-factor summary. The November 2025 policy repricings are sourced to contemporaneous market reporting only.

Checked and absent, not unexamined. No share repurchase authorisation exists to assess, so there is no authorised-versus-executed gap to quantify — the finding rests on the absence of both. No buyback, repurchase or capital-return discussion appears anywhere in the Q3 FY2025, Q4 FY2025 or Q1 FY2026 transcripts. No 30-to-50-year operating history exists to test, since the company was founded in 2012 and the Marketplace itself dates from 2014, and the only macro shock in the record coincided with expanded subsidies and special enrollment periods, which confounds the test. And no corpus or web evidence was found of a technology or business model that would render individual-market health insurance obsolete over ten years: the substitution threat identified is competitive re-entry by withdrawn carriers, not technological displacement.