Fit

Fit

Does not fit the framework (P1 not met); contested: P5.

Penumbra clears the universe (NYSE common stock, ~$12.62B market cap) but fails the one pure gate: the year-10 durability test. The revenue leg of that gate is a high-conviction call, but the year-10 adjusted-FCF-higher leg carries genuine doubt — positive adjusted free cash flow is only three years old against three larger, consolidating competitors — and a signed Boston Scientific takeover at $374/share removes the standalone decade entirely. One exclusion also fires: the market-darling screen (X4). Confidence is low, because the diagnosis criterion (P5) is contested with a 0.26 probability spread, above the framework's 0.25 line.

The answer, in one screen

The gate decides it. The setup, the yield, the balance sheet and the diagnosis below are real analysis, and none of it offsets a failed P1 — that is the framework's own rule: P1 not_met -> does_not_fit (gate; nothing offsets it). What makes this name unusual is that the price is not in a dislocation at all: at $321.19 the stock sits 5.7% above its February-2025 peak and 14% below a $374 strategic bid, so a buyer today underwrites a merger-arbitrage spread, not a fear-repriced value gap.

Sources: market cap 39,291,828 shares × $321.19 (2026-07-31); adjusted FCF yield derived from the FY2025 statement of cash flows [1]; the trial probability from the report's adversarial diagnosis trial. Adjusted FCF = reported FCF − stock-based comp − 5-yr avg acquisition spend.

Universe and exclusions

Universe — clears both lines. Penumbra is ordinary U.S. common stock listed on the NYSE (ISIN US70975L1070), incorporated and headquartered in Alameda, California — not an ADR and not a Chinese issuer, so U1 passes. At the $321.19 close on 2026-07-31 the 39,291,828 shares put market cap at $12.62B, roughly 26% above the $10B floor, so U2 passes [2]. The pass is not comfortable: the cap sits only 26% over the line and rests on a price already lifted by the takeover bid — the undisturbed pre-announcement market cap (~$11.8B at $300.97 on 2026-01-13) was closer to the threshold.

One exclusion fires — the market-darling screen (X4). This is a genuine hit, and it leads. Penumbra carries darling positioning: ~9.0x standalone revenue and ~10.3x EV/revenue (~83x EPS at the deal price), a ~20% ten-year revenue CAGR with no down year, and a bottom-left-to-top-right chart that consensus and an acquirer both love [3]. It is not the most extreme multiple-to-sales name in premium medtech, and FY2024 GAAP net income was just $14.0M after a $76.9M impairment — but the screen the framework runs (a consensus-owned secular-growth story on premium multiples) is exactly what fires here.

The other four exclusions do not fire. Penumbra is a pure medical-device manufacturer — single-use vascular catheters and clot-removal systems sold to hospitals — with no automotive or OEM exposure, so the car-company exclusion (X1) does not apply. Management's record shows one FY2024 guidance miss, acknowledged and followed by a string of 2025 beats and raises, not a promotional serial-miss pattern (X2 does not fire), though founder-CEO Elsesser owns just 1.9% and insiders 2.9% with no open-market buying [4]. Revenue rose in all ten years on record, so the structural-decline exclusion (X3) is absent. China is served through a technology-licensing arrangement inside the ~10% distributor channel; total international revenue is $311.9M (22.2% of company), so the China-dependence sensitivity (S1) is a name to flag, not a dependence.

Pattern match

Of the framework's four setups, Penumbra is closest to a quality tech/medtech franchise on a fear dip — an essential, category-leading device maker whose stock fell 25.8% in 2025. But it fails that pattern's entry checks and reverses one of them. The 2025 fear-dip has fully round-tripped: the stock closed above its February peak by December and now trades near all-time highs, so the fear window is closed, not open. And the specific fear the pattern looks for — testable and temporary — is here entangled with a real growth-algorithm question (thrombectomy deceleration) rather than a passing scare. It is emphatically not the healthcare/insurance forecasting-error pattern (no 1:1 guidance-cut anchor; guidance was beaten, then withdrawn for the deal), not cyclicals-at-the-bottom, and not a monopoly with franchise pricing power. It fits none of the four cleanly.

The pillar ledger

Source: deterministic fit tally; per-criterion arithmetic from the surviving claim ledgers cited in each section below.

Year-10 gate (P1) — the decisive result

Not met. All four jury seats and the name-masked seat returned not_met; the trimmed-mean year-10 probability is 0.55 with a spread of 0.24 — cross-family agreement. The split inside the gate is the point. Year-10 revenue higher is a high-conviction call: revenue has risen for ten straight years, from $263.3M (2016) to $1,403.7M (2025), a ~20.4% CAGR into a still-mostly-untreated stroke market (~200,000 of ~700,000 annual U.S. ischemic strokes are thrombectomy-treatable, most untreated today) [5] [6]. What cannot be asserted with very high conviction is year-10 adjusted free cash flow: positive adjusted FCF is only three years old — $31.6M / $101.1M / $115.7M in 2023–2025 — against three larger competitors (Boston Scientific, Medtronic, Stryker now including Inari Medical, Terumo) that can outspend Penumbra on development and marketing [7] [8]. Penumbra's moat is IP, clinical evidence and execution over a ~22-year history rather than the dominant scale, regulatory-entry barrier, or capital intensity the framework prizes; execution is not a moat. Under the framework, any proper doubt resolves the gate to not met, and a signed $374 takeover means there is no standalone Penumbra equity to hold in year 10 regardless. Full treatment: the Durability tab.

Strongest counter-fact: gross margin ~67%, no debt, and operating income up from $6.1M (2022) to $189.2M (2025) argue a much larger business a decade out generates more cash, not less — the revenue leg is not in doubt, and consensus revenue still climbs toward ~$2.04B by FY2028. It moves the volume question; it does not settle the year-10 cash question with very high conviction, and no consensus estimate reaches year 10.

FCF consistency (P2) — not met

Not met. The deterministic fcf_stability feature is not_computable (fewer than five consecutive positive adjusted-FCF years; stock-based comp absent from the feed), so the test was run by hand on the filed statements [9]. Reported FCF was negative or near-zero in six of the ten years through 2022 (−$26.4M, +$0.2M, +$19.2M, +$4.5M, −$58.0M, −$11.7M, −$75.0M) before inflecting to +$82.1M / +$147.3M / +$174.9M in 2023–2025. A stable rolling five-year average has not yet formed. Counter-fact: the inflection is mechanistically durable — revenue scaling over a largely fixed cost base drove the operating-income jump — and it is a one-directional growth-investment curve, not the every-5-to-8-years underwriting dip the framework tolerates for insurers and banks. Detail on the Yield tab.

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

The setup is the reverse of the framework's. P3a is not met: the 2025 drawdown had no dated adverse trigger — the stock fell 25.8% from its 19 Feb 2025 peak ($303.76) to its 5 Nov 2025 trough ($225.54) while every quarter beat consensus (EPS surprises +8.5%, +24.5%, +4.7%), and the only dated event that repriced it was the opposite of adverse, the 15 Jan 2026 Boston Scientific agreement [10]. P3b is met — the peak-to-trough leg printed a volume spike of 2.07x the pre-peak 180-day median, clearing the ≥2x capitulation reference line — though the Dislocation tab reads 2.07x as a modest, orderly bleed rather than a washout, and the drawdown has since fully round-tripped, so the reading is moot for a buyer today. On valuation the gap is an order of magnitude: FY2025 adjusted FCF yield computes to 0.92% — 708 to 808 bps below the 8–9% fortress bar — because adjusted FCF ($115.7M) sits against a $12.62B cap (a ~109x price-to-adjusted-FCF multiple), so P3c is not met [11]. The balance sheet is fortress (net cash), which selects the 8–9% bar, not the 10% default [12]. The forward path is not met at p=0.02: consensus forward FCF yields only ~1.4% (FY2026) rising to ~2.4% (FY2028) on the current cap, so no mean-reversion to the bar exists within 1–3 years — reaching 8% would take ~$1.0B of adjusted FCF against ~$116M today, an 8.7x rise. Counter-fact: consensus FCF is rising, not falling ($173M → $300M FY2026–FY2028), and the sell side is not fearful — but rising conversion on 17%+ growth is exactly what a ~109x FCF multiple already pays for (Yield).

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

Outlast (P4a): met. Penumbra carries no borrowings and ~$545M of cash and marketable investments; total liabilities are entirely payables, accruals and leases (~$27M annual lease payments against $238.7M operating cash flow), so it can outlast any operating problem with no forced debt paydown [13]. The allocation-headroom half of the pillar is now moot: the pending merger bars repurchases and new debt without Boston Scientific's approval [14].

Repurchase engine (P4b): not met. The share count rose every year — diluted shares 33.5M (2016) to 39.3M (2025), +17.4% — driven by $46–59M of annual stock-based compensation, and the sole buyback was a single $100.4M ASR in 2024 at ~$194/share [15] [16]. A rising share count on SBC with no shrinking-count engine is the specific pattern the framework excludes from the self-help pillar — the same mechanism that drops Accenture once adjusted — and the buyback flywheel is legally frozen by the merger besides. The absurdity check runs the wrong way: at the current price it would take ~72 years of raw FCF (or ~109 adjusted) to retire the float, the inverse of the ~3-year signature the check is built to catch. Counter-fact: the one buyback was executed at ~$194, well below the current $321 close and the $374 takeout, and the deal itself values the equity above any price it has ever traded — shareholders are being rescued by a buyer even though self-help is off the table. Dividend cover (P4c) is not_applicable: Penumbra has never paid a dividend and does not anticipate one. Detail on Self-Help.

Diagnosis (P5) — contested

Contested. This is the criterion that pins confidence at low. The independent trial put the probability the post-2024 reset is temporary at P(temporary) = 0.61 (mean 0.57), but the panel is genuinely split: the spread is 0.26 (above the framework's 0.25 line) and the ruling is reading-order sensitive — the judge who read the temporary brief first landed at 0.42, below 0.5, while the permanent-first judges averaged 0.645, a 0.225 order gap. The near-term earnings hit behind the drawdown was close to zero — consensus rose and every quarter beat through the fall — so at the November trough the ~$3.07B of erased market value was a genuine mispricing against a ~$0 earnings basis. The genuine damage arrived later and landed on margin, not sales: Q2 2026 EPS missed 14.4% as first-half SG and A expense rose to $391.3M and operating income slipped even as revenue grew 15.3%, and the line under question is core thrombectomy growth decelerating ~20% → 16% → 12.3% [17] [18]. Both sides are laid out on the Damage Math tab; the ruling is carried as contested and not overridden. Counter-fact: management frames the SG and A step-up as a completed sales-force build after which operating leverage should appear — if true, that converts the margin hit from structural back to timing and tilts the diagnosis temporary; even capitalized, the permanent EPS shave is modest (~$0.3–0.5B, roughly 2–4% of the current cap).

Instrument context (I1) — not verifiable

The tally records I1 as not_verifiable. Listed options exist on PEN (a $12.6B optionable NYSE name), but a current implied-volatility level could not be sourced — options-data pages returned access-denied and the web-research preload failed — so no IV is estimated. The pending fixed-price merger has collapsed 30-day realized volatility to ~5.7% (2026-07-31) against a ~37% historical median, consistent with a stock pinned to a $374 deal. This criterion never blocks the pillar verdicts; here it is moot, because the P1 gate already decides the fit. Full instrument facts on the Clock tab.

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

The tally did not compute a re-rating price: "Re-rating math unavailable because the applicable bar or normalized adjusted FCF is missing." The balance-sheet-scaled bar resolves to fortress (8–9%), but framework-adjusted, normalized adjusted FCF is not_computable in the feature file (stock-based comp absent from the XBRL cash-flow feed across all years), so there is no adjusted-yield anchor to invert into a target price.

What can be said, as framework arithmetic rather than a target: at the low end of the 8% fortress bar, the equity would need roughly $1.0B of adjusted FCF to justify today's $12.62B cap on yield alone — against $115.7M produced in FY2025, an 8.7x rise — and consensus forward FCF reaches only ~$300M by FY2028. A 3x re-rating from here is not a yield-driven proposition; it would require the market to pay an even higher multiple on a much larger cash base, on a name already trading at ~109x adjusted FCF. The base-rate context from the Clock tab is a reminder that this name can swing hard — its two deepest historical drawdowns were −63% (2021, ~24 months to round-trip) and −52% (2023, ~31 months) — but the most recent recovery was completed by the January 2026 takeover, not by operating re-recognition: the day before the deal the stock was still ~9% below its 2023 peak. None of this is a recommendation; the gate has already ruled the name out.

Contested and undetermined

One criterion is recorded contested: P5, the diagnosis. Both model families returned contested; the split is real (probability 0.61, spread 0.26, order gap 0.225) rather than a labelling artifact, which is why it is carried as contested and not rounded to a clean temporary reading. Nothing was marked cannot-determine: every pillar reached a verdict, and the features that could not be computed (adjusted FCF, adjusted-FCF yield, FCF stability, float-retirement years, balance-sheet class) were rebuilt by hand from the filed FY2025 statements rather than left open. The exclusions were not contested here — X1, X2, X3 and S1 were unanimously not_met and X4 unanimously met (the darling screen fired) across both families and the masked seat.

Provenance

Source: fit tally provenance block and the adversarial trial tally.

Two families of models voted, and they agreed on the verdict that matters — the year-10 gate failed on every seat, masked and unmasked. The name-mask probe moved nothing (no gate criterion changed, zero probability gap), so the answer is not a prior in disguise. Confidence is nonetheless low, and honestly so: the diagnosis that would decide whether the 2025 dislocation was ever a real mispricing is contested, with judges landing on both sides of even money depending on which brief they read first.

Falsifier ledger

These are the standing what-would-change-this conditions carried from the jury and the diagnosis trial's flip-conditions. They are redundant by design — harvested independently from multiple seats — and reproduced verbatim so each threshold, direction, and window stands on its own:

Data gaps

The run could not answer several things, and the verdict is stated with them in view. The deterministic feature file could not compute adjusted FCF, adjusted-FCF yield, FCF stability, yield baseline, float-retirement years or balance-sheet class — stock-based compensation was absent from the XBRL cash-flow feed for every year and the balance-sheet debt/cash fields were null — so every adjusted figure was rebuilt by hand from the filed FY2025 statement of cash flows (SBC $59.2M) and the balance sheet (zero borrowings, net cash confirmed directly). Revenue before FY2023 was reconstructed from segment US+international totals because income.json carried null revenue. There is no reported short-interest series for PEN (FINRA returned none), and no 2025-dated consensus-revision snapshot, so the estimate-versus-price lead/lag through the drawdown is inferred from consistent beats. A current implied-volatility level and long-dated option-chain detail could not be verified (options pages returned access-denied; the web-research preload failed; the tech feed provides realized volatility only). Finally, the merger's antitrust status — the FTC Second Request (2026-03-16), EU/other non-U.S. clearances, the reverse termination fee, and the updated close timing — is web-sourced (Boston Scientific Form S-4/424B3, MLex) and is not restated in the corpus PDFs, which predate the Second Request.