Fit

Fit — Amdocs Limited (DOX)

Outside the framework's universe (U2 not met); contested: P2

Amdocs sits outside the framework's universe. The scale line (U2) is not met — USD market capitalization is ~$6.18B against the $10B minimum, unanimous across all four jurors — and under the framework's rules a universe miss ends the screen before any pillar can qualify it. Confidence is high: two model families agreed, the trial was order-stable, and load-bearing spreads were at most 0.15. No exclusion was hit and the name-mask probe raised no prior-driven-risk flag. One criterion is contested — P2 (FCF consistency) — where the two families split met vs cannot-determine.

Universe and exclusions — unsoftened

Amdocs clears the listing test but fails the scale test, and the scale test is decisive.

U2 (scale) — not met (unanimous, both families). At the $55.28 close on 2026-07-30 the 111.746M ordinary shares outstanding are worth $6.18B, roughly $3.82B below the $10B line. The shortfall is drawdown-driven rather than structural: at the 2025-06-12 peak of $94.08 the same share count implied ~$10.5B, so on a pre-dislocation basis the company sat at or above the line. But the framework measures the figure today, and today it is below. This is the gate the tally applied — U2 not_met → out_of_universe — and nothing downstream overrides it.

U1 (listing) — met (unanimous). Amdocs' ordinary shares have traded directly on the Nasdaq Global Select Market since December 20, 2013 [1]; it is a Guernsey-incorporated foreign private issuer filing Form 20-F [2] — a Channel Islands (European) domicile with a US primary listing, read as a directly-listed foreign private issuer rather than an ADR.

No exclusion was triggered. X1 (auto OEM) — Amdocs is a single-segment provider of BSS/OSS, digital and cloud software and managed services to communications and media companies, explicitly described as a "systems integration" vendor, with no automobile-manufacturing activity [3]; not a hit. X4 (consensus-saturated story) — DOX trades at ~1.36x FY2025 sales after a ~47% drawdown with only ~3% forward revenue growth expected, the antithesis of an extreme-multiple, saturated-growth narrative; not a hit. X2 (promotion pattern) and X3 (structural decline) are evidential and are treated in the pillar ledger below; both resolved not_met. S1 (China dependence) is a sensitivity flag only and was not raised: geography is disclosed as North America (65.8%), Europe (15.6%) and Rest of World (18.6%), China is not a reported geography, and no development facility sits in China [4].

Note the exclusions and S1 show a label split — three jurors recorded not_met and the fourth recorded not_hit — but that is wording, not a factual dispute: every juror found the exclusion did not apply.

Pattern match

On surface features this has the shape of the framework's FCF-dislocation setup: a dated adverse catalyst (two guidance-day drops), a ~47% peak-to-trough drawdown, an adjusted FCF yield within ~125 bps of the balance-sheet bar, and an inverted signature where earnings rose while the multiple compressed. That is the closest resemblance among the constructive setups.

It qualifies as none of the four patterns, because the disqualification happens at the universe screen — a sub-$10B name is out before pattern qualification is even reached. Two pillar checks would independently block it as well: the year-10 durability gate (P1) is not met, and today's adjusted yield (P3c) sits below its bar. The resemblance is real but does not convert into a fit. This is framing only; it does not alter the tally's verdict.

The pillar ledger

Source: deterministic fit tally (per-criterion aggregate verdicts, trimmed-mean probabilities, and spreads).

Year-10 durability (P1) — the gate that would have decided a universe-qualified name

Not met (unanimous, both families; p_year10 0.61, spread 0.06). The revenue disqualifier is false — FY2025 GAAP revenue fell 9.4% to $4,532.9M, but that is one year, not the three consecutive high-single-digit declines the disqualifier requires (consecutive_decline_years = 1). The durability case is genuine: ~66% of revenue is recurring managed services renewing near 100%, remaining performance obligations are ~$6.4B, the model is asset-light, and consensus models FCF rising from $651.6M (FY2026) to $940.2M (FY2029) [5]. What puts the year-10 leg in genuine doubt is that organic growth is only ~1.5% (about half of the ~3% headline is inorganic), the core CSP end market is mature, SaaS/cloud challengers attack the bespoke-integration model — CSG Systems' FY2025 10-K names Amdocs directly as a competitor [6] — the three-year GenAI growth engine remains undelivered, and ~46% of revenue sits in two carriers (AT&T 25.9%, T-Mobile 19.9%) with T-Mobile a named FY2026 headwind [7]. The gate resolves on genuine doubt about "very high conviction," and the doubt is present.

Strongest surviving counter-fact: a lower year-10 FCF is hard to picture — FCF has held $530–698M through the disruption, consensus models it rising, and the ~−3.3%/yr buyback keeps compounding per-share value even at ~3% top-line growth. Full treatment: The Carrier's Backbone.

FCF consistency (P2) — the contested criterion

Contested (claude family met; codex family cannot_determine). Raw free cash flow held a positive $452.6M–$715.4M band across FY2020–FY2025 with no negative year, a ~$610.0M mean and ~15% coefficient of variation [8]. The split is about which metric: the spec's exact P2 measure is a rolling five-year average of adjusted FCF, and that window closes only from FY2024 (the corpus reaches acquisition history back to FY2020), leaving just two fully-windowed adjusted years — FY2024 $389.4M and FY2025 $447.9M, both positive and rising. The claude jurors read the six-year raw-FCF stability plus the two rising adjusted years as met; the codex jurors held that a two-point fragment cannot establish the five-year rolling metric and returned cannot_determine. Neither reading disputes a number; they disagree on whether the available window suffices. This is framing — P2 is not a gate — and it does not change the universe verdict.

Strongest surviving counter-fact: the raw-FCF stability is partly a capital-intensity story — capex roughly halved from ~$227M (FY2022) to ~$104M (FY2025) while equity-based compensation more than doubled to $104.9M [9], so the SBC-adjusted picture is less flattering than the raw line. Full treatment: The Carrier's Backbone.

Dislocation + yield (P3)

P3a identifiable event — met (unanimous). Two dated adverse prints anchor the fall. The Q4 FY2025 release (Nov 11, 2025) initiated a soft Fiscal-2026 outlook of only 1.0%–5.0% constant-currency growth after the 9.4% reported FY2025 revenue decline [10]; the stock fell ~7.9% the next session on ~6.9x median volume. The Q1 FY2026 print (Feb 3, 2026) fell ~8.6% even though revenue rose 4.1% YoY, a T-Mobile multi-year renewal was signed, and guidance was reiterated [11]. Strongest surviving counter-fact: the −47% drawdown spans 378 days across three legs, and the steepest (June 2026 to the $49.87 trough) carried no dated company catalyst — corpus news links it only to sector-wide AI-restructuring sentiment, so the setup reads as much like an orderly multi-leg de-rating as a single fear event. Full treatment: The Engineered Decline.

P3b capitulation — met (unanimous; reference line, not a gate). The capitulation gauge measured a 3.11x volume multiple against the 2x reference line, with single-day event spikes of ~6.9x (2025-11-12) and ~5.3x (2026-02-03). The caveat: the sustained-volume peak sat in early-March 2026 after the February leg, not at the June trough, whose down day ran only ~3.1x — so there was no classic single-day blow-off at the low.

P3c yield vs bar — not met (unanimous). At FY2025 year-end net debt was $325.0M against ~$927M EBITDA (0.35x), a fortress class that sets the bar at 8.5% [12]. Hand-computed adjusted FCF yield (FCF − SBC − 5yr-avg acquisitions) is 7.25% current and 7.48% on a three-year average — short of the bar by ~125 and ~102 bps [13]. The deterministic feature returned unknown (the feed lacked FY2025 debt/cash and the SBC/acquisition add-backs), so both the class and the yield were rebuilt from the primary cash-flow and balance-sheet statements. Strongest surviving counter-fact: on the raw (unadjusted) definition the trailing FCF yield is 10.4% and clears the bar comfortably — it is only the ~$197M SBC-plus-acquisition haircut that pulls Amdocs under the line; and by Q2 FY2026 borrowings rose to ~$900M against ~$214M cash (net debt ~0.7x), shifting the class toward moderate (10% bar) and widening the gap [14]. Full treatment: Priced for Decline.

P3d forward path — met (unanimous; p_forward_bar 0.615, spread 0.02). Applying the same haircut to consensus FCF, adjusted forward yield runs 7.35% (FY2026), 8.74% (FY2027) and 10.94% (FY2028) — crossing the 8.5% bar in FY2027 on the current depressed price, no mean reversion required [15]. Strongest surviving counter-fact: the crossing is mechanical on a static price — it rests on consensus FCF rising to $737M (FY2027) off a year in which revenue fell 9.4%, and any upward re-rating (the whole point of a dislocation entry) would compress the yield back below the bar. Under the Q2 FY2026 moderate class and 10% bar the crossing slips to FY2028. Full treatment: Priced for Decline.

Balance sheet + self-help (P4)

P4a outlast + allocation — met (unanimous). The only funded debt is a single $650.0M Senior Notes issue (2.538% fixed) that does not mature until June 2030, the $800M revolver runs to July 2029 and was undrawn and covenant-compliant at Sept 30, 2025, and investment-grade ratings are intact — so debt service does not compete with repurchases [16]. Strongest surviving counter-fact: headroom is thinning, not solvency — returning more than 100% of FCF drew cash and short-term investments from $514.3M (FY2024) to $325.0M (FY2025) to ~$214M (Mar 2026) and now leans on an $800M commercial-paper program launched March 2026 [17].

P4b repurchase engine — met (unanimous). Buybacks are executed cash (~$490–563M/yr, $2.11B over FY2022–FY2025) that shrank diluted shares 9.6% (123.65M → 111.75M, ~−3.3%/yr); the hard-fail test is negative (share_count_trend.rising = false), and management returned more than 100% of FCF through repurchases and dividends into Q2 FY2026 [18]. Strongest surviving counter-fact: the three most recent vintages were struck at ~$85–90/share against a ~$55 quote, and continued capacity is increasingly commercial-paper-funded rather than covered by current FCF [19].

P4c dividend cover — met (unanimous). The dividend is material to the return case (~3.8% current, ~4.1% on the proposed FY2026 step-up) and well covered: FY2025 dividends declared of $224.4M consumed ~35% of $645.1M FCF, DPS has risen every year and was maintained through COVID, and forward cover on the $710–730M FCF guide is ~3x [20]. Strongest surviving counter-fact: the dividend alone is covered, but it is the junior lever inside a >100%-of-FCF total payout, so the package as a whole is partly balance-sheet-funded. Full treatment: Priced for Decline.

Diagnosis (P5)

Met — leaning temporary (unanimous; p_temporary 0.69, spread 0.06, trial-carried). The probability that the impairment is temporary was produced by an adversarial trial — two opposing cited briefs, three independent judges (0.69 / 0.66 / 0.72) — returning a mean of 0.69 with contested = false and an order-stability gap of 0.0. The signature is inverted: price fell ~41% from its three-year high while FY2025 earnings rose — GAAP diluted EPS +18.8% ($4.25 → $5.05) and non-GAAP EPS +8.5% ($6.44 → $6.99) [21]. The mechanism that "broke" the top line was a deliberate, margin-accretive phase-out of low-margin non-core work (ex-phase-out revenue grew 3.1%), which anniversaries rather than compounds [22], and the company reiterated FY2026 guidance while consensus FCF rises through FY2029.

Strongest surviving counter-fact: the recovering base is a slow-grower — management concedes roughly half of FY2026's ~3% constant-currency growth is inorganic, leaving organic growth near ~1.5% [23] — and T-Mobile (~20% of revenue) re-signed yet is still guided to lower spend because its "spending appetite is lower," showing budgets can fall even where renewals hold [24]. On a DCF-lite at r=10% the market-implied FCF (~$618M) sits inside the permanent band, so the value gap exists only if the impairment is temporary. Full treatment: Priced for Decline.

Instrument context (I1)

Not verifiable (unanimous; facts only, never blocks a pillar). DOX's 30-day mean implied volatility was 49.55% as of 2026-07-30 (AlphaQuery), inside the ~55 acceptable reference line, with a bearish put-over-call skew (put IV 53.30% vs call IV 45.81%) and IV ~20 points above realized volatility. The criterion is not_verifiable because a qualifying long-dated instrument (expiry ≥ 12 months, target ≥ 18) could not be confirmed: the only reachable dated chain (Public.com/Xignite) showed expirations through 2027-01-15 (~5.5 months), while Nasdaq, Yahoo, MarketChameleon and Optionistics were bot-walled or timed out. I1 drives only the watchlist overlay, which does not apply to an out-of-universe name. Full treatment: Priced for Decline.

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

The tally records: "Re-rating math unavailable because the applicable bar or normalized adjusted FCF is missing." The bar, normalized_adjusted_fcf, implied_market_cap_at_bar, and upside_to_bar_pct fields are all null, because the deterministic adjusted-FCF-yield feature was not_computable (no SBC add-back in the feed, cash acquisitions read $0, and no complete five-year acquisition window). The framework's price-at-bar-yield test cannot be rendered from a canonical feature.

The closest available arithmetic is the DCF-lite the Damage Math pillar carried (r = 10%, consensus and FY2025 actual FCF):

Source: derived from reported financials (FY2025 Consolidated Statements of Cash Flows) and consensus FCF [25].

A 3x in three years from the $6.18B market cap implies roughly $18.5B. Even the temporary-diagnosis capitalization tops out near $6.5–8.1B, and the permanent case ($5.3–5.8B) is below today's market cap — so on the modeled paths the framework's 3x target is not reached on any scenario. It would require FCF to compound far faster than the temporary case assumes and consensus to concede a full re-acceleration of the ~1.5% organic line. For base-rate context on how deep this drawdown is and how such episodes recover, see Priced for Decline. This is the framework's target test stated as arithmetic, not a recommendation.

Contested and undetermined

Contested: P2 (FCF consistency). The two families split on which metric governs. Both readings agree the six-year raw-FCF series has no negative year (~$610M mean, ~15% CV) and that the two fully-windowed adjusted years (FY2024 $389.4M, FY2025 $447.9M) are positive and rising. They disagree on sufficiency: the claude jurors resolved met on that evidence; the codex jurors resolved cannot_determine because the spec's rolling five-year adjusted-FCF average cannot be formed from only two computable adjusted years. It is a window-length disagreement, not a factual dispute, and P2 is not a gate.

Nothing resolved to cannot-determine at the criterion level, so there is no single missing datapoint that flipped a verdict. One criterion, I1, is not_verifiable — IV is established from a dated source but a directly read long-dated option chain could not be pulled (vendor bot-walls); I1 never blocks a pillar.

Provenance

Field	Value
Jury seats	a = claude, b = claude, c = codex, d = codex; masked = claude
Model families	claude, codex
Universe agreement	U2 unanimous not_met across both families
Load-bearing spreads	≤ 0.15 (P1 0.06; P3d 0.02; P5 0.06)
Trial order-stability	temporary-first mean 0.69; permanent-first mean 0.69; gap 0.00
Name-mask probe	max probability gap 0.035; no gate criteria flipped → prior_driven_risk = false
Skeptics	11 survived, 5 weakened, 0 refuted, 2 unverifiable, 17 triaged-only

Source: deterministic fit tally (provenance block) and the adversarial trial tally.

The verdict was pressed and held: two independent model families reached the same universe call, and re-ordering the trial's reading sequence moved the diagnosis probability by 0.00. The name-mask probe — re-running the jury with the company's identity hidden — flipped no gate verdict and moved probabilities by at most 0.035, so the call is driven by the evidence, not by a prior about the name. The single label split (three not_met / one not_hit on the exclusions) is wording, not disagreement.

The falsifier ledger

These are the standing what-would-change-this conditions carried in the tally. The first five are the framework's generic templates (each a direction on a level already underwritten); the remainder are Amdocs-specific, most embedding their own threshold and window (a fiscal-quarter deadline, a margin or FCF floor, a customer event). Several near-duplicate each other because both the trial and the pillar briefs contributed conditions.

Data gaps

What the run could not answer, from the tally's list:

• Adjusted-FCF features are not_computable. The cash-flow feed carries no SBC field and reads acquisitions_cash as $0 despite transcript-confirmed deals, so adjusted_fcf, adjusted_fcf_yield, fcf_stability, yield_baseline and balance_sheet_class all returned not_computable. Every adjusted-FCF figure, the yield-vs-bar test, and the fortress classification were rebuilt by hand from the FY2025 (p.136) and FY2022 (p.91) cash-flow statements.
• The P2 rolling five-year adjusted-FCF average cannot be formed. Only the FY2021–FY2025 filings are in the corpus, so the trailing-5-year acquisition window closes only from FY2024 — two adjusted years versus the five a rolling average needs; the ~15% CV reported is on the raw six-year series, a proxy.
• China exposure cannot be sized. Geography is disclosed only as North America / Europe / Rest of World, so any China component is bounded by the $844.3M Rest-of-World line, not measured.
• No matched before/after consensus revision series. The near-term "hit" is inferred from reiterated guidance plus a rising consensus FCF path; a point-in-time consensus-cut magnitude around the drawdown could not be sourced from a primary page.
• No dated catalyst for the steepest leg. The June-2026 leg to the $49.87 trough has no company filing/release in the corpus — web research surfaced only sector-wide AI-disruption sentiment.
• Instrument granularity. A qualifying long-dated LEAPS with adequate open interest could not be confirmed (vendor option-chain pages bot-walled); IV itself is dated and verifiable.
• Insider transaction direction is unobservable. As a foreign private issuer Amdocs is exempt from Section 16, so the X2 second prong rests on the 2.0% standing ownership stake [26] with no transaction feed.
• U2 share count. Market cap uses FY2025 period-end shares (111.746M); if current shares differ the ~$6.18B figure shifts modestly, not enough to reach the $10B line.

Checked and unremarkable

All eight synthesis scouts returned load_bearing — none was routine or empty — so their findings feed the pillars above rather than this list: accounting/cash-quality, business-economics, capital-allocation, competition-moat, history/track-record, industry, people/governance, and valuation-expectations.

Playbook version

Playbook fcf-dislocation, version 4 (spec playbook/fit-spec, version 2), as frozen for this run.