Fit

Fit

Figures converted from Israeli shekels (₪) at historical FX rates — see data/company.json.fx_rates. NICE reports its financial results in US dollars, so revenue, EPS, FCF, and guidance figures are unchanged; only the shekel share-price and market-cap levels are converted (at ₪→$ 0.32641, applied uniformly so the −29.6% / −40.6% drawdown is preserved). Ratios, margins, and multiples are unitless and unchanged.

Outside the framework's universe (U2 not met); contested: U1, X1, X2, X3, X4, S1

The framework does not proceed on NICE: at a market value of about $6.0B the company sits roughly 40% below the $10B universe floor, and a size miss stops the screen before any pillar can carry it. Verdict: outside the framework's universe. Confidence is low — under the name-mask probe five screening criteria (U1 and the four exclusions) resolved on different labels between the two model families, a prior_driven_risk signal that caps confidence regardless of how clean the underlying facts are. No hard exclusion is triggered; watchlist_only is off. This is a screen result, not a judgment on the business.

Universe and exclusions

Here is the decisive point. NICE misses on size, and nothing downstream can rescue it. At 63.323M shares and a 21 July 2026 TASE close of $95.2 (₪291.8 at ₪→$0.32641), equity value is about $6.0B, cross-checked against the NASDAQ ADR (~$95–101 × 63.323M ≈ $6.0–6.4B) and third-party market data ($5.8–6.3B). The universe floor is $10B. On every basis NICE lands ~40% short. The one honest mitigant, stated in the same breath: this ~$6.0B follows a ~40% ADR decline from a 52-week high near $175, so on its pre-drawdown market cap (~$10.5–11B) NICE would have sat right at the line. Today it does not. U2 → not met, and by the framework's own gate that ends the universe question: U2 not_met → out_of_universe.

Source: market value derived from fit_features.market_cap.native (₪18,477,651,400) × company.json.fx_rates 0.32641, cross-checked to the NASDAQ ADR; the FY2025 20-F reports 60.43M ordinary shares outstanding, so the 63.3M feature basis runs market cap ~7% richer than the primary count.

The remaining screens are clean on the substance:

• U1 — Listing (contested). NICE is a bona-fide U.S.-listed security: NASDAQ ADRs since 1996, a Form 20-F filer as a foreign private issuer, explicitly not a Chinese ADR [1]. The letter of the rule names "US-listed stocks, or European companies with a US-listed ADR"; NICE is Israeli-domiciled, not European, so the geography test is met in substance and ambiguous only on wording — which is why the seats split rather than agreed.
• X1 — Auto/OEM (contested). No automotive or hardware-OEM exposure: the two reportable segments are Customer Engagement software ($2,459,974K) and Financial Crime & Compliance software ($485,425K) [2]. The car-company exclusion does not bite.
• X2 — Promotion pattern (contested). Management's promise-versus-delivery is mixed but not promotional — it kept its FY2025 cloud and EPS guides while missing an 18% FY2024 cloud target and walking back two mid-year raises — with clean governance and thin 0.7% insider ownership. Not a framework exclusion; the full record is in Self-Help.
• X3 — Structural decline (contested). Revenue rose every year for seven, $1,573.9M (2019) to $2,945.4M (2025); consecutive_decline_years = 0, three_year_hsd_decline = false. The disqualifier is absent, though growth decelerated to +7.7% and the legacy services and product lines are already shrinking within the mix. See Durability.
• X4 — Consensus-saturated story (contested). NICE trades at ~2.05x sales and ~10x earnings after a ~40% ADR fall driven by AI-disruption fear — a value multiple, not a darling's. The reverse risk is that the de-rating is deserved; that is the Damage Math question, not an exclusion.
• S1 — China dependence (contested). APAC is 5.3% of revenue in total ($154.7M of $2,945.4M), so China is at most a low-single-digit share and is not separately disclosed as material; the large APAC workforce is India-based engineering and support [3]. The real geopolitical exposure is Israel, which the China screen does not address.

None of these five is a genuine two-sided disagreement about whether the exclusion fires — on both model families' readings, none fires. They register as "contested" because one family recorded "not met" and the other "not hit," and that labeling split is what the name-mask probe surfaced. It is reported as contested here because the tally records it as contested; it is not softened, and it does not change the gate.

Pattern match

NICE fits the quality tech monopoly/duopoly on a fear dip setup only in part, and none of the other three cleanly. The dip is real and event-dated (a 40.6% peak-to-trough fall on a forward-guidance reprice, Dislocation), the business is a durable category leader with ~39 years of history, and the fear is specific and testable — that agentic AI erodes NICE's per-seat pricing. What the pattern requires and NICE does not supply: the setup's precedents (Meta, Amazon, Google at ~$90 in 2022) were $250B–$1T names inside the universe, and the "monopoly/duopoly" leg is a contested CCaaS oligopoly, not a structural gate. It matches none of the cyclical-bank, high-dividend-yield, or insurance-forecasting-error patterns. So: a partial match to one pattern, on a name the size screen has already put out of the universe.

The pillar ledger

Each pillar below states the criterion, its reference line, where NICE sits in arithmetic, and — in the same treatment — the strongest surviving fact the other way. Reference lines, never grades.

Source: ruchir/fit_tally.json criterion aggregates; supporting arithmetic derived from the surviving claims and the FY2025 20-F, cited in each pillar treatment below.

Year-10 gate — P1

The one pure gate is binary: will year-10 revenue and adjusted FCF both be higher, with very high conviction? The jury says no — all four seats returned not met, trimmed-mean probability 0.61, spread 0.07, both families agreeing. The arithmetic behind the doubt: of the framework's five conviction sources, only long operating history applies cleanly (founded 1986, ~39 years). Capital intensity gives no moat (capex/revenue = 3.2%). There is no regulatory entry barrier on Customer Engagement, which is 83.5% of revenue. Market structure is a contested oligopoly — NICE competes against Amazon Connect, Avaya, Cisco, Five9, Genesys and TalkDesk plus AI-native entrants [4] — not a monopoly or duopoly. And NICE's own 20-F flags that AI may "decrease in demand for solutions priced based on the number of human agents deployed" [5]. The strongest fact against the doubt sits in the same frame: a ~39-year survivor that already navigated on-premise-to-cloud, still compounds cloud revenue double digits (76% of revenue), and targets over $1B of AI revenue by 2028 is strong evidence year-10 revenue is higher — the doubt is about the conviction level the binary gate demands, not the direction. Full treatment in Durability.

Consistency — P2

The test is a stable rolling five-year average of adjusted FCF. The verdict is cannot determine, and the reason is a specific missing input: adjusted FCF exists for only three consecutive years (~$148M / ~$391M / ~$175M, 2023–2025 — lumpy on acquisition timing), and stock-based compensation is absent from the feature build for the earlier years, so the deterministic rolling average and coefficient of variation cannot be computed. One jury seat read the lumpiness as not met; three could not settle it. Note that the reader-facing "reported FCF has been positive and rising for seven years, so consistency holds" argument was refuted by the skeptic — the framework's basis is adjusted FCF, not reported, so the reported-FCF record supports nothing here. The named missing datapoints, verbatim from the jury:

The counter-fact: the five-year averaging of acquisitions is itself the framework's remedy for lumpiness, so volatility in adjusted FCF is expected rather than disqualifying — which is precisely why the honest answer is cannot determine, not not met. See Yield.

Dislocation and yield — P3

The dislocation is genuine and the entry conditions are met; the yield is not. P3a (identifiable event) — met: a single dated trigger, the 6-May-2026 Q1 report, on which the stock fell 18.8% in one session — even though the quarter itself beat and management raised full-year EPS guidance to $10.98–$11.18 [6]; the adverse signal was the forward growth/margin trajectory (Q2 guided to +5.5%, cloud NRR 107%). P3b (capitulation) — met: volume spiked 3.3x over the fall, the heaviest session running 13.8x its average near the June trough — a washout signature, albeit on a pre-peak baseline truncated to ~30 trading days by the six-month price feed. P3c (yield vs bar) — not met: NICE's net-cash balance sheet selects the fortress bar of 8–9%, and on the framework basis (reported FCF − SBC − 5-year-average acquisitions) FY2025 adjusted FCF is $175.0M, a 2.9% yield on ~$6.0B — roughly 560 bps short; even the charitable ex-Cognigy treatment reaches only 5.7% [7]. The counter-fact stands beside it: reported FY2025 FCF yield is ~10.3% and the SBC-only-adjusted yield is ~7.9%, so a screen that skips the M&A adjustment would show NICE at or above the bar. P3d (forward path) — not met: consensus reported-FCF yields of 9.8%–13.8% (FY2026–FY2029) clear the bar, but on the adjusted basis reaching 8% by ~2028 needs the acquisition engine to stop, which NICE's serial-acquirer record argues against — estimated probability ~0.305, spread 0.03, both families agreeing. Full workings in Yield.

Balance sheet and self-help — P4

Capital allocation is the framework's flywheel, and here it is intact. P4a (outlast + headroom) — met: NICE entered the dislocation debt-free after settling its $460M notes in cash at maturity in September 2025, with $417.4M of cash and short-term investments (cash $379.4M + ST investments $38.0M) [8] and a $300M undrawn revolver; net debt/EBITDA is −0.49x. The cushion is thinner than a year ago, though — liquidity fell from ~$1.62B to $417.4M to fund Cognigy and the note repayment. P4b (repurchase engine) — met: buybacks are executed, not just authorized — $488.9M in 2025 (up from $144.9M in 2022) plus a record $253.3M in Q1 2026 [9] — with share count falling. The counter-fact: SBC of $146–182M a year and the part-stock Cognigy deal hold the net reduction to ~1.6%/yr, so the float is retiring slowly. P4c (dividend cover) — not applicable: NICE pays no ordinary dividend and states no plans to initiate one [10]; there is nothing to cover, and the return case rests entirely on repurchases. Detail in Self-Help.

Diagnosis — P5

The question is whether the impairment is temporary. The verdict is met — the market erased 29.6% of NICE's equity value peak-to-current (40.6% peak-to-trough), while a transparent two-scenario NPV of the near-term hit destroys only ~1.6% of pre-shock value if temporary and ~10–18% if permanent; the residual gap is a multiple de-rating, not a cut to cash flows. The profile's adversarial trial — two cited briefs, three blind judges — put the probability the impairment is temporary at 0.61 (range 0.58–0.62, spread 0.04, not contested). The counter-fact carried in the same ruling: NICE shows real structural-decline signals — cloud growth decelerated to ~12% ex-Cognigy, NRR at 107% with management-flagged near-term pressure [11], and a seat-to-consumption pricing reset the filing warns can commoditize its products — which is exactly why the judges landed at 0.61 rather than resolving it either way. The two cases are laid out fairly in Damage Math.

Instrument context — I1

Facts only, no advice. Listed options on the NASDAQ ADR run out to a 21-Jan-2028 LEAP (~18 months), clearing the framework's ≥12-month (target ≥18) line, with near-money open interest in the hundreds to low thousands. But the tally records I1 as not verifiable: the only implied-volatility reading available (~58–62%, elevated on the reference lines) is a near-dated August-2026 ATM figure that spans the 5-Aug earnings, from a delayed public feed rather than real-time exchange data, and a clean long-dated IV could not be confirmed. So the instrument exists; the volatility level the framework would read cannot be verified from the corpus. See Clock.

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

The tally's re-rating math is unavailable, by its own note:

Re-rating math unavailable because the applicable bar or normalized adjusted FCF is missing.

Both inputs are missing for good reasons — the name misses the universe screen, and fit_features could not compute a native USD market cap or a normalized adjusted FCF (FX was not applied in the feature build; SBC was absent). So the framework does not open a target price. The closest available arithmetic points the other way: on the reconstructed adjusted basis, FY2025 adjusted FCF of $175.0M against ~$6.0B is a 2.9% yield, and reaching even the 8.5% fortress bar would require the price to fall to roughly $2.1B — not a 3x up but a further ~65% down — unless adjusted FCF roughly triples first via a halt to acquisition spend. The base-rate context from Clock: NICE's prior guide-driven drawdowns round-tripped in about four months (the ~20% November-2025 fall recovered to a new high by March 2026), but the current 40.6% episode is roughly double the deepest bracketable precedent, so that round-trip cannot simply be extrapolated. Stated as arithmetic against the framework's reference lines; not a recommendation.

Contested and undetermined

Contested (six criteria). U1, X1, X2, X3, X4 and S1 are all flagged contested. On the merits, none is a live dispute about the outcome: the listing is a genuine U.S.-accessible ADR (Israeli, not European, by the letter), and none of the four exclusions plus the China screen is triggered on either family's reading. The contest is a labeling divergence — one model family recorded "not met," the other "not hit" — and that split is what the name-mask probe caught and what holds confidence at low. Reported as contested because the tally records it as contested.

Undetermined (one criterion). P2 (FCF consistency) is cannot determine — the deterministic rolling-five-year adjusted-FCF series cannot be built because SBC is missing for the earlier years and only three consecutive adjusted-FCF years exist (the named datapoints are listed under the P2 treatment above).

Provenance

Item	Result
Jury composition	4 seats + 1 masked; families claude (a, b, masked) and codex (c, d)
Agreement / spread	P1 0.61 (spread 0.07), P3d 0.305 (0.03), P5 0.61 (0.04); both families agreed on P1, P3c, P3d, P5
Trial order-stability	temporary-first mean 0.58, permanent-first mean 0.615, gap 0.035
Name-mask probe	gate criteria U1, X1, X2, X3, X4 resolved differently under mask; max probability gap 0.005 → prior_driven_risk = true
Skeptic counts	17 adjudicated — 13 survived, 2 weakened, 1 refuted, 1 unverifiable — plus 22 triaged-only

Source: ruchir/fit_tally.json provenance block and ruchir/trial/tally.json.

Two sentences of plain reading. The verdict was pressed hard — a four-seat two-family jury, an adversarial temporary-versus-permanent trial with three blind judges, a claim-level skeptic that refuted one reader-facing claim and weakened two more, and a name-mask probe that re-ran the screen with the company's identity hidden. That last probe is why confidence is low: with the name masked, five of the screening criteria came back on different labels, so the framework flags a risk that a prior was doing some of the work — even though the size miss that decides the verdict is unambiguous either way.

The falsifier ledger

These are the standing what-would-change-this conditions carried from the tally — the first five are the framework's generic templates, the rest the trial's name-specific flip conditions with their thresholds, directions and windows:

Data gaps

• No native USD market cap in fit_features (reporting currency USD but price currency ILS, FX not applied), so all market-cap USD figures are derived by applying ₪→$ 0.32641 uniformly; a per-period FX split would break the −29.6% / −40.6% drawdown percentages, so a single rate is used deliberately.
• Adjusted FCF and its yield are not_computable in fit_features: SBC is absent from the feature build (present in the 20-F at $146.0M for FY2025) and there is no clean five-year acquisition window, so the acquisition drag is bounded, not precise; the figures used here are the Yield tab's reconstruction from the filed cash-flow statements.
• The rolling five-year adjusted-FCF stability series (fit_features.fcf_stability) is not computable — fewer than five consecutive adjusted-FCF years exist (the exact missing inputs are listed under P2).
• Share-count basis differs: fit_features uses 63.3M while the 20-F reports 60.43M outstanding at year-end and management cited ~58.5M by end-March 2026, so per-share figures run ~7% richer on the feature basis than on the primary count.
• Price/volume history is a partial six-month web feed (23 Jan – 21 Jul 2026), so the $135.2 peak is a window high rather than a verified multi-year top, and the capitulation baseline is truncated to ~30 trading days.
• China revenue is not separately disclosed; the 5.3% APAC total is an upper bound, not an exact China figure.
• A clean long-dated (Jan-2028) implied-volatility reading was not separately verifiable; the ~58–62% figure is a near-dated August-2026 ATM reading spanning the 5-Aug earnings from a delayed public feed.
• No official short-interest data exists for the TASE-listed security, so forced/short-driven selling cannot be quantified.