V-Score Card

TICKER:          SNAP (Snap Inc.)
V-SCORE:         1.87
VERDICT:         COLLAPSED (V < 2.0)
CONVICTION:      κ = (1.87 − 3.0)⁺ = 0.00
BASKET WEIGHT:   w = 0

DIMENSIONS
  C  Compound Cognition      (w=0.25):  2    →  0.50
  E  Irreducible Infra       (w=0.22):  2    →  0.44
  U  Ecosystem Breadth       (w=0.18):  2    →  0.36
  A  Distribution            (w=0.12):  2    →  0.24
  M  Ecosystem Gravity       (w=0.15):  3    →  0.45
  F  Friction (penalty)      (w=−0.06): 2    → −0.12

ARITHMETIC
  V = 0.50 + 0.44 + 0.36 + 0.24 + 0.45 − 0.12 = 1.87
  G₁ = 𝟙[E>1] = 𝟙[2>1] = 1                       PASS
  G₂ = 𝟙[A>1 ∨ C+E+U≥12] = 𝟙[2>1 ∨ 6≥12] = 1    PASS
  V = 1.87 × 1 × 1 = 1.87

FAST SCREEN (Bustamante): 0/3
  ✗ Proprietary data product that cannot be synthesized locally
  ✗ Regulatory mandate requiring routing through SNAP
  ✗ Transaction-embedded (SNAP is not a payment/settlement rail)

Regime Context

15-WEEK FACTOR DECOMPOSITION (2025-12-17 → 2026-04-07)

r_SNAP = α + 2.08×SPY + 0.88×XLC + ε

         α̂ (ann)    σ_idio    IR        %Idio    β_SPY
SNAP    −101.3%     46.8%    −2.278     60.4%    2.08
META     −25.4%     21.3%    −1.190     30.8%    1.69
PINS     −94.2%     51.3%    −1.835     87.1%    0.98

ρ_intra (20d rolling, SNAP/META/PINS/GOOG):
  Jan: 0.10 → Feb: 0.23 → Mar: 0.54 → Apr: 0.68

ATTRIBUTION (15wk)
  Total SNAP return:   −34.9%
  Factor (regime):      −9.0%  (24%)
  Idiosyncratic:       −28.4%  (76%)

All social/ad-sector IRs are negative. ρ_intra has risen 7x in 12 weeks, compressing idiosyncratic signal. When ρ → 1, IR → 0 mechanically for all names — IR measures the regime, not alpha.

But SNAP's decline is 76% idiosyncratic. Even in a correlated selloff, three-quarters of the drawdown is SNAP-specific. The regime is amplifying a pre-existing structural weakness, not creating a false discount.

V_SNAP  = 1.87   (structural, regime-invariant)
IR_SNAP = −2.278 (regime-contaminated, not thesis signal)
δ_SNAP  = V − V_market ≈ 0  (no mispricing — market converging to structure)

Dimension Analysis

C = 2 | Compound Cognition

The AR lens ecosystem has genuine compound knowledge: 5,927 issued patents spanning AR, spatial computing, computer vision, and generative AI (10-K lines 926-931). Five generations of Spectacles hardware, a purpose-built Snap OS, and proprietary on-device ML models represent 6-18 months of re-derivation cost for a competitor starting from scratch.

But nobody is starting from scratch. Apple ARKit (2017), Google ARCore (2018), and Meta Spark (2017) already provide comparable or superior AR capabilities. The patents protect implementation, not the capability itself.

More critically, the AR stack generates less than 5% of revenue. The revenue-generating core — ephemeral messaging, Stories, Spotlight, ad delivery — has a re-derivation cost measured in weeks. WhatsApp, iMessage, and Telegram replicate messaging. Instagram Stories replicates ephemeral content. TikTok's recommendation algorithm is superior to Spotlight.

SNAP's own AI chatbot ("My AI") runs on third-party LLMs with a Perplexity integration (10-K lines 2969-2970, 7110). The company reports that 40% of new code is AI-generated (Q4 transcript line 212) — confirming that the tools commoditizing software are already inside SNAP's own engineering process.

C=2: compound cognition exists in the AR periphery but is non-load-bearing. The revenue core is weeks-derivable.

E = 2 | Irreducible Infrastructure

The dominant discriminator — and the kill shot. In calibration data, E shows a 3.8-point gap between dead companies (avg E=0.4) and survivors (avg E=4.2). SNAP's E=2 places it on the wrong side.

"Substantially all of our network infrastructure is provided by third parties, including Google Cloud and AWS" (10-K line 2967). This is Tier 1 evidence — management's disclosure under penalty of fraud. SNAP does not own data centers, does not own network infrastructure, and has committed $3.4 billion in non-cancelable cloud payments to Google and Amazon (10-K lines 7794-7801). The infrastructure dependency flows the wrong way: SNAP is locked into its vendors, not customers locked into SNAP.

No task in SNAP's domain has c_ℓ = ∞. Every function — messaging, content, ad targeting, AR, AI chat — is locally computable. The social graph creates apparent irreducibility, but that is M-type resistance (gravity from counterparty coordination), not E-type resistance (infrastructure that cannot be replicated). Any server can route messages; only SNAP's server has your friend graph.

No regulatory mandate requires use of SNAP. The FTC consent decree and FDA Spectacles regulation constrain SNAP's behavior but do not mandate that any entity route through SNAP.

Revenue durability is thin. NRR/GRR is not disclosed because advertising is transactional: "Most advertisers do not have long-term advertising commitments with us" (10-K line 1175). Advertising represents approximately 87% of total revenue.

E=2 is the ceiling, not the floor. The score survives only because an ad marketplace technically requires centralized clearing (buyers meet sellers at an exchange). If that function is judged locally computable — agents managing cross-platform ad buys directly — E drops to 1 and V collapses to zero via Gate 1.

U = 2 | Ecosystem Breadth

SNAP offers approximately 10 product surfaces: Camera, Visual Messaging, Snap Map, Stories, Spotlight, Subscriptions, Spectacles, AI Platform, Advertising, and Developer Tools (10-K lines 559-635). These serve three user types: consumer, marketer, and creator/developer.

Every individual workflow is replicable elsewhere. Messaging maps to iMessage, WhatsApp, and Telegram. Stories maps to Instagram Stories. Spotlight maps to TikTok and Reels. Snap Map maps to Google Maps and Life360. My AI maps to ChatGPT, Gemini, and Perplexity directly.

There is no enterprise breadth. No horizontal platform play. No cross-departmental data flows creating superlinear switching cost. SNAP serves one domain (consumer social media) with features that share a single user type doing related activities. Contrast with U=3 calibration examples like Atlassian (Jira for engineering, Confluence for documentation, JSM for IT support) or Palo Alto (firewall, SIEM, SOAR, endpoint) — platforms serving different departments with distinct data needs.

A = 2 | Distribution

P(agent encounters SNAP first) ≈ 0 for all major tasks. For messaging, agents route to Slack, Teams, or email. For ad buying, agents route to Google Ads (50% market share) or Meta Ads (25%). For content distribution, agents route to YouTube, TikTok, or Instagram.

Camera-first, mobile-first design is structurally anti-agent. AI agents cannot take photos, cannot interact with AR lenses, and cannot experience visual messaging. As intelligence routes through text-based and API-first infrastructure, SNAP's visual interaction paradigm becomes a structural liability. The compounding flywheel (Ȧ = β·A) runs in reverse: β < 0.

SNAP's Perplexity deal ($387.9M through Q1 2027, 10-K lines 7110-7113) positions SNAP as a distribution channel for OTHER agents — agents consuming SNAP's eyeballs, not agents routing through SNAP's infrastructure. The direction of flow matters. SNAP consumes AI services; agents do not consume SNAP services.

Camera Kit (68M MAU, Q3 transcript line 70) and the self-serve ad API prevent A from dropping to 1. Agents can technically interact with SNAP. They just have no reason to prefer it.

M = 3 | Ecosystem Gravity

The strongest dimension and the only score above 2. The 474M DAU social graph creates genuine gravitational mass: friend connections, Streaks (years-long coordination commitments), 1T+ Memories, 450K+ creators with 5M+ Lenses, 200M+ monthly gamers, and 24M Snapchat+ subscribers growing 71% year-over-year.

Then management tells you the answer: "Snapchat is free and easy to join, the barrier to entry for new entrants in our business is low, and the switching costs to another platform are also low" (10-K line 1084-1085). Tier 1.

North America DAU is declining 5% year-over-year while growth shifts to Rest of World markets averaging $1.29 ARPU versus North America's $9.80 (10-K lines 5103-5131). The gravity is eroding in the market that generates the revenue.

The demographic amplifies the risk: "majority of our users are 18-34 years old. This demographic may be less brand loyal and more likely to follow trends" (10-K lines 1086-1087).

Social graph gravity is historically the most fragile form of M. MySpace had 115M users when it peaked. Network effects in consumer social create tipping-point dynamics — the gravity decays nonlinearly when friends start leaving.

F = 2 | Friction (Penalty: −0.12)

Low friction. The consumer app is intuitive. The self-serve ad platform is improving: "simplified onboarding, improved campaign workflows" (Q4 transcript line 106). Camera Kit SDK is clean and no longer requires mandatory Snapchat branding (Q3 transcript line 70). SNAP is actively integrating with Wix and WooCommerce for e-commerce advertisers.

Moderate friction from mobile-only design, camera-first UX that is unusual versus competitors, and absence of enterprise tools. No admin dashboard, no SSO, no CRM integration.

Thermodynamic Summary

A1 (Intelligence Conservation): Intelligence flows to minimum energy. For every task in SNAP's domain, cheaper alternatives exist or are emerging:

Tool Death Theorem applies. 𝒟(SNAP) consists entirely of computable tasks. c_ℓ(τ,t) = g(τ)/M(t) → 0 as model capability M → ∞. Since c_s ≥ c_s_min > 0 (anchored by $2.67B cost of revenue and $3.4B in cloud commitments), there exists t* such that c_ℓ < c_s for all τ.

Survival condition (∃τ with c_ℓ = ∞): Not met. The social graph is the only candidate, but social graphs are network effects (M-type resistance), not computational barriers (E-type resistance). They create gravity, not irreducibility. And gravity erodes — NA DAU is already declining.

Durable revenue: ≈25%. Snapchat+ (24M subscribers, +71% YoY) and the AR creator ecosystem have moderate stickiness from social graph inertia and creator investment. Exposed revenue: ≈75%. Advertising (≈87% of total) is transactional with zero contract lock-in and smaller, less-targeted inventory than Meta or Google.

What Would Change This

To escape COLLAPSED requires C or E reaching 3. Both paths depend on Spectacles achieving consumer product-market fit — speculative after five generations and nine years without a consumer launch.

Conviction and Basket Verdict

V    = 1.87       COLLAPSED
κ    = 0.00       (V − 3.0)⁺ = 0 — below KEEP threshold
IR   = −2.278     Regime-contaminated (ρ_intra = 0.68, α t-stat = −1.23 NS)
δ    ≈ 0          Market converging to structural reality, not overshooting
w    = 0          No position in basket

IR does not gate the verdict. The negative IR reflects a sector-wide risk-off regime (ρ_intra rose from 0.10 to 0.68 in 12 weeks) compounded by SNAP-specific structural repricing (76% of the 15-week decline is idiosyncratic). V-Score is regime-invariant — scored against structural properties, not price. The IR confirms that the market is discovering what the structure already shows.

No edge. δ ≈ 0 means the market's implied structural assessment approximately matches V = 1.87. Forward P/E of 8.16x on breakeven EPS history, with the entire C-suite selling into weakness (CTO: $10.6M, GC: $958K, CFO: $723K in Feb-Mar 2026), is consistent with DEAD ZONE pricing. There is no mispricing to exploit.

COLLAPSED. No conviction. No weight. Pass.

Sources: SNAP 10-K (filed 2026-02-05), SNAP 10-Q (filed 2025-11-06), Q3 2025 earnings transcript (2025-11-05), Q4 2025 earnings transcript (2026-02-04), yfinance market data (2026-04-08). All dimension scores validated against primary source citations.