V-Score Card

TICKER:             SAP
V-SCORE:            3.90
VERDICT:            EMBEDDED (upper bound)
κ (conviction):     (3.90 − 3.0)⁺ = 0.90
GATE 1 (E>1):      PASS (E=4)
GATE 2 (A>1∨Σ≥12): PASS (C+E+U = 14 ≥ 12)
FAST SCREEN:        3/3 — Proprietary data ✓, Regulatory lock-in ✓, Transaction embedding ✓

Dimensions

Calculation

V = 0.25(5) + 0.22(4) + 0.18(5) + 0.12(3) + 0.15(5) − 0.06(4)
  = 1.25   + 0.88   + 0.90   + 0.36   + 0.75   − 0.24
  = 3.90

Gates: 𝟙[E>1] · 𝟙[(C+E+U)≥12] = 1 · 1 = 1
V = 3.90 × 1 = 3.90

Prior Adjudication Delta

Dimension Deep Dives

E-Score Stress Test: Is the Infrastructure Real?

Challenge: "Nobody leaves" is not the same as "law requires them to stay." SAP's irreversibility is practical (switching cost, data gravity, consultant dependency) not legal (no NRSRO, no clearing mandate, no regulatory designation).

Evidence that E=4 holds (not E=3):

The switching cost hierarchy is finite but steep:

Zero Fortune 500 core ERP defections in 3,962 earnings transcripts searched. Oracle and Workday would prominently cite large SAP displacement wins if they occurred. The absence is the signal.

E=3 threshold: "replicable in 2-3 years." The evidence says 3-6 years minimum for large enterprise, with zero completed examples. That's E=4.

What would change E: First Fortune 500 core ERP defection (95% probability none by Dec 2026). Celonis trial outcome Mar 2027 (forced open APIs could erode analytical moat, but doesn't touch core ERP switching cost). AI migration tools compressing the 3-6yr timeline below 2yr.

C-Score Stress Test: Will Frontier Models Re-Derive It?

Challenge: The rubric was written before Claude 4 and GPT-5. Can frontier models make 53 years of crystallized cognition re-derivable?

Decomposition of SAP's cognition stack:

A frontier model in 2028 can plausibly handle L1+L2 (85%). The question is whether L3+L4 (15%) is load-bearing.

Answer: Yes, for the revenue that matters. Cloud ERP Suite (EUR 18.1B, 86% of cloud revenue) is concentrated in large enterprises where L3 is essential. ExxonMobil consolidated 10+ ERP systems with 65M lines of custom code to single S/4HANA. That's L3 cognition — 97% fewer profit centers, 70% fewer cost centers, German chemical plant reporting + Nigerian transfer pricing + Japanese quality standards in one system.

Mid-market vulnerability is real but doesn't change C. Companies operating in 3-5 countries don't need L3. Microsoft Dynamics 365 growing "high teens" is the evidence. But V-Score measures structural survival of the company, not growth optionality. SAP's core revenue base requires the full cognition stack.

C=5 holds. The agent-reads-the-law thesis fails on: (1) audit trail/SOX compliance requires process, not just computation; (2) cross-module interaction effects can't be derived module-by-module; (3) at 99% accuracy, a Fortune 500 with 10M transactions gets 100K errors. SAP's production-tested cognition is orders of magnitude more reliable for covered edge cases.

Sensitivity: What If Both Scores Shift?

Even the adversarial worst case (both dimensions downgraded) stays EMBEDDED with Gate 2 passing at exactly 12. κ drops from 0.90 to 0.43 — a 52% conviction haircut — but the basket inclusion verdict is unchanged.

Regime Analysis (T = 15 Weeks)

REGRESSION: SAP ~ SPY + IGV  (Dec 13, 2025 → Mar 28, 2026, n=70)

  α̂ = -39.1% ann     (p = 0.535 — NOT SIGNIFICANT)
  β_SPY = -0.20       (p = 0.594 — NOT SIGNIFICANT)
  β_IGV =  0.92       (p = 0.000 — DOMINANT FACTOR)
  R² = 0.43
  σ_idio = 31.8% ann

  IR = α̂ / σ_idio = -1.23    ← REGIME-CONTAMINATED, DO NOT GATE VERDICT
  ρ_intra = 0.51              ← ELEVATED (risk-off block-selling)
  %Idio variance = 57%        ← BELOW 75% TARGET (regime effect)

Interpretation: SAP moves 0.92:1 with IGV. β_SPY is zero after controlling for software sector. The trailing α is statistically indistinguishable from zero — the entire return is explained by sector contagion. The 15-week window contains no idiosyncratic signal because ρ_intra = 0.51 is attenuating all stock-specific information.

IR = -1.23 measures the regime, not SAP. When ρ_intra normalizes, idio signal reasserts.

Four-factor selloff decomposition:

1. Q4 CCB miss (Jan 29): -16% single day. 1pp shortfall, "more pronounced than anticipated" (Asam Q4 L78)
2. SaaSpocalypse sector rout: IGV -27% in window. AI agent seat compression fears
3. Iran war / macro shock (Feb 28+): Oil $107, European recession risk, 46% revenue exposed
4. JPMorgan downgrade (Mar 23): OW→N, PT EUR 260→175

None of the four factors invalidate structural dimensions. EUR 77B backlog intact. Zero defections. ECC migration tailwind continues. V = 3.90 is unchanged while price dropped 33%.

Cumulative returns in window:

Thermodynamic Summary

SAP is the transactional backbone of the global economy — GL, payroll, supply chain, BOM for 400K+ customers across 22 industries in 100+ countries. Intelligence cannot flow around SAP because the structured business data (petabytes per enterprise), cross-module process logic (25+ departments), and regulatory compliance encoding (53 years, 100+ jurisdictions) create an energy well too deep to escape. The switching cost is not infinite but it is economically irrational: $100M-$1B and 3-6 years for core ERP replacement, with zero completed Fortune 500 examples in the public record.

The E downgrade (5→4) is the key analytical move: practical irreversibility is not legal mandate. SAP sits 0.10 below FORTRESS. The path back requires A improvement (Joule closes coverage gaps, MCP adoption, standalone AI revenue → A=4 adds +0.12 → V=4.02). The path to erosion requires either a Fortune 500 defection (E→3) or frontier model re-derivation of the regulatory tail (C→4) — neither is probable within 24 months, but both merit monitoring.

Durable revenue: ≈90%. Cloud ERP Suite (EUR 18.1B) + Software Support (EUR 10.5B) protected by switching cost, contract commitment, and ECC migration forcing function.

Exposed revenue: ≈10%. Extension Suite (EUR 2.6B, ServiceNow encroachment at interface layer) + IaaS (EUR 0.3B, declining) + mid-market greenfield (GROW segment, frontier model competition by 2028).

Conviction Weight

κ = (V − 3.0)⁺ = (3.90 − 3.0)⁺ = 0.90

w_SAP ∝ κ = 0.90  (normalized: w_SAP = W_S · 0.90 / Σ_j κ_j)

κ is regime-invariant. IR = -1.23 does not gate the verdict — it confirms the regime (risk-off, ρ_intra = 0.51, sector block-selling). The structural signal (V = 3.90, δ = +0.40) is orthogonal to the trailing price.

Edge: δ = V − V_universe = 3.90 − 3.50 = +0.40. Market applies uniform software discount during selloff. SAP's structural advantage survives the discount. Maximum δ at maximum indiscriminate selloff = optimal entry timing.

Basket Verdict

KEEP. V = 3.90 (EMBEDDED) → include at target weight. κ = 0.90 → near-maximum conviction within EMBEDDED tier.

Catalyst timeline: Q1 earnings Apr 23. Cloud revenue and CCB guidance will reveal whether spending freeze translates to actual revenue miss. If guidance holds → idio signal reasserts as ρ_intra normalizes. If misses → SURVIVAL factor activates, re-score F dimension.

Predictions tracking: