V-Score Card

V = 0.25(C) + 0.22(E) + 0.18(U) + 0.12(A) + 0.15(M) − 0.06(F)
  = 0.25(5) + 0.22(4) + 0.18(5) + 0.12(3) + 0.15(4) − 0.06(3)
  = 1.25   + 0.88   + 0.90   + 0.36   + 0.60   − 0.18
  = 3.81

VERDICT:         EMBEDDED (upper range)
κ:               (3.81 − 3.0)⁺ = 0.81
GATE 1 (E>1):   PASS (E=4)
GATE 2 (A>1∨Σ≥12): PASS (A=3; C+E+U=14≥12)
FAST SCREEN:     3/3 — proprietary data, regulatory lock-in, transaction-embedded
BASKET:          KEEP

Regime Context (T = 15 weeks)

IR_i  = α̂/σ_idio = −1.94  (α t-stat = −1.02, p = 0.31 — NOT significant)
ρ_intra           = 0.49   (elevated since Feb — sector-driven selloff)
%Idio Variance    = 66%    (below 75% target — factor-dominated window)
δ_i = V_struct − V_mkt ≈ 0.8–1.3  (market prices TYL as generic software)

IR does not gate. Every name in the software comp set shows negative IR over this window (range: −5.73 to −0.53, all 8/8 negative). Only one (NOW) is statistically significant. The regression is measuring the regime — a sector-wide selloff where ρ_intra rose from 0.38 in January to 0.51–0.54 in February–March and stayed there. When the sector factor dominates, idiosyncratic residuals compress toward zero and IR becomes noise.

V(s) is orthogonal to r_sector(t). The structural score is invariant to whether the sector is up 30% or down 30%. κ = 0.81 is the sizing input.

Dimension Analysis

C = 5 | Compound Cognition | w = 0.25

25+ years of exclusive public sector domain encoding across 9 functional areas × 50 states × thousands of jurisdictions. Every jurisdiction has different tax codes, court procedures, property assessment rules, and public safety reporting mandates. Tyler has encoded continuous legislative changes across all of these into software maintained through releases "necessary because of legislative or regulatory changes" (10-K L425–427).

Inter-module dependencies are superlinear. The justice continuum (Courts → Prosecutors → Defenders → Corrections → Probation) and financial ecosystem (ERP → Tax → Property → Utilities → Payments) create compound re-derivation barriers where each module's regulatory encoding depends on data flows from other modules (10-K L344–349, L326–327).

Institutional knowledge retention: 7% voluntary turnover vs 13–15% industry norm, 8-year average tenure, 30% with 10+ years, 91% workforce redeployment rate (10-K L689–735). These 7,800 employees ARE the crystallized cognition.

Adversarial check (B₀): Legal/regulatory knowledge is public and partially AI-derivable (3–8 years for the 70% following model codes). But tacit operational knowledge — "how government actually works" vs what the statute says — requires deployment experience across thousands of jurisdictions (10–15 years minimum). And the zero-error tolerance for legally binding outputs (tax bills, court orders) means "approximately correct" doesn't cut it. Re-derivation timeline compresses from 15–25 to 7–12 years with AI tools. Still astronomical. C = 5 confirmed.

E = 4 | Irreducible Infrastructure | w = 0.22

98% logo retention — "historically very low client turnover, approximately 2% annually" (10-K L530–531, L2184). ARR $2.06B growing 10.9%, implied NRR ≈112–113%. Maintenance auto-renews unless termination notice given (10-K L435–437). 89% of new contracts are subscription-based (10-K L2502).

Regulatory barriers: CJIS compliance for public safety, court records under state/federal requirements, property tax lifecycle governed by state law, Arizona Supreme Court preferred provider for all 170 courts (Q3 2024 transcript). Data portability: none — data conversion is a standalone professional services engagement (10-K L444–448).

Cloud flip irreversibility: ACV +64.5% YoY in Q4 2025, peak expected 2027–2029 (CFO Miller). Once flipped to Tyler cloud on AWS, no practical exit path — the integration depth increases, not decreases.

Adversarial check (B₀): 2% annual churn proves c_ℓ ≠ ∞. Switching cost is ≈5–10× annual contract value, not infinite. No hard federal mandate requiring Tyler specifically. Cloud hosting on commodity AWS infrastructure is not proprietary physical plant. But government procurement friction (12–24 month RFP cycles) is structural — AI doesn't speed up county commission votes. And cross-department compound switching costs (24 interconnected workflows) create superlinear exit barriers. E = 4 confirmed. Not E = 5 (no legal mandate), clearly above E = 3 (practical regulatory barriers are real).

U = 5 | Ecosystem Breadth | w = 0.18

24 distinct workflows enumerated from 10-K across every government department:

Cross-sell: current 2–3 products per client, target 10–12 (CEO Moore Q3 2025). "Our existing client base offers significant opportunities for additional sales of solutions and services that we currently offer, but that existing clients do not fully utilize" (10-K L518–520).

φ_switch ∝ |𝒲|^α, α > 1. Replacing 24 interconnected workflows simultaneously is qualitatively impossible within normal government procurement cycles.

A = 3 | Distribution & Discoverability | w = 0.12

Tyler Resident AI live in 6 states (Alabama, Hawaii, Indiana, Mississippi, Nebraska, South Carolina), ≈17K MAU in Indiana alone (Q4 2025). Agentic AI in early access Q1 2026 for permitting/licensing and supervision. R&D ramped 73% to $205M (8.8% of revenue), headcount +57% to 1,368 (10-K L2773–2782). AWS strategic collaboration, OpenAI + Anthropic partnerships named.

But: no public API marketplace, approach self-described as "measured" (10-K L1079), government sector inherently slow on AI adoption. Agents don't naturally route through Tyler. A = 3 is right: functional API, not default.

Catalyst watch: Investor Day June 9, 2026 in Frisco TX — will detail "purpose-built AI strategy." If agentic deployment succeeds and agents route through Tyler for gov-tech workflows → A = 4, V → 3.93.

M = 4 | Ecosystem Gravity | w = 0.15

45,000+ installations, 15,000+ locations, all 50 states. Enterprise contracts in 30 states with dedicated offices (10-K L192–194). 98% logo retention. No single client >10% of revenue. NIC payments flywheel: Tyler back-office clients → NIC payments; NIC payments clients → Tyler software (10-K L522–525). Serial acquirer: 11 acquisitions since NIC for ≈$400M.

CEO Moore positioning PE-owned competitors as vulnerable: "high debt, wondering what happened to multiples" (Q4 2025). Tyler: $1.67B total liquidity, $621M FCF, $1B buyback authorization.

Not M = 5: network effects are primarily internal (cross-department within jurisdiction), not counterparty. Government is one-sided customer, not 2-sided marketplace.

F = 3 | Ecosystem Friction | w = −0.06

Standard enterprise learning curve, declining. Pro services 10.4% of revenue (down from 12.3%), intentionally shrinking — CFO calls it "low negative margin business." Cloud-first since 2019 reducing per-client friction. Low-code platform enables gov workers to build applications. Tyler's own 7,800 staff deliver services, not consultant-dependent.

Government procurement friction is external (RFP cycles, budget appropriations), not self-imposed. F = 3: moderate, constructive.

Thermodynamic Summary

Government IS the transaction. You cannot process a court case, bill property taxes, manage criminal justice workflows, or accept government payments without an authorized, auditable, jurisdiction-specific system of record. Tyler IS that system for 15,000+ locations across 24 interconnected workflows spanning every government department.

What prevents intelligence from flowing around TYL: Regulatory authority + jurisdictional encoding + cross-departmental integration + transaction embedding. An AI agent can help a government worker USE Tyler. It cannot REPLACE Tyler. The system of record is the moat.

Revenue protection: 87% durable (SaaS 33.4% + transactions 34.7% + maintenance 19.1%), 13% exposed (pro services 10.4%, declining toward ≈5%). Revenue quality improving structurally toward 90%+ durable within 2–3 years.

Conviction Weight

κ = (V − 3.0)⁺ = (3.81 − 3.0)⁺ = 0.81

w_TYL = W_S · κ_TYL / Σ_j κ_j   (basket normalizes)

κ is a pure structural signal. It does not depend on the 15-week factor regression, the sector selloff, or the measured IR. The structural moat properties that generate κ = 0.81 — 25 years of crystallized cognition, 98% GRR, 24 interconnected workflows, practical regulatory barriers — exist independent of whether software is up or down this quarter.

Sensitivity

Gate-proof: even worst-case C=4, E=3 yields V = 3.34, C+E+U = 12 ≥ 12 → Gate 2 PASS. Score cannot be killed by any single dimension downgrade.

Basket Verdict

KEEP. V = 3.81, EMBEDDED.

The market applies a uniform software discount (δ ≈ 0.8–1.3). V discriminates structurally. Tyler's system-of-record moat in government — crystallized cognition, irreducible infrastructure, 24-workflow ecosystem breadth — is orthogonal to whether IGV is down 29% over 15 weeks.

Next catalyst: Q1 2026 earnings April 29. Investor Day June 9 (A dimension update). Cloud flip peak 2027–2029 (E dimension strengthening). C-suite insider buying cluster $20M+ on Feb 27 at ≈$334 — Tier 1 evidence (LR 5–10) of management conviction at current levels.

Evidence base: TYL 10-K FY2025 (filed 2026-02-18), Q4/Q3 2025 earnings transcripts, Form 4 insider filings, worldview (6 evidence items, cumulative LR 20.56). Adversarial review (B₀) stress-tested E and C — both confirmed at scored levels.