V-Score Card

TICKER:           CCCS (CCC Intelligent Solutions)
V-SCORE:          3.65
VERDICT:          EMBEDDED
κ (conviction):   0.65
BASKET:           KEEP

DIMENSIONS:
  C = 4  (w=0.25)  →  1.00    ← CHALLENGED from 5 (data licensed, not proprietary)
  E = 4  (w=0.22)  →  0.88    ← DOWNGRADED from 5 (no regulatory mandate)
  U = 4  (w=0.18)  →  0.72
  A = 4  (w=0.12)  →  0.48
  M = 5  (w=0.15)  →  0.75
  F = 3  (w=−0.06) → −0.18
  ─────────────────────────
  RAW                  3.65
  × G1 (1) × G2 (1) = 3.65

GATES:
  G1 (E > 1):          PASS (E = 4)
  G2 (A > 1 ∨ Σ ≥ 12): PASS (A = 4; C+E+U = 12)
  Bustamante:           3/3 (proprietary data, structural lock-in, transaction-embedded)

Score evolution: Research Agent scored V = 4.12 (FORTRESS). Strict rubric applied E: 5 → 4 (no regulatory mandate), bringing V to 3.90. Adversarial challenge applied C: 5 → 4 (base estimating data licensed from MOTOR, competitor Mitchell proves re-derivability), bringing V to 3.65.

Dimension Analysis

C — Compound Cognition: 4 (challenged from 5)

CCC has operated in the P&C claims domain for 45 years with 30 interconnected workflows spanning estimating, DRP routing, subrogation, casualty, parts ordering, and shop management. The system encodes 62,000 insurer-specific audit rules across 13,000 jurisdictions, 2M labor rate profiles, 7.4M part SKUs, and 5.5B live part quotes, processing 5.7B database transactions daily.

Why 4, not 5. The adversarial review identified three structural issues:

First, CCC's base estimating data — labor times and parts pricing — is licensed from MOTOR Information Systems (Hearst), not proprietary. Mitchell built an equivalent proprietary database in-house starting in 1946. Audatex has its own (DBRM). The foundational data layer that the research scored as "decades of accumulated knowledge" is available to any licensee.

Second, the claimed 5 compounding layers conflate network infrastructure (E) and data gravity (M) with cognition. Layers 2 (network connections), 3 (transaction routing), and 5 ($2T historical data) are properly scored under E and M. Only the AI layer (400+ models) and insurer rule configurations represent genuine crystallized cognition.

Third, Mitchell's competitive existence empirically disproves the "15-20 year re-derivation" claim. Mitchell processes claims for 300+ insurers and 20,000+ shops with its own estimating database, DRP management, and AI capabilities. The three largest US MSO chains — Caliber (1,600 locations), Crash Champions (650), Classic Collision (300) — all standardized on Mitchell Cloud Estimating between 2023-2025.

CCC's genuine cognitive moat: the AI model layer (400+ models trained on the network's transaction data), workflow orchestration logic, and the integration layer connecting 30 products. Re-derivation: 5-7 years with equivalent data access, not decades. This is consistent with C = 4 (strong domain encoding, re-derivation measured in years, multiple compounding layers — but not the "superlinear, decades-to-replicate" threshold of C = 5).

Sources: 10-K L249, L422, L461; Q4 transcript L58-59; Q3 transcript L34-35; CIC Estimating Committee Feb 2024; Mitchell/Caliber/Crash Champions PRs 2023-2025.

E — Irreducible Infrastructure: 4

CCC operates the dominant bilateral network connecting insurers and collision repair shops in the US: 300+ insurers (27/30 top auto carriers), 30,500+ repair facilities, 6,000+ parts suppliers, and 14/15 top OEMs. The DRP (Direct Repair Program) network is the mechanism by which insurers route claims to approved shops — CCC describes itself as the "architectural backbone of insurance DRP systems."

Coordination tasks — DRP routing, multi-party settlement, real-time parts ordering, cross-network performance measurement — require the bilateral network and cannot be performed by a local AI model operating on isolated data. These represent approximately 75-80% of CCC's revenue.

Why 4, not 5. No explicit regulatory mandate names CCC. Unlike Verisk (ISO statistical agent designation in 50 states) or Fiserv (Durbin Amendment, FFIEC examination), CCC's lock-in is structural and contractual, not regulatory. The bilateral network creates high switching costs — estimated at 20-50x operating costs — but these are finite, not infinite.

Mitchell provides a viable (if smaller-scale) alternative for all coordination tasks CCC performs. State Farm, the largest US private auto insurer, reversed its CCC-only mandate in September 2025 and now allows Select Service shops to choose between CCC and Mitchell nationwide. Multi-homing exists: 30% of shops ran multiple estimating systems (2019 survey, declining from 34% in 2016).

Key distinction: c_ℓ is "very high" for coordination tasks (20-50x c_s), but not literally infinite. Mitchell does the same coordination tasks on a parallel network. The 99% GDR proves extreme stickiness, but stickiness and irreducibility are different concepts.

Sources: 10-K L250, L256, L287-288; Q4 transcript L53, L57, L145; CollisionWeek Sept 2025; Repairer Driven News May 2021.

U — Ecosystem Breadth: 4

30 distinct solutions across insurance (13 products: workflow, estimating, reinspection, total loss, subrogation, casualty, disability, workers' comp), repair (12 products: estimating, DRP management, workflow, parts ordering, payments, diagnostics), and ecosystem (5 products: parts e-commerce, OEM programs, salvage, lender, international). Spans 4+ departments within the claims/repair value chain.

Cross-module data flows create superlinear switching costs: removing estimating breaks reinspection, subrogation, and parts ordering downstream. IX Cloud event-based architecture connects 35,000+ businesses.

Why 4, not 5. Deep within its vertical but narrower than SAP (U = 5), which spans every enterprise department. CCC touches claims + repair operations across insurance and shops. Does not cover underwriting, policy admin, distribution, marketing, or HR.

Sources: 10-K L540-750; Q3 transcript L34.

A — Distribution & Discoverability: 4

Within the P&C claims domain, CCC is the default routing path: 27/30 top auto carriers, 30,500+ shops, 600M interface transactions per year, 900,000 registered users. An AI agent processing an auto insurance claim encounters CCC through the DRP network structurally, not through API discovery.

AI-specific revenue: ≈$100M (≈10% of total), growing 70%+ YoY. 400+ AI models. Management working with "seven different" frontier AI companies. New CPO (Valdez, from Dayforce/Workday) hired explicitly to scale AI-driven innovation.

Why 4, not 5. No public API ecosystem, no developer marketplace, no self-serve agent integration. CCC's routing is domain-implicit (you encounter CCC because the DRP network exists), not agent-discoverable (no API for external agents to find and call CCC services). ServiceNow (A = 5) has the developer ecosystem CCC lacks.

Sources: 10-K L287, L422, L449-450, L461; Q4 transcript L42, L50, L128, L141-143; Q3 transcript L90, L148.

M — Ecosystem Gravity: 5

Bilateral gravity — the dominant feature. Migration cost is not just for the migrating party but imposes costs on all counterparties: if State Farm leaves CCC, 30,500 shops lose State Farm connectivity; if Caliber leaves CCC, 300+ insurers lose routing to Caliber's 1,600 locations.

Accumulated state: $2T historical claims data, 62,000 insurer-specific audit rules, 200,000+ insurer-to-shop relationships with performance history, decades of calibrated estimating data across 13,000 jurisdictions. Migration cost decomposition spans data ($2T non-exportable), integrations (600M annual transactions), retraining (900K users across 35K companies), and counterparty re-coordination (200K+ relationships).

No customer exceeds 10% of revenue. 35,000+ total customers. The network as a whole cannot migrate — only individual small players can leave (99% GDR = 1% annual churn). No competitor has attempted to build a parallel bilateral network at CCC's scale.

Sources: 10-K L269, L287-288, L449, L1121; Q4 transcript L34, L59.

F — Ecosystem Friction: 3 (penalty)

Cloud-native, multi-tenant architecture. Mobile-first for shops (CCC ONE). 1,700 software releases in 2025. Enterprise insurer implementations are inherently complex — regulated industry, insurer-specific audit rules, multi-party workflows — but CCC manages its own deployments without Big 4 consultant dependency (unlike SAP). Actively streamlining packages and selling motions.

No public developer API/marketplace adds friction for external integration.

Sources: 10-K L255, L426, L736; Q4 transcript L76, L98; Q3 transcript L40, L166.

Thermodynamic Summary

The Tool Death Theorem predicts SaaS collapse when local AI cost c_ℓ(t) falls below centralized cost c_s(t) for all tasks in the software's domain. For CCC:

Coordination tasks (c_ℓ >> c_s, ≈75-80% of revenue): DRP routing, multi-party claim settlement, real-time parts ordering across 6,000+ suppliers, cross-network performance tracking. These require the bilateral network to exist. A local AI model cannot route a claim to a shop it has no connection to. c_ℓ is not infinite (Mitchell proves alternatives exist) but is 20-50x c_s — switching cost measured in years and millions, not hours and dollars.

Computable tasks (c_ℓ → c_s, ≈10-15% of revenue): Photo-based damage assessment, document summarization (Medhub), basic claim triage. Frontier vision models are approaching local capability. CCC is self-cannibalizing this segment by selling AI products into the network (Intelligent Estimating, Mobile Jumpstart).

Configuration tasks (≈10% of revenue): Insurer rule setup, workflow configuration, implementation consulting. AI-assistable but context-dependent.

Net resistance: R(s,t) ≈ 0.75-0.80 as t → ∞. Revised down from initial 0.85 estimate after adversarial challenge on c_ℓ (Mitchell demonstrates coordination tasks are high-cost-to-switch, not infinite-cost). CCC survives because its core function is orchestration across a bilateral network, not computation on isolated data.

Kill cycle: t* = ∞ for network-embedded tasks (no AI replaces the network itself). t* ≈ 3-7 years for computable tasks (15% of revenue). CCC captures much of this disruption value itself through its AI product suite.

Regime Context

15-week factor regression (Dec 16, 2025 — Mar 27, 2026):

CCC ~ SPY + IGV (70 trading days)

α̂ (annualized):     +50.3%   (t = 0.43, p = 0.666)
β (SPY):              0.178
γ (IGV):              1.096
R²:                   0.269
σ_total (ann):       68.9%
σ_idio (ann):        58.9%
%Idio variance:      73.1%

IR_i = α̂ / σ_idio = 0.85   ← STATISTICALLY INSIGNIFICANT (p = 0.666)

Return decomposition:

Total return:        −23.1%
Factor return:       −34.1%   (β × SPY + γ × IGV)
Idio return:           0.0%

100% of the 15-week move is sector. 0% is CCC-specific.

Intra-sector correlation:

ρ(CCC, VRSK):   0.315
ρ(CCC, GWRE):   0.448
ρ(VRSK, GWRE):  0.398
ρ_intra (avg):   0.387

ρ(CCC, IGV):    0.518

ρ_intra = 0.387 — moderate, not extreme. This is not a fully indiscriminate selloff where ρ → 1 and ε → 0. But the sector factor (γ = 1.096) is doing nearly all the work. The 15-week measurement window contains no distinguishable idiosyncratic signal. IR measures the regime, not the stock.

Market-implied V vs structural V:

At forward P/E 12x, EV/Revenue 4.6x, and 7.4% FCF yield, the market prices CCC as if its network moat is eroding. Comparable pricing implies V_market ≈ 2.0-2.5 (AT_RISK to weak EMBEDDED). Structural analysis finds V = 3.65 (solid EMBEDDED).

δ_i = V_i − V_market,i ≈ 3.65 − 2.25 = 1.40

The delta is the story: the market applies a uniform software discount during the selloff (IGV −27%), but CCC's structural properties — bilateral network, 99% GDR, 30 interconnected workflows, $277M FCF — are orthogonal to a 15-week sector drawdown.

Conviction Weight

κ_i = (V_i − 3.0)⁺ = (3.65 − 3.0)⁺ = 0.65

w_i = W_S × κ_i / Σ_j κ_j   (tenant normalizes across basket)

κ = 0.65 is regime-invariant. It comes from structural properties scored against primary sources (10-K, transcripts, competitor evidence), not from price action or factor returns. The software selloff does not change the bilateral network, the 99% GDR, or the 30 interconnected workflows.

The challenged C-score (5 → 4) reduced κ from 0.90 to 0.65. This reflects real structural findings: CCC's base estimating data is licensed from MOTOR (not proprietary), Mitchell proves the cognitive layer is re-derivable, and 3 of 5 claimed knowledge layers properly belong to E/M (network and gravity). The moat is real but shallower than first scored.

Basket Verdict: KEEP

V = 3.65 > 3.0 → EMBEDDED → passes survival filter.

Bear case acknowledged: Mitchell gaining MSO share (2,500+ locations), State Farm opening alternatives, base data licensed not proprietary. But no insurer has fully dropped CCC, 99% GDR holds, bilateral network effects dominate, and CCC self-cannibalizes the AI-exposed 15% through its own products.

The market-structural gap δ ≈ 1.40 is among the widest in the basket. At 7.4% FCF yield with 10.6% revenue CAGR and 99% GDR, the market prices an AI-kills-network scenario that the evidence — particularly the coordination/orchestration core — does not support.

Evidence Table