A 50-year proprietary data flywheel embedded in every top-100 P&C insurer's underwriting workflow, with 80%+ subscription revenue and a 15.6% 10-year average ROIC. Trading at 0.73x base IV with bull-case upside to $312, but starting from a 26x P/E that demands continued execution.

Plain English

Verisk is the data utility of U.S. property and casualty insurance. Since 1971, every major insurer has fed it claims and premium records. In return, Verisk publishes the standard rates, policy language, and risk models the whole industry uses. All top 100 U.S. insurers pay annual subscriptions because nobody can rebuild a 50-year database. It is a toll booth on insurance pricing, with about 80% recurring revenue, very high free cash flow, and pricing power. The risks are regulators forcing it open and big insurers building their own data.

Thesis

Verisk Analytics is a single-segment data, analytics, and technology provider serving the U.S. (and increasingly global) P&C insurance industry. Its core asset is the ISO database, started in 1971 — over 38.9 billion statistical records, 3.6 billion new transaction records contributed annually by client-insurers themselves, and policy language for 32 lines of insurance that has been litigation-tested for 50+ years. All top 100 U.S. P&C insurers are clients. Over 80% of revenue is subscription, typically prepaid annually or quarterly. This is a contributory-data toll booth: insurers feed Verisk their loss data because they need the industry-aggregated rating and loss-cost output back, which creates a cold-start moat that is essentially unbeatable de novo.

The scorer reflects this: composite 81/100, profitability 20, balance sheet 19, capital allocation 20, valuation 22. ROIC 10y avg = 15.6% (high for an asset-light data business — the leverage masks how high the underlying return on tangible capital actually is, since Verisk is technically in stockholders' deficit from buybacks). FCF conversion 5y = 111% (owner earnings exceed reported net income — the gold standard of a Buffett business). Net debt/EBITDA = -0.175 (effectively unlevered if you ignore the ASR-funded buyback debt). Share count down 1.99% over 10 years, accelerating: a $1.5B February 2026 ASR plus $126M open-market in Q1 alone, with $2.5B authorization in place.

Valuation: at $181.11 vs. IV base $246.77 (price/IV = 0.73), reverse-DCF implied growth is 6.2% — below the historical organic algorithm. IV low $165.75 suggests near-floor support; IV high $312.67 implies 73% upside. P/E TTM 26.3 vs. 10y avg 40.3 — meaningful re-rating already happened. Owning shares below ~$170 carries a margin of safety; the math compounds at ~12% from base IV plus FCF yield over a decade if the moat holds.

Moat

Verisk has the cleanest example of a contributory-data network effect in U.S. equities. Five-moat analysis:

Switching costs (HIGH). Verisk's loss-cost filings, policy language, and rating rules are filed with regulators in all 50 states; insurers building products on those filings would need to refile, retest, and re-train field underwriters to switch. The 10-K notes Verisk processes ~2,000 regulatory filings/year and maintains 6 basic homeowners coverages plus 383 national and 714 state-specific endorsements. An insurer that swaps Verisk for an alternative inherits regulatory friction across every state, every line. Buffett's 2010 letter [2] notes that 'a sound insurance operation requires four disciplines' — Verisk supplies inputs to all four (exposure understanding, loss probability, premium adequacy, walk-away discipline). Carriers don't replace the spine of their actuarial stack lightly.

Network effects / contributory data (WIDE). The flywheel: insurers contribute 3.6B records/year because they receive aggregated industry loss costs they cannot replicate from their own book. The bigger Verisk's pool, the better the predictions, the more insurers contribute. A new entrant attempting to bootstrap a competing dataset starts with zero contributors — a chicken-and-egg problem that Verisk solved over 50 years as a not-for-profit ISO bureau before privatizing. Munger's $10B + 5-year competitor stress test: assume Google or Palantir spends $10B on insurance analytics. They still cannot get the top 100 P&C carriers to dual-feed contributory data, because the carriers' value comes from the existing 38.9B-record corpus and the regulatory pre-clearance Verisk provides. Money does not buy 50 years of court-tested standardized policy language.

Intangibles (HIGH). The ISO 'statistical agent' designation in all 50 states, Puerto Rico, and DC is a regulatory intangible. Insurers are required to report premiums and losses; Verisk is the channel. This is closer to a regulated utility than a software vendor. The court-tested coverage language is itself an intangible — meanings of policy terms have been adjudicated in the courts using ISO language as the reference standard, so deviation creates litigation risk for any insurer using non-Verisk forms.

Cost advantages (NARROW). Verisk's marginal cost to serve another carrier is near zero — the database, models, and policy language are fixed-cost. The 500-person field staff visiting 300,000 commercial properties per year creates a proprietary dataset on community fire-protection capabilities that no individual carrier could economically build. But this is partly available through alternate aggregators (CoreLogic, LexisNexis Risk Solutions, Moody's RMS for catastrophe modeling), so cost advantage exists but is not insurmountable in adjacent niches.

Pricing power (MODERATE). Subscription revenue >80%, typically multi-year prepaid. Verisk has demonstrated mid-single-digit organic revenue growth historically through a mix of price escalators and module attach. The 26.3x P/E (down from 10y avg 40.3) reflects that pricing power is strong but not 'mission-critical-to-double-prices' levels — insurers do push back, and a few lines (catastrophe modeling) face credible competition from Moody's RMS and CoreLogic.

Erosion risks: (a) Regulatory action — if state insurance commissioners decided ISO's quasi-monopoly on filings was anti-competitive, the toll booth narrows. (b) AI/foundation models — large language models trained on regulatory filings could in theory replicate the policy-language layer, though not the contributory loss-cost dataset. (c) Client consolidation — if the top 10 insurers grow to 60% of premium, they have bargaining leverage to in-source. The 2025 10-K confirms 'all of the top 100' P&C insurers are clients, but doesn't disclose customer concentration above the top 10, which deserves diligence.

Buffett's 1990 letter [5] notes that financial strength 'is unmatched' as a moat in insurance — Verisk inverts this: it sells the analytical strength that lets carriers price risk correctly. The four insurance disciplines Buffett enumerates in 2001 [6] (factor evaluation, aggregation limits, moral risk avoidance, walk-away discipline) all run through Verisk's data layer.

Moat verdict: WIDE

Management

Lee Shavel became CEO in May 2022 after serving as CFO. Elizabeth Mann is CFO (since 2023). The team has executed a clear simplification: the February 2023 sale of the Energy business (formerly Wood Mackenzie), the 2022 Verisk Financial divestiture, and the 2024 Verisk 3E sale focused the company on insurance — which the 2025 10-K calls a 'single segment.' This is the right strategic posture for a moat-rich asset; you don't dilute a contributory-data flywheel by chasing tangentially related verticals.

Capital allocation choices over the last decade:

Reinvest: Modest organic capex relative to cash flow. The 'Reimagine' modernization of core forms/rules/loss costs is the major internal investment, plus continued AI/ML augmentation. ROIC 10y avg of 15.6% suggests reinvestment returns are healthy but not outsized, consistent with an asset-light data business where the marginal return on growth capex is high but the absolute reinvestment rate is small.

Acquire: Mixed history. Verisk paid premium prices for several specialty acquisitions in the 2010s (notably Wood Mackenzie at ~$2.8B in 2015) that were later divested at lower implied multiples. The pending AccuLynx acquisition (referenced in Q1 2026 10-Q with mandatory-redemption senior notes that ultimately got terminated) shows discipline — they walked away rather than overpay or stretch covenants. Recent tuck-ins in Europe (Krug, etc.) appear bolt-on scale.

Debt: Net debt/EBITDA = -0.175 per the scorecard (effectively negative). Verisk uses leverage tactically — entering ASR agreements via the revolving credit facility, then repaying from operating cash flow. The Q1 2026 10-Q describes draws on the Syndicated Revolving Credit Facility for ASR funding 'subsequently repaid in full prior to' quarter-end. Conservative.

Buybacks (the headline): Share count down 1.99% over 10 years per the scorecard — but accelerating dramatically. February 2026: $1.5B ASR with HSBC and Wells, initial 6.986M shares delivered at $182.50. Q1 2026 open-market: an additional $126.1M repurchased, but at an average price of $216.24 — a higher price than the ASR. Board increased authorization to $2.5B in February 2026. The critical Buffett question — 'P/IV when buying' — is mixed: $182.50 is below base IV $246.77 (~74% of base, attractive); $216.24 is at ~88% of base IV (acceptable but not aggressive). Not the discipline Henry Singleton showed at Teledyne, but well above average for an S&P 500 staple. The pace of repurchase matches a management team that thinks the moat is intact and the stock is reasonable rather than screamingly cheap.

Dividends: $0.50/share quarterly, ~$65M/quarter, modest payout ratio. Sensible for a compounder — return some cash, retain the rest for buybacks where IRR is highest at fair prices.

Communication quality: Investor presentations are clear about the 80% subscription mix and organic growth algorithm. The post-divestiture 'one segment, Insurance' disclosure is a sign of management focus and unwillingness to obfuscate. The 10-K front-page list of competitive differentiators (proprietary data, expertise, relationships, scale) is honest and not hyperbolic. No earnings-management red flags evident in the filings provided; FCF conversion of 111% over 5 years is consistent with conservative accounting.

Concerns: (1) Stockholders' deficit on the balance sheet from aggressive buybacks at sometimes-elevated prices is fine economically but obscures the absolute return on equity invested. (2) The historical Wood Mackenzie / Verisk Financial cycle (acquire-divest-write-down) is a yellow flag, though under prior CEO Scott Stephenson, not Shavel. (3) Buying back stock at $216 in Q1 2026 (above current $181) shows some price-insensitivity, but is forgivable given the long-term IRR math.

Capital allocator: B

Industry

Porter's Five Forces on the U.S. P&C insurance data and analytics industry as Verisk participates in it:

Threat of new entrants — LOW. Building a contributory-data set spanning 38.9 billion statistical records and the regulatory plumbing to act as a 'statistical agent' in all 50 states takes decades and the cooperation of competitors. Verisk's origin as the not-for-profit ISO bureau (1971) gave it a one-time, unrepeatable institutional starting position. A well-funded entrant could attack a single niche (e.g., catastrophe modeling, where Moody's RMS competes credibly) but cannot replicate the across-the-stack position. Regulatory barriers (state filings, 50-state statistical-agent designations) plus the cold-start contributory-data problem make this near-impossible to recreate.

Bargaining power of suppliers — LOW. The 'suppliers' are the insurers themselves, who contribute the data. They are also the customers. This dual role neutralizes supplier power: an insurer that wants Verisk's industry-aggregated loss costs has to contribute its own data, and the value of contributing is greater than the cost of contributing. Verisk's 500-person field staff and 250 insurance experts/lawyers are labor inputs but specialized; talent is competitive but not concentrated.

Bargaining power of buyers — MODERATE. The top 10 P&C insurers (Progressive, State Farm, Allstate, Berkshire/GEICO, Travelers, Liberty Mutual, etc.) collectively control a large share of premium and have leverage. Some — Progressive most notably — have invested heavily in proprietary data and analytics and could self-supply more functions. But even Progressive uses ISO loss costs as a baseline. The mid-tier and small carriers are essentially captive: they have neither the scale nor the data to build internal alternatives. Net: large carriers can negotiate terms; the long tail cannot. The 80%+ subscription, multi-year-prepay revenue mix limits buyer power in the short run.

Threat of substitutes — MODERATE. The substitutes are: (a) carrier in-sourcing (only feasible for the top 5–10), (b) competing data vendors (LexisNexis Risk, CoreLogic, TransUnion's TLOxp, Moody's RMS for cat modeling), and (c) emerging AI-first startups. None of these threats is symmetric across Verisk's product line — cat modeling has real competition; ISO forms and rules effectively don't. AI-driven foundation models could in theory commoditize parts of the policy-language layer over a decade, but the contributory-data and regulatory-filing layers are substitution-resistant.

Industry rivalry — LOW within Verisk's core; MODERATE in adjacencies. Within the ISO-style core (forms, rules, loss costs, statistical reporting), Verisk has effectively no peer. In claims solutions, anti-fraud, and specialty analytics, rivalry is more active (Guidewire on policy administration, Duck Creek, multiple cat modelers). Verisk's strategy of 'extending into adjacent insurance workflows' invites more competition than the legacy core, which is a long-term margin watchpoint.

Value pool location and trajectory: The U.S. P&C insurance industry collects ~$900B in premiums annually and spends a small but growing fraction on data, analytics, and technology. Verisk captures roughly 0.3% of total industry premium as revenue — there is room for that share to expand as more decisions become data-driven (telematics, climate, underwriting AI). Climate change increasing catastrophe frequency is a tailwind for cat-modeling and exposure analytics — Buffett's 2025 letter [3] notes 'additional capital entered the market' in reinsurance, which means more carriers needing better risk data. The trajectory is up and to the right, with a long runway, modulo regulatory risk.

Industry Verdict: Excellent

Inversion

I am playing a short-seller. Here is the strongest credible bear case on Verisk at $181.11.

The single event that kills this. State insurance regulators or the FTC under a more activist administration declare ISO's quasi-monopoly on rating filings anti-competitive and force structural separation — either compelling Verisk to license its loss-cost database on regulated, cost-plus terms, or forcing carriers to use independent data sources for rate filings. This is not science fiction: ISO operated as a not-for-profit advisory bureau until 2009 precisely because its function was viewed as quasi-utility. A populist 'insurance is too expensive' wave (already visible in Florida and California homeowners) could provide political cover. The Atlas Data Privacy litigation referenced in the 10-K (vrsk:AtlasDataPrivacyCorpEtAlVVeriskAnalyticsIncMember) is a hint that data-aggregation businesses face escalating legal exposure. If Verisk is regulated as a utility, the IV collapses to maybe 12x earnings — a $90 stock — overnight.

Why the moat is narrower than bulls think. The contributory-data flywheel is real for the long-tail of small/mid carriers, but the top 10 carriers (who write a majority of premium) are increasingly self-sufficient. Progressive has spent two decades building proprietary telematics and pricing analytics; State Farm and Allstate have similar internal capabilities. If the top 10 carriers (40-50% of industry premium) consolidate their data needs internally and reduce Verisk spend by 30-50%, that's mid-single-digit revenue growth turning negative even if every small carrier stays loyal. Foundation models (GPT-class systems) trained on public regulatory filings can already draft policy endorsement language at a fraction of the cost; the litigation-tested-language moat erodes one LLM-generated form at a time over the next decade. CoreLogic, LexisNexis, and Moody's RMS are real competitors in property data, identity/risk data, and catastrophe modeling respectively. Verisk's 'one segment Insurance' framing makes the moat sound monolithic, but it is actually a stack of distinct moats of varying width, and the narrow ones (cat modeling, claims fraud) are growing faster than the wide ones (ISO core).

Why management is worse than it appears. Lee Shavel is paying $216.24/share in Q1 2026 open-market repurchases — above the current price of $181 and above the IV-low of $165.75. That is not Henry Singleton discipline. The Wood Mackenzie / Verisk Financial / Verisk 3E acquisition-and-divestiture cycle under the prior CEO destroyed value (acquired with shareholder cash, divested at write-downs). The current AccuLynx situation — issuing senior notes for an acquisition that then fell through — burned issuance fees and management attention. Stockholders' deficit on the balance sheet is the cumulative result of buybacks at sometimes-rich prices. Compensation is heavily equity-linked, which is fine when the stock works but creates incentives to hit short-term EPS targets via debt-funded buybacks rather than reinvest in moat-widening capabilities.

What bulls are extrapolating that won't hold. (1) That 6%+ organic revenue growth continues forever. The reverse-DCF implied growth of 6.2% is not nothing — at scale ($3B+ revenue, ~$900B serviceable industry premium), Verisk's revenue/premium ratio of ~0.3% would need to nearly double for organic growth to sustain that pace through pricing alone, and unit growth requires expanding into adjacencies (life, international, specialty) where moats are narrower. (2) That the 80%+ subscription mix is locked in. Subscriptions can be renegotiated; multi-year deals come up for renewal. If 2026-2027 sees a P&C industry margin compression from rate cycle peaks (Buffett notes in 2025 [3] that primary insurance pricing is plateauing and reinsurance facing 'significant price declines'), carriers will press Verisk on price. (3) That ROIC stays at 15.6%. As Verisk pushes into life insurance, international, and adjacencies, returns on incremental capital will compress because those markets are more competitive than ISO.

Valuation trap (multiple compression / regime change). The 10y average P/E of 40.3 reflects an era of zero rates and 'quality compounder' premium multiples. We are in a different regime. Even today's 26.3x is rich for a business whose terminal organic growth is plausibly 3-4%, not 6%+. EV/FCF of 25.4x bakes in continued FCF growth that re-investment-rate math (15.6% ROIC × low reinvestment rate = mid-single-digit FCF growth) does not obviously support. If the multiple compresses to 18-20x earnings — consistent with utility-like data businesses (Moody's, S&P Global trade richer; specialty actuarial/data services trade lower) — and earnings grow at 4% rather than 6%, fair value is $130-140, not $181. The base IV of $246.77 assumes a specific terminal-multiple choice; sensitivity to that choice is high.

If I am right, the stock could be worth $90-130 within 3 years.

Lollapalooza Bias Check

Biases active in me as the analyst right now:

Authority bias: The Buffett canon excerpts about insurance discipline [2][3][6] are not directly about Verisk — they are about Berkshire's underwriters. I am letting their authority color my view of Verisk's quality because Verisk sits in the same ecosystem. Verisk is not Berkshire-managed and does not bear the underwriting risk; it sells data to underwriters. I am treating Verisk's adjacency to Buffett-quality insurers as if it were Buffett-quality itself, which conflates two different positions in the value chain.

Anchoring: I am anchored to the scorer's IV base of $246.77. That number is the output of a model with assumed terminal growth and discount rates. If I had built the IV myself with a 7% discount rate and 3% terminal growth (more pessimistic than the implied 6.2%), the IV would be 30%+ lower. The price/IV ratio of 0.73 looks attractive only if the IV is right. I should weight IV-low ($165.75) more heavily than I am instinctively doing — the price is barely above IV-low, not 'screaming buy' territory.

Confirmation bias: Once I framed Verisk as 'a Buffett moat business,' I started looking for evidence to confirm the moat (proprietary data, network effects, switching costs) and underweighting evidence against (top-10 carrier insourcing, AI substitution, regulated-utility tail risk). The 10-K is written by Verisk, so its 'competitive differentiators' section is by definition pro-Verisk. I am reading marketing copy as fact.

Recency bias: The Q1 2026 $1.5B ASR is fresh and feels bullish — management bought a lot at $182.50, near today's price. But one quarter of buyback activity is not capital-allocation discipline; the 10-year share-count change of just -1.99% suggests historical buybacks have been roughly net-zero against dilution from equity comp. I should weight the 10-year history more than the most recent quarter.

Social proof / authority: Verisk is in the S&P 500 and has a 'quality compounder' reputation among institutional investors. That popularity is itself a signal — but in the wrong direction. Crowded compounder names have historically de-rated when the regime changed (witness MSCI, S&P Global, Moody's all going through painful 30-40% drawdowns when rates rose). I should not assume the consensus has correctly priced the regime risk.

Deprival super-reaction (counter-bias): I have to check that I'm not being too clever-bear. The keystone-species and contributory-data arguments are real. The risk of 'losing' a moat call by being too cautious is just as real as the risk of overpaying.

Net effect: my conviction should be medium, not high. The moat is real; the price is okay-not-great; the regime risk is non-trivial.

10-Year Outlook

Ten-year outlook test:

Same fundamental business model? Yes. Verisk in 2036 will still be aggregating P&C insurance loss data, publishing rating filings, maintaining policy language, and serving claims and catastrophe analytics. The form factor will shift (more API/cloud delivery, more AI-augmented workflows), but the underlying contributory-data and regulatory-statistical-agent functions are stable. The 1971-2026 history is 55 years of essentially the same business; another decade of the same is the base case.

Customer base larger? Probably modestly. Domestic top-100 P&C insurers are already 100% penetrated by Verisk; growth has to come from (a) life and annuities expansion, (b) international markets (UK, Europe), (c) adjacent specialty lines (cyber, parametric, climate-resilience). All are real but lower-moat than the U.S. P&C core. 5-15% customer base expansion is plausible; 30%+ is not.

Profit per customer higher? Likely yes, modestly. The 'Reimagine' modernization and AI-augmented analytics suggest higher ARPU per insurer. Pricing power on subscription renewals adds 2-4%/year. Cross-sell of claims, fraud, and catastrophe modules into existing rating-data accounts is the bull case. 20-30% profit-per-customer expansion over a decade is reasonable.

Moat wider? Probably stable, not meaningfully wider. The U.S. P&C core moat is already maximal — you can't get more than 100% of the top 100 carriers. International expansion adds a second moat from scratch in each new geography, which takes a decade per region. AI is a wash: it widens the moat where Verisk has more data, narrows it where competitors can train on public data.

Single biggest threat? Regulatory action in the U.S. — either antitrust intervention, or state-by-state reform of how ISO functions in rate-filing. This is low-probability but high-impact. Second-biggest: top-5 carrier in-sourcing reducing the pool of contributable data and forcing Verisk to compete with its own customers' analytics teams.

Confidence: I can describe the 2036 Verisk in plain English, name its top three profit drivers, and predict its moat shape with reasonable accuracy. The downside scenarios are bounded (utility-multiple regime is the worst credible case, not zero). This is solidly within circle of competence.

CONFIDENCE: medium

Position Guidance

• Recommendation: Hold (with bias to Buy on weakness)
• Conviction: Medium
• Target buy price: $170 (~3% below IV-low of $165.75 plus a small margin; below this the moat-vs-price math is genuinely attractive)
• Target trim price: $310 (~99% of bull-case IV $312.67; above this even optimistic assumptions are baked in)
• Position sizing: 2-4% of portfolio for a high-quality compounder bucket; size up to 5-6% only if price falls toward $150 (clear margin of safety vs. IV-low). Below current $181 but above $170, accumulate slowly. Above $230 (~93% of base IV), do not add. Above $290, trim.
• Holding period: 5-10 years minimum; this is a compounder thesis, not a special situation.
• Key monitor items: (1) regulatory rumblings in P&C insurance, (2) buyback price discipline (avg P/IV when buying), (3) organic revenue growth holding 5%+, (4) top-10 customer concentration trends, (5) M&A discipline post-AccuLynx.