Verisk Buys McKenzie Intelligence to Close the Gap Between a Catastrophe and Its First Loss Read

Rob Newbold, president of Verisk's Catastrophe and Risk Solutions unit, announced on July 29, 2026 that Verisk had acquired McKenzie Intelligence Services, a UK geospatial firm whose GEO platform delivers verified post-event damage insights within 48 to 72 hours (McKenzie Intelligence Services). The deal landed the same day Verisk reported $806 million in second-quarter revenue, up 4.3% year over year (Verisk, July 29, 2026).

What McKenzie Actually Adds to the Model

McKenzie Intelligence Services was not a catastrophe modeling company before Verisk bought it; it was a data and analysis layer that sat downstream of one. The company's GEO platform pulls satellite imagery, radar sensor data, open-source reporting, and machine-learning-assisted analysis, cross-checked by former military intelligence analysts, into a single pipeline that produces three successive reads on an unfolding event: an initial exposure layer identifying at-risk locations within 24 hours, a comprehensive portfolio-level damage overview within 24 to 48 hours, and verified, financially confirmed damage insights within 48 to 72 hours (McKenzie Intelligence Services). The platform tracks 13 hazard types, from hurricanes and hail to wildfire, flood, earthquake, and civil unrest, and part of its raw imagery has come from ICEYE, the synthetic-aperture-radar satellite operator whose flood and wildfire products already ran on a data-supply agreement into McKenzie's GEO platform before the acquisition (ICEYE, McKenzie Intelligence Services). That supply relationship matters for what Verisk did and did not buy: it acquired the analysis, triage, and client-delivery layer, not the underlying satellite constellation feeding it.

Newbold's framing of the deal was explicit about where it plugs in. "By bringing MIS's real-time geospatial intelligence together with Verisk's catastrophe models, risk analytics, and claims solutions, we can give clients a more complete view of unfolding events so they can assess impacts, prioritize response and support policyholders more effectively," he said (Verisk, July 29, 2026). Forbes McKenzie, the firm's founder, described the deal in similar terms: combining MIS's intelligence with "Verisk's risk modelling and claims solutions" to help insurers serve policyholders during their most difficult moments (Verisk, July 29, 2026). Both quotes describe MIS feeding two things at once: the pre-event catastrophe model that generates a loss distribution, and the post-event claims and analytics workflow that turns an actual storm or earthquake into a dollar figure. MIS now sits inside Verisk's Catastrophe and Risk Solutions division, alongside the newly launched Synergy Studio cloud platform and Verisk's Property Claim Services (PCS) industry loss index.

From a Six-Hour Wind Field to a Ninety-Day Revision Cycle

The acquisition is legible mainly against the cadence it is meant to replace. PCS, Verisk's decades-old industry loss index, has historically issued its initial estimate and then revised it periodically, continuing to update the figure roughly every 90 days until the team settles on a final number (Verisk PCS). That cadence produced, for example, Verisk's Extreme Event Solutions estimate that Hurricane Ian's onshore property losses would fall between $42 billion and $57 billion, a $15 billion range issued in the days after landfall in 2022 that reflected genuine uncertainty about wind, storm surge, and inland flood damage still being surveyed on the ground. Moody's RMS runs a faster but still multi-stage process on the modeling side: HWind, its tropical cyclone tool, issues live wind-field footprints and loss estimates every six hours while a storm is active, then produces a detailed event reconstruction, incorporating the hazard as best understood after the fact, within one to two weeks of the event's conclusion (Moody's RMS).

McKenzie's GEO platform is built to beat both benchmarks on the ground-truth side of that equation. Rather than modeling a probable loss distribution before the event or reconstructing a hazard footprint after it, GEO observes actual damage: it delivers a verified, financially confirmed damage picture in 48 to 72 hours, well inside PCS's 90-day full-cycle window and comparable to or faster than RMS's one-to-two-week post-event reconstruction. ICEYE's own wildfire product, built partly on the same imagery pipeline that used to feed McKenzie as an independent vendor, delivers building-level damage detection every 24 hours during an active wildfire (ICEYE). The table below lines up the cadences these tools have historically operated on.

Estimate Source	Update Cadence	What It Captures
Verisk PCS industry loss index	Revised roughly every 90 days until a final figure	Aggregate industry-wide insured loss
Moody's RMS HWind	Every 6 hours while a storm is active; detailed reconstruction 1–2 weeks post-event	Hazard footprint and modeled loss
ICEYE wildfire building damage detection	Every 24 hours during an active event	Building-level physical damage, satellite-observed
McKenzie GEO (now Verisk)	24 hours (exposure layer); 24–48 hours (portfolio overview); 48–72 hours (verified financial impact)	Property- and portfolio-level damage, financially confirmed

The pattern across all four is the same direction of travel: the industry has been compressing the time between an event and a defensible loss number for years, and Verisk's acquisition is a bet that owning the fastest ground-truth layer, rather than licensing it from an independent vendor, is worth more now that carriers expect an answer inside a claim cycle rather than a reserving quarter.

The Reserving Clock Verisk Is Trying to Move

For a reserving actuary, the practical effect of a faster ground-truth read is not that the ultimate loss changes; it is that the first credible IBNR estimate moves earlier in the claim-development curve. A carrier's initial cat loss pick after a hurricane or earthquake traditionally leans on a blend of the vendor model's pre-event probable maximum loss, early claims counts from the carrier's own book, and whatever PCS or RMS figure is available at the time the quarter closes. When that external anchor takes 90 days to settle, or a market estimate revises materially between two quarter-ends, the actuary's Bornhuetter-Ferguson expected-loss-ratio input is itself moving underneath the reserve, and a change in the vendor's published range between the point estimate is set and the point the financials close can force a mid-quarter reserve revision that has nothing to do with new claims data.

A property-level damage feed that reaches verified status in 72 hours changes that math by giving the actuary a ground-truth anchor before the quarter, not during it. It does not shorten the tail of a genuinely long-developing casualty or liability claim, but for the short-tail property lines where cat losses concentrate, it means the expected-loss-ratio input to a Bornhuetter-Ferguson estimate can be set closer to actual observed damage and further from a modeled probability distribution. That is a meaningfully different reserving posture: less "what does the model say the loss should be" and more "what does the imagery say the loss already is." Carriers that have leaned on real-time cat accumulation monitoring to replace the weekly PML run are already moving in this direction on the exposure-management side; McKenzie's acquisition extends the same logic to the loss-estimation side of the same event.

One Vendor, Two Roles: The Model and Its Own Check

The acquisition also creates a structural question that did not exist while McKenzie operated independently: Verisk now owns both the ex-ante catastrophe model that estimates a probable loss before an event and the ex-post geospatial feed that many carriers have used to validate or challenge that estimate once the event has actually happened. When those two functions sat in different companies, a carrier that thought Verisk's pre-event loss cost looked too high or too low had an independent data point to test it against. With MIS inside the same corporate structure that builds the AIR and Touchstone model suites, that independence is gone by construction, even if nothing about the underlying analysis changes on day one.

This is not a new pattern in catastrophe risk data; it is the same concentration dynamic actuaries have flagged in AI model-provider concentration, where a handful of foundation-model vendors sit underneath a large share of insurers' automated underwriting and claims tools. Regulators have already articulated the underlying concern in the catastrophe-model context specifically: no single vendor's catastrophe model is considered definitive, which is why a multi-model approach that blends independent views is the standard defense against relying too heavily on any one vendor's assumptions (NAIC Catastrophe Modeling Primer, March 2025). That defense works when the validating data source is genuinely independent of the model it is checking. A carrier running Verisk's AIR models and now also buying Verisk's post-event damage confirmation from the same corporate parent has, in practice, narrowed the independence of its own check without necessarily realizing the boundary moved.

The mitigating detail is that McKenzie's own imagery supply, including the ICEYE satellite feed, remains outside Verisk's ownership, so the acquisition consolidates the analysis and delivery layer rather than the entire data-collection chain. A reserving or pricing actuary relying on a Verisk-branded post-event estimate can still ask what raw imagery and sensor sources fed it, and a carrier that wants a genuinely independent second opinion can still turn to ICEYE directly, to Moody's RMS's HWind and HD model suite, or to a facultative reinsurer's own claims triage. But the convenience of a single-vendor, single-invoice loss read is precisely what makes the independent check easy to skip in practice, particularly for mid-market carriers without a dedicated cat-modeling team to interrogate the source data behind a headline number.

The Deal Inside the Quarter

Financial terms of the McKenzie acquisition were not disclosed, and Verisk said the transaction is not expected to have a material impact on its financial results (Verisk, July 29, 2026), consistent with a bolt-on data acquisition rather than a scale deal. The number that does matter is the backdrop it landed against. Verisk's Q2 2026 organic constant-currency revenue grew 5.8%, subscription revenue rose 8% and now makes up 83% of total revenue, and adjusted EBITDA reached $464 million at a 57.5% margin (Verisk, Q2 2026 results). The company reaffirmed full-year guidance of $3.19 billion to $3.24 billion in revenue and continued an aggressive capital return program, retiring $1.7 billion of stock year to date through a mix of accelerated and open-market repurchases, with $800 million remaining on its authorization (Verisk, Q2 2026 results). On the same call, management said XactAI licensees had grown nearly tenfold to roughly 7,000 users, part of the broader AI product push that also includes new Claude connectors built for insurance analytics workflows, a build actuary.info covered when Verisk shipped MCP connectors for Anthropic's Claude (Verisk, Q2 2026 results).

A bolt-on acquisition that does not move the needle on Verisk's own P&L is nonetheless a strategic tell about where the company sees its moat. Verisk has spent the past several quarters expanding the AI and data layer around its core model suite, shipping seven new AI modules in Q1 2026 and building a growing carrier adoption pipeline around them. McKenzie fits that pattern precisely: it is not a new peril model, it is a new speed advantage layered onto the existing catastrophe franchise, financed out of a balance sheet that just returned $1.7 billion to shareholders in the same six months.

Where This Leaves Moody's RMS, ICEYE, and the Independent Shops

The competitive read on the deal splits along two lines: model vendors and independent event-response specialists. Against Moody's RMS, the acquisition narrows a specific gap. RMS's HWind and event-reconstruction tooling is strong on the hazard-footprint side, tracking wind speed and storm surge in near real time, but its detailed post-event reconstructions still run on a one-to-two-week cycle. Verisk buying a firm whose core competency is compressing that reconstruction into 72 hours is a direct answer to that gap, and it comes on the heels of Moody's own HD model launch for severe convective storm, suggesting both vendors are racing to close different parts of the same speed-and-granularity problem from opposite ends.

Against the independent event-response shops, the calculus is starker. ICEYE, which supplied part of McKenzie's imagery under a standing agreement, remains an independent satellite operator that can sell directly to carriers, and its wildfire and flood products stand on their own. But smaller, single-purpose geospatial and claims-triage vendors that competed with McKenzie on service now face a well-capitalized incumbent with a distribution channel into essentially every large US and European carrier through Verisk's existing model licenses. A carrier that already pays Verisk for AIR models or Synergy Studio access gains a low-friction upsell path into McKenzie's event-response layer that an independent competitor cannot easily match on bundling economics alone, even if it can match GEO on raw analytical quality.

A Faster Read Is Also a Faster Repricing Trigger

The pricing implication runs in parallel to the reserving one. Reinsurance retrocession markets and parametric triggers already price off industry loss indices, and the lag between an event and a defensible loss figure has historically been part of what kept post-event repricing conversations from happening inside the same week as the loss. A 48-to-72-hour verified damage read compresses that lag meaningfully relative to a 90-day PCS cycle, which means a cedant, a retrocessionaire, or a parametric trigger administrator has a usable number to negotiate against well before the quarter in which the event occurred has even closed its books. For property catastrophe treaty pricing specifically, that shortens the gap between "an event happened" and "capacity providers can reprice the next layer they are asked to write," a dynamic that matters most in a soft market where reinsurers are already competing on speed of quote as much as on rate. It also raises the stakes on getting that fast number right: a verified 72-hour estimate that a market treats as authoritative for repricing purposes carries more consequence if it is later revised than a PCS figure everyone already understood to be provisional for 90 days.

What Actuaries Inherit With a Vendor Loss Pick

The governance question this acquisition raises is not whether McKenzie's imagery-based damage estimates are accurate; independent event-response geospatial analysis has a reasonably strong track record against ground survey data. The question is what a reserving or pricing actuary is actually relying on when a Verisk-branded loss figure lands in a reserve committee packet or a treaty renewal negotiation, and whether that reliance is documented the way any other material assumption would be. An actuary who takes a vendor's cat model output has, at minimum, some visibility into model methodology through vendor documentation, model change logs, and increasingly through platforms like Synergy Studio that expose model versioning directly. An actuary who takes a vendor's post-event damage estimate as an input to an IBNR pick or a treaty repricing decision, without having seen the underlying satellite imagery, the analyst review process, or the confidence intervals behind that 48-to-72-hour figure, is accepting a black-box number on a compressed timeline precisely because the timeline pressure discourages the kind of diligence a slower estimate would invite.

That is a governance gap worth naming explicitly rather than assuming away. As McKenzie's output moves from an independent third-party check into a Verisk in-house product line, the burden shifts toward carriers and their actuaries to ask the same questions of a 72-hour damage read that they would ask of any other vendor model: what is the confidence interval, how has the estimate performed against final settled losses historically, and does the carrier have any source of validation that sits outside the vendor relationship that produced both the pre-event model and the post-event check. Those questions do not require distrusting Verisk's analysis. They require recognizing that the industry just lost one of its remaining independent post-event data points, and priced that loss of independence as a deal too small to move a $3.2 billion revenue company's guidance.