Skip to content

Exploration 01 — Market scan (business · technical · architecture)

Status: living working doc — wave 1 done 2026-07-25 (Claude, web research); wave 2 targets listed at the end. Runs in parallel with the manual UI-pattern + business-idea scan (Giorgos; UI-focused predecessor: risk-assessment-tool/docs/plan/ui-revamp.md, 30 products, patterns extracted). Axes: business model / pricing · persona served · delivery form · data moat · architecture pattern · what nobody does (white spots).

Wave 1 findings

1. Business-model archetypes that actually exist

Archetype Exemplar Mechanics Relevance to us
Free consumer funnel → enterprise data/API licensing First Street (riskfactor.com) Free per-property portal for every US address; revenue (~$9M/yr) from enterprise API + data packages incl. self-service on AWS Data Marketplace; ended in acquisition by MSCI Direct template for the citizen persona: the free funnel is the marketing, monetization is B2B. Also names the realistic exit shape: acquisition by a financial-data player
Enterprise SaaS, six-figure contracts Climate X Spectra, Jupiter/XDI, Munich Re, Moody's RMS SaaS + enterprise API, custom pricing, full-portfolio deployments at six-figure annual commitments The ceiling we do NOT compete with. Confirms the mid-market gap below
Open-core SaaS on open science CLIMADA Technologies (ETH spinoff, delta-climate™) Commercial SaaS on top of open-source CLIMADA; "open-core… full transparency"; venture award 2024; partnerships (Wüest Partner, real estate) Closest strategic twin and direct competitor. Validates the model we're contemplating; differentiation must come from elsewhere (see white spots)
Industry-owned open framework + commercial model marketplace Oasis LMF Not-for-profit owned by ~30 (re)insurers; 18+ model suppliers, 90+ models; commercialized via Nasdaq's cloud NRMC service; Moody's RMS now integrating Oasis-based models Insurance already trusts open cores with commercial layers. Parking-lot idea: an Oasis-compatible output/packaging path as an insurer-credibility channel
Publicly funded free tooling CLIMAAX (Horizon Europe, ends Dec 2026), OS-Climate/FINOS physrisk Free toolboxes/libraries; CLIMAAX served 67 EU regions with CRA workflows; physrisk actively maintained (pushes July 2026) with physrisk-api/-ui Competes for the decision-maker & researcher personas at price zero. CLIMAAX's funding cliff (Dec 2026) may leave regions with unmaintained workflows — a post-project vacuum worth watching

2. Regulatory drivers (the demand side — strongest wave-1 finding)

  • Greece, Law 5116/2024: from 1 Jan 2025, businesses with turnover > €500k are required to insure physical assets (buildings, equipment, inventory, vehicles) at ≥ 70 % of value against fire/flood/earthquake nat-cat. A regulatory-mandated customer base for property nat-cat risk assessment in the home market — insurers must price thousands of new policies, businesses must buy them.
  • EIOPA / Solvency II ORSA: climate materiality assessment + two climate scenarios (below-2 °C / well-above-2 °C) across three time horizons are expected in every insurer's ORSA since 2022; EIOPA's July 2025 monitoring shows most insurers now comply — but mid-size insurers do this with consultants and spreadsheets, not six-figure platforms. An "ORSA climate scenario module" (portfolio × two scenarios × three horizons → documented, reproducible report) is a concrete, regulation-shaped product wedge.
  • Context signal: Greece is investing in nat-cat tech nationally (e.g. first country with a dedicated wildfire-detection satellite array, 2026).

3. White spots — confirmed, weakened, or killed

Hypothesis Verdict Evidence
Mid-market / regional insurers priced out of enterprise platforms but under regulatory mandate Confirmed — the primary wedge Six-figure enterprise pricing above; EIOPA ORSA + GR 5116/2024 demand below; nothing in between except consultants
Self-hosted, white-label, deployable-per-client platform ~~Confirmed~~ → Narrowed (reassessment 2026-07-26) Refuted as an absolute: Verisk Touchstone ships on-prem, OasisLMF is BSD-3 self-hostable, CelsiusPro sells white-label insurer platforms. Survives only as a claim against the modern climate-physical-risk SaaS set (Climate X, Jupiter, CLIMADA Tech) — a positioning statement, not a vacancy
Multi-engine consensus (same scenario, N engines, agreement view) ~~Confirmed~~ → KILLED as headline (reassessment 2026-07-26) Model blending/comparison is decades-old insurance practice (IFoA working-party paper; Guy Carpenter advisory; Nasdaq NRMC offers 300+ models; Moody's integrates Oasis models) — and the market has moved toward single-model on cost. Our roster also lacks same-peril plurality (CLIMADA = our parity reference; OpenQuake = seismic-only). Keep engine plurality as the ratified product commitment and consensus as a feature; never claim "nobody offers this"
SE-Europe / Greek localized data, language, and hazard mix Confirmed as GTM advantage, corrected as "moat" No localized player found — but the underlying data is open (ESRM20, EFAS, JRC), so anyone can ingest it. A moat only compounds from non-public assets: claims-data partnerships, HAIC relationships, Greek-language workflow depth
"Open-source physical climate risk platform" as a differentiator per se KILLED OS-Climate physrisk (FINOS, active), CLIMAAX toolbox, CLIMADA itself. Openness alone is table stakes in this segment; differentiation must be product completeness (tenancy, UI, reports, ops) and locality

4. Architecture patterns worth adopting (cross-checked with the UI scan)

  • Precompute → serve: consumer-grade players (First Street, Munich Re on-demand) precompute per-location scores nationally and serve lookups instantly; the model run is a factory job, not a user action. Confirms the answer-layer thesis in 00-frame.md.
  • Dual-track delivery (portal + API with identical semantics) — Fathom, First Street, Climate X. Our unpublished SDK is this track, waiting.
  • Score rollups over raw values everywhere (grades/ratings with explainability drill-down) — Munich Re score tree, Cervest A–F, First Street 1–10.
  • Editions/tiers on one shell (Munich Re "editions") — licensing-gated feature bundles map cleanly onto our per-client instance model.

Implications carried forward (input to 02-personas)

  1. The wedge persona is the regional/mid-market insurer (regulatory pull, pricing gap, Sword channel alignment). Citizen funnel is the marketing surface, not the business. Decision-maker/researcher personas face free-tool competition (CLIMAAX, physrisk) and monetize later, if at all.
  2. The two defensible differentiators available to us: self-hosted white-label instances and multi-engine consensus — both already architectural commitments. Greek/SE-Europe data depth is the moat that compounds with time.
  3. An ORSA climate module and a 5116/2024 compliance pack are candidate first-revenue products that require no new science — packaging + reports over the existing pipeline.

Wave 2 findings (2026-07-26)

5. Munich Re on-demand — the pay-per-lookup archetype exists

Munich Re sells subscription-free single-location climate/hazard reports for credits (48 credits per lookup; € price per credit not publicly verified), with two report editions on one credit pool, worldwide coverage. Confirms: (a) there is retail demand for per-location answers below the enterprise tier; (b) the delivery form is a composed report, not a dashboard — reinforcing the report-composer primitive.

6. OS-Climate physrisk — capability assessment

Hazard indicators onboarded from public sources (CMIP/CORDEX-derived), generic vulnerability-model configuration, methodology formally documented. But by their own description the UI is a hosted "sandbox"/demo and physrisk-api/-ui are an "example API and user interface." Assessment: physrisk is a finance-sector stress-testing library with a reference stack, not a product — no tenancy, no report workflow, no white-label story. The product distance between physrisk and us is large and in our favor; their public hazard-indicator datasets (AWS Open Data registry) may be reusable as ingest sources.

7. CLIMAAX — reusable under Apache-2.0 (verified first-hand)

Checked the GitHub org directly: per-hazard workflow repos (FLOODS, STORMS, HEATWAVES, DROUGHTS, SNOW, HEAVY_RAINFALL, FIRE, MULTI) are Apache-2.0, actively pushed as of July 2026. Their CRA workflows (Jupyter, EU datasets) are legally reusable as methodology/ingest recipes for the decision-maker persona. Post-project (Dec 2026) maintenance remains unclear — the vacuum thesis stands, unconfirmed.

8. Greek market structure — two findings that CUT AGAINST the insurer-first lean

  • Hazard-mix misalignment (the material one). Greece's dominant peril is earthquake (~€22 bn residential loss at 1-in-200; risk-based pricing done today on CRESTA zones), and Law 5116/2024's mandated trio is fire / flood / earthquake. Our hazard set (RF, TC, WF, WS) covers flood and wildfire but not earthquake — and TC is near-irrelevant for Greece. Serving the Greek insurer wedge credibly requires a seismic answer: GEM/OpenQuake-derived hazard (note Phase 7.2 already plans a GEM exposure adapter — the seismic-hazard question is bigger and unscoped). This is now an explicit open point for session 2, not a footnote.
  • Compliance may not need granularity. 5116/2024 compliance can plausibly be met with zonal (CRESTA-style) pricing that insurers already do; the case that they need per-asset modeling rests on competition/loss-ratio pressure, not on the law itself. Must be validated in stakeholder interviews (aligns with the "real market need" open point already minuted in the parallel corporate track).
  • Confirmations for the thesis: Greece's protection gap is stark (1990–2019: ~9 % of fire losses, ~4 % windstorm, ~8 % flood/earthquake losses insured — Bank of Greece), and HAIC itself frames the challenge as building an insurance culture — literally the same phrase as our insurer-persona goal.

9. B2B2C embedded risk — white spot is regional, not global

In the US this is a mature category: Guidewire HazardHub (500+ per-address risk attributes inside underwriting flows), Verisk LOCATION, plus climate-risk APIs (Climate X, First Street) feeding quote flows. No equivalent found for Greece/SE Europe. Refined conclusion: the embed play is proven elsewhere and vacant locally — lower product risk, but the differentiator is locality, not novelty.

10. Reinsurer free/bundled tools — a real threat to the insurer wedge

Swiss Re CatNet (10k+ registered users; versions bundled for clients) and Munich Re NATHAN give cedents hazard-map lookup + climate modules, often via the reinsurance relationship. Greek insurers may already "have a tool." The honest counter: these are hazard atlases, not portfolio EP/pricing/ORSA workbenches, and never white-label or self-hosted — but whether that distinction matters to a mid-size Greek insurer is a question for interviews, not assumption.

Implications carried forward (input to 02-personas) — revised after wave 2

  1. The regional/mid-market insurer wedge survives wave 2 but with two validation gates: (a) the earthquake gap must have a credible roadmap answer before any insurer pitch; (b) stakeholder interviews must confirm demand for per-asset granularity beyond zonal compliance and beyond bundled reinsurer tools.
  2. Citizen funnel as marketing surface: unchanged, strengthened by Munich Re's pay-per-lookup precedent (there is even a paid retail market, not just free).
  3. Defensible differentiators unchanged: self-hosted white-label instances, multi-engine consensus, SE-Europe locality/language/data.
  4. ORSA module remains the cleanest first-revenue candidate (regulation-shaped, scenario-based — plays to the existing scenario engine, sidesteps the per-asset granularity question entirely).
  5. New candidate requirement for the roadmap: seismic hazard support (likely via the adapter path — GEM/OpenQuake ecosystem — rather than in-engine).

Reassessment corrections (2026-07-26, adversarial verification)

An independent refuter pass (3 verifier agents + completeness critic, web-grounded) corrected the scan's load-bearing conclusions:

  1. The wedge is not empty — EIOPA gives away a free tool inside it. EIOPA publishes the open-source CLIMADA-App explicitly for small/medium insurers and national authorities (nat-cat + climate risk, incl. earthquake; updated June 2026) — built on the same science our engine reimplements, and absent from waves 1–2 of this scan (EIOPA open-source tools). Productized mid-market ORSA tooling also exists: Ortec Finance ClimateMAPS (sold for "Solvency II, ORSA and TCFD"), Milliman ORSA tooling. Corrected wedge: no self-hosted, portfolio-level nat-cat workbench at mid-market price; pricing power must come from workflow/report/audit value, not tool scarcity.
  2. ORSA economics corrected. Draft RTS: full scenario analysis at least every 3 years when material — not annual — and small/non-complex insurers are carved out; EIOPA's July 2025 monitoring says most insurers already cover climate in ORSA (both risk types). It is a replacement sale, and the realistic buyer is the actuarial-consultancy channel, not the insurer directly. From ~Jan 2027, sustainability-risk plans add transition-risk content — the physical-only pack covers a shrinking share of the filing.
  3. OpenQuake is feasible but is a quarter-plus model-integration project, not "an adapter": model selection/calibration/validation + semantic mapping. Good news: ESRM20 is a verified open Greek seismic risk model, EIOPA itself lists OpenQuake as recommended tooling, and GEM sells commercial licenses as the AGPL fallback. Sidecar isolation posture holds (unmodified OQ, process boundary, source-offer honored per instance).
  4. Both kills re-confirmed (openness-as-differentiator; B2B2C novelty), with one caveat now on record: single-digit nat-cat penetration may signal absent demand rather than unclaimed opportunity — 5116/2024 is the bet that demand is being legislated into existence; interviews must test exactly that.

Wave 3 targets (if needed after session 2)

  • Seismic feasibility: OpenQuake outputs → our hazard model; who else wraps GEM.
  • Munich Re credit pricing in €; Guidewire HazardHub pricing as embed benchmarks.
  • Greek data access specifics: ELSTAT microdata terms, NOA/FloodHub, EFAS licensing.
  • Interview-question kit for insurer validation (feeds the corporate-track demo).

Decided / Parked / Killed (waves 1–2)

Decided - Openness alone is not a differentiator (see killed hypothesis); lead with self-hosted white-label + locality + multi-engine. - Regulatory wedge (EIOPA ORSA, GR 5116/2024) enters the persona discussion as the leading business hypothesis — conditioned on the two wave-2 validation gates (earthquake roadmap answer; interview-confirmed demand for per-asset granularity). - CLIMAAX workflows are Apache-2.0 and legally reusable (verified on GitHub). - The insurer pitch must position against bundled reinsurer tools (CatNet/NATHAN) as "workbench vs. atlas" — and that positioning must be tested, not assumed.

Parked - Oasis-compatible packaging as insurer-credibility channel. - CLIMAAX post-project vacuum (revisit ~Q4 2026). - Seismic hazard support — scoping question opened, feasibility unexplored (wave 3).

Killed - "First open-source physical-risk platform" positioning. - "B2B2C embedded risk is globally novel" — mature in the US (HazardHub, Verisk LOCATION); the local vacancy is the opportunity, novelty is not.

Sources (wave 1)

Sources (wave 2)