Parametric weather triggers

A research-stage technical surface for parametric weather triggers, driven by the same scoring engine. Goable is not an insurer, MGA, reinsurer or actuary, and does not quote, bind, price or settle risk. These endpoints are a surface a licensed risk carrier could operate, not an insurance product Goable runs.

Where this sits

The endpoints are gated to a contracted Scale tier. Goable makes no claim of conformance, readiness or compliance with any insurance regulatory regime; the surface is designed with parametric-risk model-governance frameworks in mind, nothing more. Talk to us about a research collaboration.

The contract

A policy on this surface is a parametric contract over a single spot or a 2-50-spot portfolio, each with its own fixed payout amount. The trigger is either a score threshold (scoreBelow) or a verdict threshold (verdictAtOrBelow), evaluated hourly across the recurring annual coverage window: each time the condition holds for consecutiveHours in a row (subject to a cooldownHours gap between events and an optional maxPayoutsPerYear cap), that spot's flat payout amount fires. The trigger is evaluated deterministically by the same historical-replay engine that priced it. Pricing, the decision to issue, and settlement of funds sit with a licensed carrier operating the surface, not with Goable.

POST /v1/underwriting/quote

Returns a technical premium figure for a single spot or a 2-50 spot portfolio (a research-stage signal, not an insurance quotation from Goable). The engine replays the covered window against the ERA5 historical archive, walks the trigger over each year to build the expected-payout distribution, applies the per-spot tier risk multiplier, and loads the figure by loadingFactor (default 15%). A licensed carrier's actuary would own any real tariff. Portfolios can mix currencies: each spot carries its own payout, and the response's expectedPremium.byCurrency breaks the figure down per currency (no FX conversion, each currency settles independently).

{
 "portfolio": [
 { "point": { "lat": 36.013, "lng": -5.604 },
     "activity": "kitesurfing",
     "spotId": "tarifa-balneario", // optional, your own label
     "payout": { "amount": 150, "currency": "EUR" }
   },
   … 1-49 additional spots
 ],
 "coverageWindow": { "monthFrom": 6, "dayFrom": 1, "monthTo": 9, "dayTo": 30 }, // recurring calendar window, not a fixed date range
 "trigger": { "kind": "scoreBelow", "scoreBelow": 55, "consecutiveHours": 4 },
 "loadingFactor": 0.15
}

trigger is a discriminated union: either { kind: "scoreBelow", scoreBelow: <0-100> } or { kind: "verdictAtOrBelow", verdict: "unsafe" | "not_feasible" | "poor" | "marginal" | "fair" | "favorable" | "excellent" }. Both accept optional consecutiveHours, cooldownHours, and maxPayoutsPerYear. There is no per-spot "limit" field: the payout schedule is driven entirely by the trigger plus the per-spot payout amount, walked across every hour in the coverage window that satisfies the trigger.

Response carries policy (the echoed + resolved terms, including each spot's tier/tierSource, see below), expectedPayouts.byCurrency (mean / p10 / p50 / p90 distribution), expectedPremium.byCurrency (fair + loaded + tierMultiplierWeightedAvg per currency), modelConfidence, advisoryLevel, issuable, and underlying (the historical score distribution the figure was priced against). There is no cryptographic cohort hash on the quote: reproducibility comes from the persisted policyTerms snapshot (see Settlement below), not a hash.

POST /v1/underwriting/policy/bind

Convert a quote into a bound policy record. Submit the quoteId from a recent quote (default TTL 24h) + coverageYear + premiumCollection (external / stripe / invoice_due). The response echoes the resolved policy (a snapshot of the priced terms, the reproducibility anchor for settlement) + the quoteId. Fires the underwriting.policy.bound webhook. Whether to bind, issue or price a real policy is a licensed carrier's decision, not Goable's.

import { DriftActiveError } from "@goable-io/sdk"

try {
 const policy = await goable.bindPolicy({ quoteId })
 if (policy.driftAdvisories?.length) // watch-level, non-blocking
 console.warn("drift watch", policy.driftAdvisories)
} catch (err) {
 if (err instanceof DriftActiveError) {
   // refused: cell has open warning/critical
   console.error(err.openDriftEvents)
 }
}

curl -X POST https://api.goable.io/v1/underwriting/policy/bind
 -H "X-Goable-Key: $GOABLE_KEY" \
 -H "Content-Type: application/json" \
 -d '{
 "quoteId": "e12b…",
 "coverageYear": 2027,
 "premiumCollection": "external",
 "premiumPaid": 1750
 }'

The quote response, echoed on every policy read, surfaces the resolved tier per spot with a tierSource flag: "catalog" when the tier came from the versioned catalogue, "classifier"when the nightly data-driven classification overrode the catalogue at full confidence. Any tier > 1 lifts the tierMultiplierWeightedAvg above 1.0, see the tier multipliers section below.

POST /v1/underwriting/policy/:id/evaluate

Replays the bound policy's trigger against the historical provider (ERA5) and inserts new payout events into the policy record. Idempotent on repeat calls: the same event set surfaces twice with inserted: 0 the second time. Fires the underwriting.policy.triggered webhook only when new events land in a given evaluation. Response includes the up-to-date policy record + all payout events on the policy so far + the inserted and skipped counts for this call.

Tier multipliers

Sub-spot resolution matters for the pricing figure. A tier-1 spot (≥150 paired outcomes, calibrated) prices at the base multiplier; tier-2 (cluster-level) and tier-3 (regional fallback) carry uncertainty loadings. The tierMultiplierWeightedAvg in the quote response shows the portfolio-wide multiplier. Configure rates in apps/api/config/tier_risk_multipliers.yaml.

Settlement: POST /v1/underwriting/policy/:id/settle

Settlement is not automatic. A separate admin cron (POST /cron/expire-windows, platform_admin only) sweeps and marks stale untriggered policies as expired, but the payout math itself runs via POST /policy/:id/evaluate (replays the bound policy against the ERA5 historical archive, idempotently inserting new payout events). Settlement is finalized by whoever operates this surface as a licensed carrier, calling this endpoint with a settlementReference. It requires the platform_admin scope: recording a settlement is a carrier operation, not something a policyholder tenant does to its own policy, so clients never call it directly. Goable is not the carrier and does not settle risk. Fires the underwriting.policy.settled webhook.