STATUS
UTC --:--:--
Exploratory research direction

A research-stage weather-decision signal for parametric risk

Model weather triggers, replay them against history and measure forecast skill. This is an exploratory direction a licensed risk carrier could build on, not an insurance product Goable operates.

Where this sits today

Goable is founder-led and pre-revenue. Parametric underwriting is an exploratory research direction on a T2 horizon, not a shipped product. Goable is not an insurer, MGA, reinsurer, coverholder or actuary, and does not quote, bind, price or settle risk. What follows sketches a research-stage technical surface a licensed risk carrier could build on, resting on the parts Goable genuinely does today: forecast verification, provider-skill measurement, historical replay and SPC drift methodology.

01
The why

Traditional vs parametric

Outdoor-business interruption is a classic claims-handling nightmare: subjective loss assessment, photo evidence, multi-month settlements. Parametric flips the contract: pre-agreed trigger, objective measurement, and a settlement a carrier can process quickly. The leg that is hardest is a credible trigger model. That is the layer this research direction explores, as a signal a licensed carrier could price and issue on.

Traditional indemnity
  • Adjuster visits the spot
  • Photos + customer statements
  • 60 to 90 day settlement
  • Loss ratio swings by tens of points per season
Parametric trigger
  • Wind under 12kt for 4h means the trigger is met
  • An agreed evaluation source confirms the outcome
  • A carrier settles quickly on confirmation
  • Loss ratio bounded by trigger frequency
02
Illustrative example

One trigger, Tarifa kite-school, August 2026

Illustrative only. A kitesurfing school in Tarifa wants weather-cancellation cover for its August teaching season: 31 days, 4-hour windows of at least 12kt sustained wind. The sketch below shows how a technical signal could be assembled from historical exceedance, per-cell forecast skill and operator loss loading. A licensed carrier's actuary would own any tariff. Goable prices nothing here.

Trigger definition
POST /v1/underwriting/quote
{
 spot: "kitesurf-tarifa-balneario",
 trigger: {
 metric: "wind_kt",
 operator: "lt",
 threshold: 12,
 duration_hours: 4
 },
 coverage: {
 from: "2026-08-01",
 to: "2026-08-31",
 loss_given_trigger_eur: 11400
 }
}
Illustrative endpoint shape for a research-stage surface a licensed carrier could operate. Not an endpoint Goable runs as an insurer.
Technical signal (illustrative)EUR 2,847.20
  1. Forecast skill (Brier)
    per-cell historical, refit nightly
    0.92
  2. Historical exceedance
    reanalysis multi-decade base rate
    8.4%
  3. Tier risk multiplier
    per-spot tier + tierSource echoed on the response
    1.18
  4. Loss given trigger
    operator-supplied LGT
    EUR 11,400
  5. Carrier loading + IPT
    a licensed carrier's actuary discretion + insurance premium tax
    1.42x
Illustrative research sketch, not an insurance quotation, offer, solicitation or binding terms. Goable does not quote, bind, price or settle risk.
Reading the sketch: Base hazard cost = exceedance x LGT = 0.084 x EUR 11,400 = EUR 957 per day-of-coverage on a perfect-skill basis. Adjusted by 1/0.92 forecast-skill loading to EUR 1,040. Tier multiplier 1.18 reflects the location's historical busy-season cancellation profile, for EUR 1,228 fully-loaded per-day. This is not multiplied flat across the 31-day window: the exceedance rate already reflects how often a trigger actually fires, so the aggregate exceedance-weighted figure, IPT included, lands at EUR 2,847.20. Every multiplier traces to a published metric, so a carrier's actuary could validate the methodology on its own terms.
03
Illustrative demo

Pick a spot. Pick a trigger. See the signal.

An illustrative research demo. It replays your exact trigger against 20+ years of an agreed historical source at the spot you choose. It is not an insurance quotation, does not bind any carrier, and Goable does not price, quote or settle anything from it.

Illustrative research sketch, not an insurance quotation, offer, solicitation or binding terms. Goable does not quote, bind, price or settle risk.

Set a spot, an activity, and a trigger, then hit calculate.

An illustrative research demo. It replays 20+ years of real ERA5 weather at that exact spot. It is not an insurance quotation, and Goable does not quote, bind, price or settle risk.

04
How it could work

Where Goable stops and a carrier begins

In this research sketch, Goable would supply the weather-decision signal, the historical analysis and a drift monitor. A licensed carrier would own pricing, issuance, binding and settlement. None of the steps past the signal are operated by Goable today.

1

Define a trigger

In this research sketch, a carrier would define a parametric trigger: the cell, the metric, the threshold, the duration and the evaluation rule. Everything downstream traces back to this definition.

2

Analyse historical frequency

Replay the exact trigger against decades of an agreed data source to measure how often it would have fired. Same scoring curves the /v1/score endpoint uses today.

3

Assemble a technical signal

Forecast skill, historical exceedance, loss-given-trigger and per-cell tier combine into a technical weather-decision signal. It is research-stage input for a carrier to study, never a premium, quotation or offer from Goable.

4

A licensed carrier prices, issues and binds

The carrier's actuary validates the methodology, sets the tariff, and the licensed carrier issues or binds any policy. Goable does not price, quote or bind. This step sits entirely outside Goable.

5

Replay against the data hierarchy

After a window, a bound cohort could be replayed against a contract-defined evaluation source or documented hierarchy of sources. Idempotent on repeat calls; only new events count.

6

Emit an auditable evaluation signal

The research surface could emit a trigger evaluation event carrying an evidence hash. A licensed carrier, not Goable, decides and settles. Every event is reproducible from the bind snapshot.

Design intent: do not bind on a degrading model. The sketch has a bind step refuse (422 DRIFT_ACTIVE) when the engine has an open warning or critical drift event on the resolved cell. Watch-level drift returns as an advisory. This is a research-stage guardrail a carrier could adopt, not a live underwriting control Goable runs.

05
Evaluation source

The source would be agreed at binding

In the sketch, a policy's evaluation source is agreed at binding by the carrier. Depending on the product it might use an approved observation source, a defined reanalysis dataset, or a documented hierarchy of sources. Reanalysis is not treated as automatic contractual truth: the contract decides.

evaluation_basisillustrative snapshot
{
"policy_id": "pol_a7b2...",
"evaluation_source": {
"source_type": "reanalysis",
"provider": "ECMWF",
"dataset": "ERA5",
"dataset_version": "2025.1",
"rule": "agreed_at_binding",
"fallback_hierarchy": [
"approved_station_observation",
"reanalysis_dataset"
]
},
"bind_snapshot": {
"cohort_hash": "sha256:9f3c8a...",
"calibration_version": "cal_2026_07",
"frozen_at": "2026-07-15T09:00:00Z"
}
}
Reproducible by hash

A bind snapshot would freeze the cohort hash, calibration version and evaluation rule. Every later evaluation cites that hash, so a carrier could re-derive the outcome from the open catalog and the agreed source, without Goable's cooperation.

06
Honest status

What is real today, and what stays with a carrier

Goable has not been audited for regulatory compliance and is not an insurer, MGA, reinsurer, coverholder or actuary. Only the research-foundation rows are real today; everything insurance-facing sits with a licensed carrier and is not built.

Historical trigger replayReal research today
Forecast skill & drift monitoringReal research today
Technical weather signalExploratory (T2)
Contract binding integrationCarrier-operated, not built
Claims settlement integrationCarrier-operated, not built
Insurance issuanceLicensed carrier only
07
Model governance

Research a carrier's review could draw on

This research is designed with insurer, MGA and delegated-authority model-governance frameworks in mind, so a carrier could study it against its own requirements. Goable makes no claim of conformance, readiness or compliance with any regulatory regime, and this is not an attestation or compliance product.

Drift monitoring methodology

The SPC drift charter is real research today. A carrier could study it as a model-monitoring input inside its own governance review, subject to its own validation.

Model-governance inputs

The drift charter, cohort hashes and per-cell skill estimates are structured to feed a carrier's model-governance work, never to replace it or to stand as compliance.

Data-security posture

Pseudonymous and tenant-scoped, audit-log retention configurable per DPA, EU residency by default. A carrier maps this to its own obligations.

Reproducibility

Every technical signal is a function of (cohort hash, calibration version, profile maturity). Same inputs, same number, forever.

Want the SPC drift charter as a JSON-LD or PDF artifact?See the charter
08
Illustrative shape

Evaluation as an event, settlement as a carrier's

Illustrative only. In the sketch, a window expires and an evaluator runs, emitting a trigger evaluation event carrying the policy id, the evaluation basis hash and the triggered outcome. A licensed carrier, not Goable, decides and settles, and reconciles by evaluation-basis hash rather than by spreadsheet.

HMAC-SHA256, same delivery contract as score webhooks
  • Reasonable retry posture (5 attempts, exponential backoff)
  • The signal is bound to the original binding's cohort by hash
  • Idempotent on policyId (replay-safe)
trigger evaluation eventillustrative
{
"event": "underwriting.policy.trigger_evaluated",
"policyId": "pol_a7b2...",
"data": {
"tenantId": "tn_...",
"evaluatedAt": "2026-08-14T11:02:14Z",
"triggerMet": true,
"signal": "trigger_condition_met",
"referenceAmount": 11400,
"referenceCurrency": "EUR",
"evaluationBasisHash": "sha256:9f3c8a...",
"carrierAction": "the licensed carrier decides and settles"
}
}
09
Responsibilities

Where Goable stops and a licensed carrier begins

  • Trigger modelling and definition support
  • Historical trigger-frequency analysis against an agreed source
  • Forecast-skill verification and drift monitoring
  • A technical weather-decision signal (never a premium, quotation or offer)
  • Open methodology for technical due diligence

10
Get in touch

A research collaboration, not a product you can buy

There is nothing to purchase here and no self-serve trigger endpoint. If you are a licensed risk carrier, MGA or reinsurer weighing parametric weather cover, we would share the forecast-verification and drift methodology and let your team probe the weather-decision signal on its merits, as technical due diligence at a research stage. You carry the risk; Goable contributes a signal and open methodology.

Talk research with us Partner programme SPC drift charter