← Back to Noah Predict
FOR INSURERS

Political risk intelligence for insurers.

Structured insight into political violence, terrorism, civil unrest and personnel risk — designed for underwriting and exposure assessment.

Insurance desk NOAH PREDICT · Predictive political risk intelligence
The lag problem

By the time a view reaches a quarterly briefing, the signal is weeks old.

Broker consensus, country-risk reports and expert surveys are useful reference points. They are also slow. Political violence does not deteriorate on a quarterly reporting cadence. By the time a view appears in a published briefing, the underlying signal has been detectable in the wider information system for days — often weeks.

Noah sits underneath that reporting layer. Continuous ingestion of roughly 1.6 million sources per refresh, paragraph-level AI sentiment, and a structured physics layer measuring velocity, trajectory, divergence, volatility and coupling. The output is decision-ready: a composite 0–100 score with documented weights, a peril breakdown, sub-national divergence, peer comparison, and a concrete underwriting verdict.

Every run is traceable to paragraph-level source and cross-referenced against six independent external benchmarks.

What lands on your desk

Every run carries the same underwriter-grade structure.

Noah is not a chatbot and not a news summariser. It is a structured-output system whose format is designed for underwriting use.

01 · Composite

0–100 score with documented weights

A weighted composite across nine perils, with the exact weights published in weights_doc on every run. No black box. Current weighting: Terrorism 0.20 · War 0.17 · Political Violence 0.15 · K&R 0.13 · SRCC 0.10 · Non-Payment 0.08 · Currency 0.07 · MedEvac 0.06 · Expropriation 0.04.

02 · Peril sheet

Nine perils, each scored 0–10 with signal basis

Every peril carries a signal_basis attribution (which signal families drove the score and in what proportions) and an action_trigger — a concrete underwriting action such as “Load K&R, PV/T” or “+15–25% load; deductible USD 25k → 50k”.

03 · Sub-national

District-level scoring with hard-data counts

Each district carries a composite 0–100 plus kidnap_freq_90d, ied_within_5km_un, ingo_convoy_incidents_90d, and hospital_mci_capacity — the numeric inputs an in-house exposure team would want to see.

04 · Verdict per coverage

Per district, per line of business

A verdict_per_coverage block for each district, each row reading bind, load N%, exclude or decline for K&R, MedEvac, PA, PV/T, Property and War. Working-paper-ready.

05 · Peer comparison

Calibrated against the neighbouring book

Every country-level composite can carry a peer set drawn from the returned packet for the territory being analysed — with per-peril sub-scores and a rank note per peer when Noah has genuinely comparable evidence. The number you see should be anchored against the live frame, not a hardcoded demo country.

06 · Horizons

Forecast read over 7 days / 4 weeks / 3 months

Three horizons on every report. Each comes with a direction label (rising / flat / falling) and a one-line rationale tied to signal velocity. The short, medium and renewal-window reads in one place.

07 · Invalidation

What would change Noah’s mind

Three to five calibrated pivot points with expected magnitude — “al-Shabaab captures a district capital, composite rises to 93+”, “aid-worker K&R with ransom >USD 2m, re-rate +30% minimum”. Falsifiable, not hedged.

08 · Executive verdict

Ends with a concrete pricing action

Every brief ends in underwriter language: “+15–25% on K&R renewals; exclude Lower Shabelle on PV/T; retain USD 1m MedEvac sub-limit; no new binders in Middle Juba without fac re.” One sentence, five decisions.

Benchmarked against

Six independent external anchors, attached to every bundle.

Noah is designed to be cross-checked. Every country-level bundle ships with an _external_anchors block so you (or your reviewing LLM) can independently verify the read against reference points that already sit in your workflow.

Official advisory
FCDO

UK Foreign, Commonwealth and Development Office travel advice. Level, regions-advised-against, summary text.

Official advisory
US State Dept

Travel advisories at Level 1–4 with the nine indicator codes — Terrorism, Crime, Kidnapping, Civil Unrest, Health, Natural Disaster, Time-limited Event, Wrongful Detention, Other.

Peace index
Global Peace Index

IEP annual ranking 1–163 with overall score and three domain sub-scores: ongoing conflict, societal safety, militarisation.

Governance
World Bank WGI

Worldwide Governance Indicators on six axes: political stability, rule of law, control of corruption, government effectiveness, regulatory quality, voice and accountability.

Sanctions
OFAC & UK HMT

Sovereign-level designations and named-entity designations via Risk Atlas, with last-update timestamp. For the compliance layer, not the headline.

Event data
ACLED

Armed Conflict Location and Event Data — event counts, fatalities, top actors and top admin-1 regions over 30, 90 and 365-day windows.

Marsh-style posture Noah produces the structure of a Marsh World Risk Review — WRR-adjacent peril tables, peer comparison, district-level verdicts, rate-band language — without licensing Marsh WRR data. Rate-band suggestions are Noah-derived, not Marsh-sourced. The phrase “Marsh-style” denotes the shape of the output, not the data source.
Where it lands

Four moments in the underwriting cycle.

Noah is read and cited at four points in the cycle. Each moment maps to a specific subset of the output.

Moment 01

Risk selection

Is this environment deteriorating, stabilising, or transitioning? Does the peer comparison support the rate I’m being asked to write at?

  • Composite 0–100 with 90-day delta and 12-month baseline percentile
  • Peer comparison table with per-peril sub-scores
  • Regime classification (fragile calm, pressure release, political crisis, stabilisation, regime instability)
  • External anchor agreement matrix
Moment 02

Renewal decisions

Do last year’s pricing assumptions still hold? Is the book drifting in a direction that warrants a re-rate?

  • Horizon views at 7d / 4w / 3m calibrated to the renewal window
  • Peril-by-peril action triggers (load, exclude, retain, decline)
  • Invalidation triggers with expected composite movement
  • Executive verdict ending in a concrete pricing action
Moment 03

Exposure monitoring

Where is the book concentrating? Which sub-region of which country is moving faster than the national headline?

  • Sub-national divergence: district composites with hard-data counts
  • Tremors and relief: granular findings that passed the four-gate specificity filter
  • Cross-peril couplings (when two perils co-move, exposure stacks)
  • Source-mix transparency panel
Moment 04

Challenge & validation

Does the internal model’s read hold up against an independent signal-led view? Where does the analyst’s draft disagree with the corpus?

  • Contrarian read, mandatory on every run
  • Blind spot, named for each answer (what the corpus may be missing)
  • Full provenance trace to paragraph-level evidence
  • Downloadable bundle with methodology for peer review
The key point

It does not replace underwriting judgement.

Noah does not replace underwriting judgement.

It improves it — by providing a clearer view of how risk is actually moving, before consensus catches up.

Every output carries a mandatory contrarian read, a named blind spot, and a confidence band that is explicit rather than hedged into the prose. The audit is the point: this is a tool for underwriters who want to be able to show their working.

Better visibility.

Better decisions. Better outcomes.

Elsewhere on Noah Predict

More from the system.

Free while in preview
Official systems show the current state of risk.
Noah shows where it is moving.

Google sign-in. Three free analyses a day. No card required.