What we do

---

Analysis, models, even results with uncertainty intervals, do not directly yield a decision. We provide user-friendly interfaces for your team to explore our outputs. 

Our decision-support tools give the freedom to define scenarios, weigh different options, assess the potential impact and cost of different choices and, importantly, to validate against what did, and didn't, happen.

The analyses behind our tools synthesise your data — your surveillance, your exposure, your constraints — with our own curated data, using cutting-edge machine learning methods to maximise insight and minimise bias.

Our offer: we provide a range of licensing and front-end customisation options to meet your needs.

We generate AI-driven probabilistic forecasts to support decision-making, describing the likelihood of a range of outcomes, not a single prediction. 

We set the resolution, in space and time, against what the decision requires and what the data can support — availability, representativeness and quality. 

We work with diverse and often incomplete data streams: inconsistent definitions, administrative geography that shifts, and the unglamorous reconciliation that this requires.  We can also assess what a new data source is worth before you commit to it. 

Our offer: a validated pipeline with quantified uncertainty, running in your environment or ours, documented for your team to own.

We build loss and exceedance models for perils where observation history is limited and the claims record is short, thin or no longer representative — European wildfire, biological hazards, and risk to standing forest. We derive frequency and severity from the processes that generate them, calibrating with a range of state-of-the-art analytical approaches. 

We package our models for the open Oasis loss-modelling framework, already run by reinsurers and insurance-linked securities investors, so the model sits inside your existing pipeline.

Our offer: event loss tables, exceedance-probability curves and average annual loss, with the documentation for full validation. 

We validate models for carriers who need to justify a view of risk under Solvency II or Lloyd's oversight, and for anyone who wants an independent read on a model they depend on.

We separate process risk from parameter risk: one is priced, the other is reduced by better data or held as an explicit margin. We test whether stated intervals are calibrated. Where models disagree, the disagreement is reported as a quantity in its own right.

Where a cover is parametric, we quantify basis risk: how closely the trigger index tracks the loss it stands in for, and the conditions under which the two diverge. Estimated at design, the size of the gap is known before the cover is written.

Our offer: an independent validation report, a decomposition of the uncertainty in the estimate, calibration testing, and a written view defensible to a regulator or a validation committee.