Methodology · Research · Innovation
Aviation risk is not adequately understood by evaluating individual measures in isolation. Andre's models assess the dynamic interaction between the risks in a pilot, an aircraft, an operation and its surroundings, because that interaction is where accidents are made and where they can be prevented.
Interactive
Move the sliders to see how elevated risks interact. The visualisation is illustrative only. The dimensions the models actually use, their weightings, thresholds and scoring are proprietary and are not disclosed on this site.
Principle
Risks interact; they do not add up
Move the sliders. Two moderately elevated risks are often more dangerous than one severe risk on its own, because each removes the margin the other depended on.
Principles
A single measure, however important, never explains an aviation exposure. The models assess how each risk changes the others: a strength in one area can absorb a weakness elsewhere, and two moderate weaknesses together can be more dangerous than one severe one.
Every dimension in the models traces back to accident data, to the 983 official investigation reports analysed at Cranfield, to published human-factors research and to years of underwriting outcomes. Weightings are revised as the evidence changes.
An assessment is a snapshot. Operational data, flight data monitoring and recency measures allow the models to recognise a changing risk between renewals, which is the point at which intervention still makes a difference.
The models record the evidence behind each conclusion so that two assessors reach the same answer and an underwriter, a reinsurer or a court can see how it was reached.
Origins
The models grew out of two bodies of experience: more than a decade of underwriting pilots and aircraft for a South African life insurer, and the analysis of 983 official general-aviation accident reports from the United Kingdom and South Africa for Andre's Cranfield MSc thesis.
The thesis found correlations between pilot age, total and recent experience, licence type, phase of flight and injury severity, and a significantly higher incidence of powerplant-related occurrences in South Africa. Several of those correlations are exactly the kind of fact that a single-measure assessment misses: a pilot's age matters more at low recency; an engine failure matters more over hostile terrain; a training flight carries a different risk signature from a cross-country flight in the same aircraft.
In underwriting practice the models replaced exposure proxies, such as annual hours flown, with a structured assessment of the interacting risks, supported by logbook scrutiny, licence and medical verification and territorial data. They are the basis of the aviation underwriting philosophy Andre formulated for South Africa's largest life insurer from 2018.
The method was then built into software: an assessment tool that scores each dimension, records the evidence behind the score, and produces a consistent, auditable risk opinion. SkyGuard extends the same approach to live operational data. Flight data monitoring supplies actual behaviour rather than declared intent, so that risk can be reassessed as operations change rather than once a year at renewal.
The models, their structure, scoring methodology, underwriting thresholds, questionnaires and software implementations are the intellectual property of Andre Roos and SkyGuard. This page describes the approach in general terms only; the dimensions assessed, proprietary algorithms, thresholds, source code and confidential underwriting methodology are not disclosed, and no commercial results or loss ratios are claimed here.
Enquiries are treated in confidence. Initial consultations establish scope, independence and conflicts before any engagement.