ESSAYS · WHITE PAPERS · APPLIED RESEARCH

Research on motor pricing, fraud and insurance AI.

Executive summaries of my longer research: the commercial question, the operating logic and the decision an insurer can take next. Download the original papers for the full evidence and methodology.

01 FRONTIER RESEARCH

Motor pricing and fraud research.

These papers examine how geographic data can improve motor pricing and how evidence from several sources can support fraud investigations.

JOHAN STRYDOM · 7 JANUARY 2026 · 25 PAGES · UK MOTOR PRICING

Beyond Postcodes

Advanced Geo-Enrichment for UK Motor Insurance Pricing

Download white paper Read the essay

EXECUTIVE SUMMARY

A postcode is a container of risk signals, not a cause of loss.

Traditional postcode factors compress mileage, road environment, behaviour, claims severity, climate and price response into one historic label. That may fit yesterday's loss experience, but it obscures why risk differs and can weaken stability, fairness and governance.

The paper proposes decomposing location into seven independently governed layers. Exposure is separated from risk per mile; frequency is separated from severity; and correlated tail risk is made visible rather than hidden in an average postcode relativity.

For executives, the opportunity is broader than model lift. The same transparent feature system can support pricing, portfolio steering, capital allocation, reinsurance and product design - while leaving the insurer in control of which signals belong in each decision.

Postcode rating is not obsolete - but postcode thinking is.

THE MODELFrom a label to decision-ready risk layers

STARTPostcodeA useful join key

01Exposure & usage

02Per-mile accident risk

03Severity amplification

04Behaviour & moral risk

05Temporal dynamics

06Tail & correlation

07Elasticity & optimisation

FrequencySeverityTail riskCapitalGrowth

STEFANUS KORFF & JOHAN STRYDOM · 12 JANUARY 2026 · 14 PAGES · FRAUD & VALIDATION

Guardian Guru

A New Era in UK Insurance Fraud Prevention

Download white paper Read the essay

EXECUTIVE SUMMARY

Fraud rarely reveals itself through one signal.

Rules, pooled databases and digital-identity networks remain essential, but they are strongest at known patterns or one dimension of risk. Novel schemes can sit across documents, images, narratives, devices and relationships - leaving each individual signal below a referral threshold.

Guardian Guru is presented as a coordinated investigation team. Specialist agents examine different forms of evidence in parallel, then corroborate their findings before a central decision is made. Validation Guru checks customer, vehicle and document data at entry, helping stop false information before it becomes a policy or claim problem.

The executive case is fewer fraudulent payouts without turning every genuine customer into a suspect: faster low-risk journeys, more focused investigations, evidence-led referrals and a full audit trail for human review.

£1.16bnfraudulent UK general insurance claims detected in 2024, as quoted in the paper from ABI reporting.

THE OPERATING MODELCorroborate signals before acting

Validate the evidenceIdentity · vehicle · MOT · device · documents

02Specialists work in parallel

IdentityTextVoiceVisionBehaviourNetworks

Corroborate and explainEvidence-linked risk decision · human-readable rationale

ApproveVerifyReferInvestigate

A NOTE ON THE RESEARCH

These papers set out frameworks and product propositions. Any predictive uplift, operational saving or fraud outcome must be established on the insurer's own portfolio, data and implementation.

THE EXPONENTIAL INSURANCE BRIEF

One technology development each week, with its implications for pricing, claims and insurance products.

Subscribe free Free on Substack, with links to the underlying sources.

Research on motor pricing, fraud and insurance AI.

Motor pricing and fraud research.

A postcode is a container of risk signals, not a cause of loss.

Fraud rarely reveals itself through one signal.

Get the Brief each week.