
🛡️ Insurance · Data Flow
How an insurer scores incoming claims for fraud: data enrichment, rules, an ML model and a special investigation unit queue.
Drawing diagram…
Data flow for claims fraud scoring: new claims from the claims system are enriched with policy data, the customer's claim history, garage and hospital networks and industry fraud registry hits, then scored by business rules (claim soon after policy start, repeated garages) and an ML model. Low scores go straight to settlement, high scores to the special investigation unit (SIU) queue; investigator findings become labels for model retraining.
flowchart LR CL[New Claims] --> EN[Enrichment] POL[(Policy Data)] --> EN HIS[(Claim History)] --> EN NET[(Garage and Hospital Networks)] --> EN REG[(Industry Fraud Registry)] --> EN EN --> RU[Rules: early claim, repeat garages] EN --> ML[ML Fraud Model] RU --> SC[Combined Score] ML --> SC SC -->|Low| ST[Straight-through Settlement] SC -->|High| SIU[Investigation Unit Queue] SIU -->|Findings| LB[(Labels)] LB -->|Retraining| ML
The states of an insurance claim from first notice of loss to payment or closure, including investigation and appeals.
The core system of an insurer: quoting, underwriting, policy issue, billing and claims, used by customers, agents and underwriters.
How a life or health insurance proposal is underwritten: automated rules, medical tests where needed, and the underwriter's decision.
How a customer buys motor insurance online: quote, add-ons, vehicle and previous-policy checks, payment and instant policy issue.
A relational schema for an insurer: customers, policies, coverages, insured objects, premiums, claims and payments.
What happens between a policyholder, the insurer's app, a surveyor and a garage when a car insurance claim is made.