AIDA cheat sheet

Combined ratio = loss ratio + expense ratio; a value below 100% indicates underwriting profit before investment income.

Loss ratio = incurred losses ÷ earned premium; expense ratio = underwriting expenses ÷ written premium.

Claims analytics decision rule: route claims to special investigation units when fraud score exceeds the calibrated threshold and expected savings exceed investigation cost.

Predictive underwriting models require out-of-time and out-of-sample validation; discriminatory power is often measured by Gini or AUC, while calibration compares predicted and observed loss costs.

Data quality dimensions include completeness, accuracy, timeliness, consistency, validity, and uniqueness; acceptance thresholds are set per dimension.

Telematics and IoT analytics support usage-based insurance by converting driving behavior into rating variables; regulators may require actuarial justification and non-discriminatory use.

Fraud analytics commonly combines rules, anomaly detection, network analysis, and supervised models; alert capacity is matched to investigative resources.

Model monitoring detects population drift, label drift, and performance decay; retraining or decommissioning is triggered when metrics breach pre-set tolerances.

Customer analytics uses retention, cross-sell, and lifetime value metrics; next-best-action models rank offers by expected incremental value, not just propensity.

ROI = (cumulative benefits − cumulative costs) ÷ cumulative costs; NPV discounts net cash flows at the organization's cost of capital.

Payback period = time until cumulative benefits equal cumulative costs; shorter payback reduces strategic risk but ignores value after payback.

Use-case prioritization matrix ranks initiatives by business value and delivery feasibility; high-value, high-feasibility items enter the first wave.

A data strategy specifies data domains, ownership, quality standards, architecture, analytics capabilities, governance, and funding; missing ownership is a common failure point.

Data governance assigns accountability for data quality, access, privacy, and model risk; RACI or similar roles clarify decision rights.

Benefits realization requires a baseline, target KPI, measurement frequency, and owner; without a baseline ROI cannot be verified.

Pilot-to-scale decision rule: scale only when pilot results meet pre-agreed success criteria, operational capacity exists, and risks are controlled.

Change management addresses stakeholder incentives, skills, process redesign, and communication; adoption metrics are tracked alongside financial metrics.

Build-versus-buy decision rule: choose buy when requirements are common and speed matters; choose build when the capability is differentiating and data or process control is strategic.