Explainability & Governance
Building trust through transparent, auditable recommendations
OverviewData & FeaturesModels & AlgorithmsDecisioning PipelineReal-time ArchitectureExplainabilityFeedback LoopMetrics & Impact
"Why Recommended" Panel
1Travelers
94% match- 12% below market price for this profile
- 87% historical win rate for similar risks
- Strong retention track record (92%)
- Agent has 3x higher close rate with Travelers
2Hartford
89% match- Competitive pricing (8% below market)
- Good appetite fit for this risk class
- Fast quote turnaround (< 2 hours)
3CNA
78% match- Moderate pricing (at market)
- Strong coverage options
- Lower historical win rate for agent
Feature Importance (SHAP Values)
Price Competitiveness
35%
Carrier Win Rate
25%
Agent Performance
15%
Risk Match Score
12%
Retention History
8%
Payment Reliability
5%
SHAP (SHapley Additive exPlanations) values show how each feature contributes to the recommendation score.
Governance Framework
Audit Trail
- Log every recommendation with inputs
- Store feature values at prediction time
- Record model version used
- Retain for compliance (7 years)
Bias Monitoring
- Track outcomes by demographic
- Flag disparate impact patterns
- Regular fairness audits
- Human review for edge cases
Human Override
- Agents can override recommendations
- Override reasons captured
- Feedback loop for model improvement
- Escalation path for disputes
Trust Signals for End Users
94%
Confidence Score
Model certainty in recommendation
< 24h
Data Freshness
Age of underlying data
2,847
Similar Profiles
Training examples for this segment
87%
Historical Accuracy
Past prediction success rate