Methodology

We use a hybrid rule-based and data-driven approach to ensure scores are:

• Deterministic: Same inputs produce same outputs
• Explainable: Every score can be traced to its sources
• Auditable: All calculations are logged and versioned
• Neutral: No external influence on scores

Raw data from different sources is normalized to ensure fair comparison:

• Crime data normalized per 1,000 residents
• AQI interpolated by distance from monitoring stations
• School quality standardized across boards and types
• Commute times adjusted for time-of-day patterns
• Seasonal variations accounted for in environmental scores

Crime and safety scores receive special treatment to ensure responsible reporting:

• Crime types are weighted by severity (non-violent vs violent)
• Trends are smoothed over time to avoid short-term fluctuations
• Policing context and response times are factored in
• Confidence is reduced when data is sparse
• No incident-level mapping or sensational language

Every score includes:

• Version: Methodology version used (e.g., v2.3)
• Data Window: Time period of data used
• Methodology Hash: Cryptographic hash of calculation logic
• Last Updated: When the score was last refreshed

This ensures users can see exactly how and when scores were calculated, and track changes over time.