Overview
PORE uses a weighted aggregation model to convert weak signals from multiple operational data sources into a unified risk index. The methodology prioritizes recency, severity, and source credibility to produce a forward-looking indicator of operational risk.
Core Formula
Daily Risk Score Calculation
Risk_Score(day) = Σ [W_source × M_severity × D_recency × 10]
Where the sum is taken over all observations recorded on that day, and each component is defined below.
Parameters
W_source
Source weight reflects data quality and predictive value. Safety observations weighted higher (default 1.2) than maintenance notes (0.9).
M_severity
Severity multiplier scales contribution by consequence potential. High/Actual events (1.5×), Medium/Potential (1.2×), Low (1.0×).
D_recency
Recency decay applies exponential weighting. Recent signals matter more. D = 0.5^(days_ago / half_life).
Recency Decay Function
D_recency = 0.5^(days_since_observation / half_life)
Default half_life = 14 days
Adjustable range: 7-30 days
This exponential decay ensures older observations contribute progressively less to current risk assessments, reflecting the time-sensitive nature of operational hazards.
Trend Calculations
Moving Average Trends
7-day MA = (Σ scores_last_7_days) / 7
30-day MA = (Σ scores_last_30_days) / 30
Trend % = ((Recent_MA - Previous_MA) / Previous_MA) × 100
Moving averages smooth daily volatility and reveal underlying trends. Divergence between 7-day and 30-day MAs indicates drift.
Drift Signal Detection
Sustained Deterioration: Triggered when 7-day MA exceeds 30-day MA by >15% for 5+ consecutive days.
Volatility Spike: Detected when standard deviation of last 14 days exceeds threshold (σ > 8).
Step Change: Identified when a specific risk theme shows >8 observations in 14 days, indicating emerging pattern.
Predictive Forecast
The 14-day forecast uses exponential smoothing with trend adjustment:
Forecast(t+n) = Level(t) + n × Trend(t)
Level(t) = α × Score(t) + (1-α) × Level(t-1)
Trend(t) = β × (Level(t) - Level(t-1)) + (1-β) × Trend(t-1)
α = 0.3 (level smoothing)
β = 0.1 (trend smoothing)
Confidence intervals calculated using historical forecast error variance (95% CI = ±1.96σ).
Benchmark Methodology
Industry benchmarks derived from anonymized UKCS operator data (n=23 installations). Current index positioned relative to P25 (lower quartile), P50 (median), and P75 (upper quartile) of peer distribution.
Limitations & Assumptions
• Model assumes reporting consistency across time periods
• Does not account for exposure hours or production volumes
• Requires minimum 30 days of data for reliable trending
• Forecast accuracy degrades beyond 14-day horizon
• Benchmark data refreshed quarterly (last update: Q4 2025)