Methodology
How does the Crisis Index work?
Overview
The Nyman Intelligence Crisis Index uses proprietary machine learning techniques to monitor geopolitical risk across five global regions in near real-time. Each region receives a composite risk score from 0 to 100, derived from multiple analytical components.
Scoring Components
The composite score is built from independent analytical signals, each capturing a different dimension of crisis activity:
Event Volume
Measures the intensity of news coverage for a region, scaled to account for baseline variation across different geographies.
Conflict Intensity
Analyses the proportion of coverage containing conflict-related language, using proprietary machine learning models and semantic analytics tuned per region.
Additional scoring components are applied depending on the scoring version and region configuration.
Data Sources
The index ingests data from a wide range of global news sources spanning multiple languages and geographies. Articles are processed and scored within hours of publication, providing near real-time situational awareness.
Update Frequency
Scores are updated multiple times per day to capture developing situations. Historical data is retained for trend analysis and backtesting.
Research Foundation
The methodology builds on techniques developed through years of crisis monitoring and text analysis research, including real-time conflict detection in active war zones and systemic risk analysis for financial institutions.
Event Detection Pipeline
The event detection pipeline combines multilingual natural language processing with graph-theoretic methods to identify and track emerging crisis events across hundreds of intelligence feeds in near real-time. Articles are processed through multiple analytical stages that assess semantic relatedness, temporal proximity, and geographic coherence before being grouped into distinct events. Each event is continuously updated as new reporting emerges, with machine learning models generating structured summaries and severity assessments.
Source Intelligence Layer
Unlike simple news aggregators, Crisis Terminal continuously evaluates every source in our network across four dimensions:
- •Reliability — how consistently a source aligns with the verified event narrative, measured using a Bayesian Beta-Binomial model with credible intervals
- •First Mover — how often a source is the first to report on an event, indicating original reporting capability
- •Authority — the overall source quality score, combining Bayesian reliability (60%) and Kalman-filtered response speed (40%), with propagated uncertainty estimates
- •Coverage — total number of events a source has contributed to, indicating breadth of reporting
Events confirmed by multiple independent high-reliability sources receive higher confidence ratings. Sources are classified into tiers based on their track record:
- Verified — consistently high reliability and frequently among the first to report
- Established — strong track record with broad coverage
- Standard — monitored but limited history or lower consistency
This proprietary layer ensures that the intelligence you receive reflects verified, multi-source consensus rather than unfiltered noise. Source assessments are updated daily as new events are detected and analysed.
Limitations
- •Reflects media coverage patterns, not direct field intelligence
- •Monitors current state and does not predict future events
- •Scores should be interpreted alongside other intelligence sources