Method & Sources
How PowerFlow
measures power
A geopolitical intelligence engine that turns sourced reporting and structural data into a living model of who holds power, how far it reaches, and where it is shifting.
What PowerFlow is
Start with what shifted. Go deeper and the structure appears: who depends on whom, where leverage concentrates, where fractures are forming. PowerFlow reveals complexity as you are ready for it.
Traditional reporting
Reactive. Each cycle resets. No accumulation, no structure.
PowerFlow
Anticipatory. Every event accumulates into a living model. Structure compounds.
Not a news site
Events are inputs to the model, not the product. PowerFlow shows what shifted and why it matters structurally.
Not a dashboard
Dashboards display state. PowerFlow models dynamics: trajectories, dependencies, cascades. A snapshot versus a system.
Not vibes
Every score traces to sourced intelligence or structured data, calibrated across the whole registry. No sentiment, no crowd wisdom.
The PF Score
Every actor in the registry carries a PF Score built from two pillars: Authority (how much power an actor holds at home) and Reach (how far that power extends beyond its borders). A high score needs both.
Authority
Power at home. How much weight an actor carries and how firmly it controls it.
Reach
Power abroad. The actor’s position in the system, not its raw size.
The PF Score is the geometric mean of the two, scored 0 to 100. The geometric mean punishes imbalance on purpose. Overwhelming control with no external reach still scores low, because real power needs both.
PF Score = √(Authority × Reach)
The two pillars
Each pillar breaks into parts you can see on any actor page, in the Power Profile. Authority is built from size and control. Reach is built from position.
Authority = Weight × Grip
The power an actor can bring to bear at home. Two parts, multiplied together.
Weight. The base an actor has to work with: economy, population, and military size.
Grip. How firmly the government holds that base, measured as control and stability, not ideology. Tight autocracies hold through coercion, healthy democracies through cohesion. Only contested or failing states score low. Grip acts as a multiplier, so a small secure state uses its base fully, while a large state with a shaky hold is discounted but never zeroed.
Reach = position, not size
External consequence, built from how central an actor is to the world rather than how big it is. Three channels feed it.
Trade centrality. How much of global trade routes through an actor, measured on the bilateral trade graph. This is brokerage, not volume, so a hub the size of Singapore can outrank a larger but peripheral economy.
Relationship leverage. The pull an actor holds across PowerFlow’s own relationship graph: alliances, patronage, and dependency.
Military projection. Structural reach instruments: nuclear forces, overseas basing, intelligence tier, alliance leadership, and arms exports.
Why position, not size
Reach weights the three channels at roughly half trade, a quarter relationships, a quarter military. Trade volume would just re-encode size, so Reach measures brokerage instead: how much of the world routes through you. That is why pivotal small actors like Israel and Singapore rank well above their raw mass, and why a large but isolated economy does not.
How a score is set
A score means nothing in isolation, so PowerFlow does not set one in isolation.
One scale for everyone
Every actor’s pillars are computed on the same scale and normalized across the whole registry at once. A score is a position relative to every other actor, not an absolute verdict. This is what keeps the board from collapsing into a clump where most of the world looks the same.
A durable floor, a bounded move
The structural inputs set a durable floor: where an actor sits given its size, grip, trade position, relationships, and military weight. On top of that floor, the model reads the period’s intelligence and applies a bounded move. Authority shifts by at most a few points at a time, Reach a little more. Only a named rupture, a coup, a war’s onset, a sovereign default, a patron’s collapse, unlocks a larger jump. Day-to-day noise cannot whipsaw a score, but a real structural break shows up fast.
Every move has a reason
A score that moves cites the event that moved it. Nothing shifts on sentiment, narrative, or a feeling that an actor is having a good week.
Relationship scoring
Power does not exist in isolation. PowerFlow scores bilateral relationships between every pair of collective actors that appear together in ingested intelligence. Each relationship is scored in both directions, because the leverage A holds over B is not the same as the leverage B holds over A.
Three dimensions capture the structural texture of each relationship:
Political and strategic alignment. Allies near +100, active adversaries near -100, transactional neutrality at zero.
How much one actor can constrain or enable the other through economic, military, or diplomatic instruments.
How exposed one actor is to the other. Existential reliance at 100, minimal exposure at zero. Vulnerability, not alliance.
Sources and evidence
PowerFlow does not treat any single feed as ground truth. It combines live reporting, slow-moving structural data, its own relationship graph, and conflict records. Each layer does a different job on purpose.
Why this is not a bibliography
Most of the product leads with interpretation, not footnotes. This section shows the evidence behind that interpretation so trust is earned, without turning every screen into citations.
Live reporting
The stream of who did what, screened and translated before it becomes model evidence.
Event Registry and a curated outlet policy. Breadth of intake from monitored sources, feeding events, intel items, score movement, and briefs. No single article sets a score by itself.
Structural baselines
Slow-moving denominators that anchor actors in real capacity, behind the Authority pillar.
IMF. GDP and population, the backbone of Weight, plus reserve-currency share. The heaviest structural source in the model.
World Bank. Government effectiveness and political stability, the backbone of Grip, plus development, debt, and logistics context.
SIPRI. Military spending, part of Weight, and a coercive-apparatus signal for Grip.
V-Dem. Regime and governance context. Its democracy indices are deliberately kept out of scoring, so an actor is never rewarded or penalized for its system of government.
Relationship evidence
Directional data that makes leverage and dependency less subjective, behind relationship pages and Reach.
IMF trade and investment. Bilateral trade, the basis for trade centrality in Reach, plus cross-border investment positions.
SIPRI, OECD, and World Bank. Arms transfers, development finance, and creditor exposure: who supplies, funds, and lends to whom.
UN Comtrade. Deep-history goods trade, used as a fallback where the primary trade source is thin.
Conflict substrate
Structured depth for conflict pages and query answers, beyond the daily headlines.
PowerFlow conflict records. Each conflict carries authored historical context and a live event stream drawn from the PowerFlow pipeline.
UCDP and ACLED. Roadmapped to deepen the historical record and add ground-level political-violence events.
How evidence shows up
Interpretation first
PowerFlow is not built like an academic paper with a citation on every sentence. Most surfaces lead with explanation, then expose the evidence behind it.
Receipts where it counts
Relationship pages, conflict evidence, and Ask PowerFlow answers show their work more explicitly, because those are where unsupported claims fail fastest.
Meaning before metrics
When source detail appears, PowerFlow leads with concepts like dependence, concentration, and trend. Raw values are secondary, not the headline.
Limits and judgment
Coverage is uneven. Some actors, relationships, and conflicts are denser than others.
Slow structural data and fast event reporting do different jobs and are not treated as interchangeable.
Trade centrality is position, not alliance. Military weight is structural reach, not readiness or battlefield outcome.
Structural data strengthens a judgment without making it. PowerFlow translates evidence into readable concepts instead of dumping raw tables.
Scores are calibrated outputs, not facts handed down by a single feed. They are built to be defensible, not magical.
Data pipeline
The model combines two layers: a baseline built from structured geopolitical knowledge and structural data, and a continuous stream of sourced reporting that updates scores as events unfold. Every score movement traces to either a sourced event or a calibration pass against the full registry.
Screening
Incoming reports are scored for relevance and geopolitical signal strength. Low-signal content is filtered before it enters the system.
Extraction
Relevant reports are parsed into structured intelligence: affected actors, mechanisms of power shift, trajectory signals, and source attribution.
Relationship scoring
All collective actor pairs in the report are scored bilaterally. Alignment, leverage, and dependency update in both directions.
Actor scoring
Affected actors are rescored against the whole-registry baseline and their relationship context. The structural floor sets the durable position, and a bounded delta reflects the period.
Cascade resolution
When a patron or key dependency shifts, the cascade propagates. A proxy whose patron has been degraded sees its Reach compressed accordingly.
See it live
Every score on this page is visible on the actors themselves
The two pillars, their drivers, and the relationship graph all show up on each actor page, with the evidence that moved them.
Browse the actor directory →