How We Build the Map
ScenaPlan doesn't predict one future. It maps the space of possible futures, assigns probabilities, and updates them weekly as new signals arrive. Here's how.
Core Principle
We track ships, not tweets. Official statements, press conferences, and social media posts are noise. Observable actions — ship movements, troop deployments, financial flows, regulatory changes, logistics — are signal. When words and actions diverge, the actions are the truth. This principle governs every layer of our analysis.
We don't ask "what will happen." We ask: "what are the observable indicators that will tell us which branch of the scenario tree we're on?" Then we watch for those indicators. When they fire, we update probabilities. When they don't, we update too.
Six Frameworks
ScenaPlan is not built on one methodology. It synthesizes six established frameworks from intelligence analysis, military strategy, and systems science. Each addresses a different failure mode of geopolitical analysis.
ACH
Analysis of Competing Hypotheses · CIA / Richards Heuer
List all plausible hypotheses. For each new data point, assess: which hypotheses does it support, which does it contradict? The surviving hypothesis isn't the one with the most evidence for — it's the one with the least evidence against. Prevents narrative lock-in.
Book: Psychology of Intelligence Analysis (Heuer, 1999)
Scenario Planning
Shell Method · Pierre Wack / Peter Schwartz
Don't predict one future — map 3–4 internally coherent but mutually exclusive scenarios. Assign each early indicators that distinguish it. Prepare for all of them. Shell used this to predict the 1973 oil crisis when the CIA missed it.
Book: The Art of the Long View (Schwartz, 1991)
Superforecasting
Bayesian Updating · Philip Tetlock / IARPA
Start with a base rate. Update incrementally with each new signal. Never shift more than the evidence warrants. Track calibration over time — are your 70% predictions right 70% of the time? Public, verifiable track record is the only proof that matters.
Book: Superforecasting (Tetlock & Gardner, 2015)
OODA Loop
Decision Cycle · Col. John Boyd, USAF
Observe › Orient › Decide › Act › repeat. The actor who cycles fastest wins. Our model accelerates your OODA: you see the signal and know what it means before the market reacts. Daily scan, weekly assessment, monthly review — three nested loops.
Briefing: Patterns of Conflict (Boyd, 1986)
Systems Dynamics
Feedback Loops · Donella Meadows / Santa Fe Institute
Map reinforcing loops (escalation spirals) and balancing loops (stabilizers). Identify where the system amplifies small perturbations and where it absorbs them. Phase thresholds — specific numbers where the system jumps to a qualitatively different state.
Book: Thinking in Systems (Meadows, 2008)
Agent Simulation
Multi-Agent AI · LLM-Based Wargaming
Multiple AI agents play as key actors (each with defined objectives, constraints, and information asymmetry). They respond to the same signals from different perspectives. Emergent insights — moves no single analyst would predict — are the most valuable output.
Ref: Santa Fe Institute complexity research
Signal Scoring
Not all information is equal. Every data point is scored on four parameters before it enters the model. Only signals scoring 15+/40 affect probabilities.
| Parameter | Range | What it measures |
|---|---|---|
| Source Reliability | 0–10 | Observable actions (8–10) > official sources (5–7) > media narratives (1–4) |
| Actionability | 0–10 | Does this change a probability on the bifurcation tree? Ship movement = 9. Op-ed = 1. |
| Novelty | 0–10 | First report = 10. Fifth retelling of the same fact = 0. |
| Corroboration | 0–10 | Three independent sources = 10. Single anonymous claim = 2. |
Threshold: 15/40. Below this, the signal is noise — logged but not acted upon. Above this, it enters the model and may shift probabilities. The highest-scoring signals (30+/40) appear in the weekly briefing as "Signals Fired."
Source Hierarchy
1
Observable Actions
Data that cannot be faked. This is where truth lives.
Ship positions (MarineTraffic), military aviation (FlightRadar24), satellite imagery (Sentinel Hub), financial market data (real-time prices), regulatory filings (armypubs.army.mil, defense.gov contracts), OSINT-verified footage.
2
Official & Analytical
Facts with institutional backing. Useful but filtered through agendas.
Wire services (Reuters, AP, AFP), government sources (Pentagon, White House, IAEA), regional media (Fars News, Times of Israel, Haaretz), think tanks (Atlantic Council, CSIS, RAND, Chatham House).
3
Narrative
What each side wants you to think. Valuable for understanding the gap between words and actions — not for facts.
Major media (CNN, NYT, Al Jazeera, RT), social media (Twitter/X OSINT community, Telegram channels), commentary and opinion.
Words vs Actions Detector
The signature tool
Every week, we systematically compare what officials say with what observable data shows. When they diverge — the divergence is the signal. A president announces a "pause" while ships remain in position. A government claims "talks are progressing" while troop deployments accelerate. A country denies involvement while drones fly 1,000 km through its airspace unimpeded. The gap between words and actions reveals intent more reliably than either words or actions alone.
Probability Update Rules
| Condition | Allowed Shift |
|---|---|
| Single signal, medium confidence | ±3–5% |
| Multiple corroborating signals | ±5–10% |
| Bifurcation resolved (event confirmed) | Node › "passed," new BIF activates |
| Phase threshold crossed | Major restructure — flagged for human review |
| Maximum shift per week | ±15% (hard cap without human override) |
These constraints prevent model whiplash — the tendency to overreact to dramatic headlines. The model changes slowly and deliberately. If reality shifts faster, the cap is overridden explicitly, with documented reasoning.
Bifurcation Trees
The core visual tool. A bifurcation tree maps decision points where the future branches. Each node represents a moment where observable events will determine which path the system takes. Each branch has a probability and specific indicators.
Trees are fractal — nested at multiple levels. The top level (L0) shows the global strategic plan. L1 shows individual conflicts (Iran War, Russia containment). L2 shows tactical bifurcations within each conflict. You can zoom in or out depending on what you need.
Every week, the tree is updated: probabilities shift, nodes resolve, new branches emerge. The Mermaid diagram format serves as the source of truth — human-readable, machine-readable, version-controlled.
What We Don't Do
We don't predict. We map possibility spaces with probabilities. "70% ground operation by April 5" is not a prediction — it's a probability estimate with a specific falsification condition and date.
We don't give investment advice. Portfolio Compass is scenario analysis: "in scenario X, asset Y moves in direction Z." The reader makes their own decisions. We provide the map, not the destination.
We don't claim secret sources. Everything is built on open-source intelligence (OSINT). Every fact is sourced. Every source is linked. You can verify anything we publish.
We don't hide our mistakes. When a prediction is wrong, we say so publicly, explain why the model failed, and update. The track record is the product — hiding misses would destroy it.
Our commitment to you
Every probability has a cause. If it moved, we say why. No unexplained shifts.
Every claim has a source. Publication, date, link. No "sources say."
Every prediction has a date. Falsifiable. Verifiable. If we're wrong — the record shows it.
Every issue challenges itself. The Skeptic is not optional. Confirmation bias is the enemy.
Observable actions over statements. Always. Ships over tweets. Data over drama.
Intellectual Lineage
Structure: Heuer (ACH), Schwartz & Wack (scenario planning), Tetlock (superforecasting), Boyd (OODA). Systems: Meadows (feedback loops), Wallerstein (world-systems), Santa Fe Institute (complexity & emergence). Strategy: Friedman (geopolitical realism), Mearsheimer (offensive realism). Resilience: Taleb (antifragility, black swans).
ScenaPlan is not derivative of any single thinker. It is a synthesis designed for a specific purpose: navigating geopolitical uncertainty with observable data, structured methodology, and explicit probability tracking. Where these frameworks disagree — we note the disagreement rather than force consensus.