The methods inside the Policy Wind-Tunnel

Foundations

Anticipation, not prediction

Planning by extrapolating the past is like driving while looking only in the rear-view mirror. It works in calm periods and fails when change is fast, complex and non-linear.

Strategic foresight looks beyond prevailing expectations, surfaces the assumptions hidden in institutions and engages with low-probability, high-impact discontinuities before they become crises. The European Commission and the OECD use the futures cone to separate four classes of futures: possible (anything conceivable), plausible (consistent with what we know about how systems work), probable (what trends suggest) and preferable (what a society chooses to build).

Two ideas are often confused. Resilience is the capacity to absorb a shock and recover. Anticipatory governance is the capacity to detect weak signals, bring diverse intelligence together and turn forward-looking analysis into legitimate action before the shock. It runs through policymaking, budgeting, regulation, innovation and evaluation, rather than sitting in a report at the start of a planning cycle.

Policy Horizons Canada

The Horizons Foresight Method

A systems-based, iterative process that asks experts to surface their mental models of how a system works and then tests policies against structurally different futures.

FramingSet the scope, the time horizon and the policy question.
AssumptionsList the commonly held planning assumptions behind current programmes.
ScanningLook past the trends everyone already tracks, towards weak signals and low-probability, high-impact events.
System mappingUse system maps and impact trees to trace first-, second- and third-order effects across domains.
Change driversIdentify the forces that could reshape the system.
ScenariosBuild structural scenarios that go beyond the classic two-axis matrix, so wicked problems are not squeezed into four boxes.
Results“Wind-tunnel” existing and proposed policies, then loop back to expose vulnerable assumptions and replace them.

The final step feeds back into the second: the scenarios are tools to challenge baseline beliefs, not forecasts.

Used in Module 3: wind-tunnel and assumption premortem

Singapore · Centre for Strategic Futures

Scenario Planning Plus (SP+)

Singapore adopted Shell-style scenario planning in the early 1990s. The 1997 Asian financial crisis, the 2001 attacks and SARS in 2003 showed that standard scenarios miss non-linear shocks, so the method was upgraded.

Stage	Objective	Instruments	Output for policymakers
1. Defining focus	Classify the nature of the problem to pick the right tools	Cynefin framework	No methodological mismatch, e.g. no linear models in complex domains
2. Environmental scanning	Look for shifts in velocity and emerging discontinuities	Emerging Issues Analysis; the RAHS scanning platform	A shortlist of emerging strategic issues, plus tracked blind spots
3. Sense making	Turn fragments into a coherent systemic picture	Driving Forces cards; SWOT	A map of how trigger events alter trends and affect groups
4. Developing possible futures	Challenge groupthink with plausible and normative futures	20-year national scenarios; backcasting	Narratives and reverse-engineered policy pathways
5. Designing strategies	Shape interventions that survive turbulence	War-gaming; whole-of-government integrated risk management	Stress-tested options and cross-ministry protocols
6. Monitoring	Track leading indicators and strategy performance	Early-warning systems; signpost matrices	Triggers that alert leaders when a scenario is materialising

One feature is easy to copy and rarely copied: in voting exercises, low-consensus issues are kept and tracked rather than dropped, because consensus filters blind organisations to the most disruptive threats.

Used in Modules 1 and 4: scanning, signposts and Cynefin triage

Weak-signal triage

The Horizon Loop

An eight-step discipline that stops horizon scanning from becoming an archive of interesting news that produces no decisions.

AnchorBound scanning by a specific institutional dependency and the assumption behind it, not by a broad topic.
CollectOn a fixed schedule, quality over volume: five to ten strong items per category beat a hundred weak ones.
LogA two-minute record: event date, STEEP category, challenged assumption and the critical “if this scaled” mechanism.
ScreenSix filters: novelty, source credibility, centrality of the challenged belief, scalability, impact and timeframe.
ClusterAt least three independent signals sharing one causal mechanism. Three articles about the same event are one signal. A theme must pass four tests: something changed, it can reach a dependency, exposure is unmeasured, it could alter a decision.
InterpretAsk what drives the theme and whether the trajectory will persist.
Hand overGive the theme to a named decision-maker with a signpost watch-list.
HygieneRetire any signal older than two years that never joined a cluster.

Used in Module 1: signal triage

Weimer-Jehle · ScenarioWizard · Cascade Institute

Cross-Impact Balance analysis

A way to move from qualitative themes to scenarios that hold together, by letting the logic of the system rather than the loudest voice decide which combinations are plausible.

The system is described by a set of descriptors, each with a few discrete future states. Experts judge, for every pair, whether a state of one descriptor promotes or inhibits each state of another. A scenario is internally consistent when, for every descriptor, the state it contains receives the strongest combined support from the other states in the same scenario.

The wind-tunnel's solver tests every combination in your browser. For each one it adds up the impacts each state receives, measures the gap between the chosen state and the best-supported alternative (the inconsistency), and keeps only the combinations within your tolerance. It ranks them by total impact: how strongly the scenario reinforces itself.

CIB was formalised by Wolfgang Weimer-Jehle and is implemented in the ScenarioWizard software; the Cascade Institute uses it to model polycrises. Eliminating contradictory stories before the wind-tunnel, robust decision-making and premortems is what makes the later steps credible.

Used in Module 2: scenario generator

Matching method to context

Cynefin and the four contextual variables

Over-engineered models miss political windows; superficial brainstorming lacks evidential standing. Four variables decide how a foresight exercise should be built.

1. Complexity domain

Using Dave Snowden's Cynefin framework: clear (best practice), complicated (expert modelling), complex (structural scenarios, CIB and safe-to-fail sandboxes), chaotic (stabilise and war-game) or disorder.

2. Time horizon

6 to 24 months for sandbox cohorts and signal hygiene; 5 to 10 years for technology radars and crisis-management transitions; 15 to 40 years for welfare, defence and net-zero choices.

3. Polity structure and strategic depth

A city-state with little room for error can centralise scanning in the centre of government. Multilevel polities such as the EU must connect standard-setting with cities and regions, which feel shocks first.

4. Authorising environment and trust

Trust varies by topic. Where importance is high but trust is low, as with health research or AI, foresight needs transparent deliberation, not technocratic authority alone.

Used in Module 4: Cynefin diagnostic

Nesta · EU AI Act · sandboxes

Anticipatory regulation

Fast-scaling technologies create a pacing problem: regulate too early and you stifle useful innovation, too late and harm becomes irreversible. Anticipatory regulation treats rules as an iterative, learning tool.

Collaborative

Co-develop the regulatory environment with newcomers, and coordinate across jurisdictions.

Lead-taking

Steer markets towards public-interest outcomes, for example with challenge prizes such as the UK's Open Up 2020 Challenge for open banking.

Systems-first

Look across legacy statutory silos, as Accelerate Estonia's cross-sectoral experimentation framework does.

Confidence-building

Generate safety evidence and ethical frameworks in public, as in the UK MHRA AI Airlock for AI as a medical device.

Goal-oriented

Use outcome-based pull incentives that signal a direction of travel instead of prescriptive rules.

Learning-centric

Use sandboxes and testbeds, often with temporary derogations of around six months, under direct supervision.

Under the EU AI Act, regulatory sandboxes do two jobs: they lower compliance costs and speed market entry for SMEs, and they feed real-world evidence back to regulators so guidance and standards can adapt. Nesta also recommends small, rapid-allocation contingency budgets governed by cross-agency panels, so regulators can act as soon as scanning detects a threat or opportunity.

Used in Module 4: regulation configurator

Standard criteria

Four evaluation matrices

Inputs and outputs are judged against criteria drawn from established governance frameworks.

Dimension	Source	Criteria	Function
A. Weak signals and emerging risk	Horizon Loop	Six signal filters; four emerging-risk tests	Turns noise into actionable clusters and retires unclustered signals after 24 months
B. Sectoral policy sustainability	WRR Report 104	Financial (GDP share, per-capita cost, budget ceilings), personnel (workforce ratio, carer burden, psychosocial load), societal (solidarity, trust, acceptance of hard choices)	Checks that a policy stays in balance without crowding out other public values
C. Intergenerational well-being	Well-being of Future Generations (Wales) Act, Integrated Impact Assessment	Economic, social, environmental and cultural well-being, weighed equally	Stops short-term policies from pushing costs onto future generations
D. Institutional anticipatory capacity	Integrated Anticipatory Governance Framework	Strategic, institutional, democratic and adaptive capacity	Audits whether an administration can turn futures intelligence into legitimate action

B and C in Module 3 D in the readiness check

The methods inside the wind-tunnel