Scenario Planning (2x2)

Quick facts

Mechanism: Produces a scenario set, 2-4 contrasting and internally consistent short narratives of alternative external futures named by two critical-uncertainty axes, plus a robustness read of the strategy across the worlds (which moves survive every world, which early signals show which world is arriving, which options to keep open).

Produces: scenario-set

Evidence: P · Domain: Strategy and Opportunity

Use when the planning horizon is long and the decisive forces are outside control and not reliably predictable, and a single implicit forecast is driving the strategy. Not forecasting and not a single preferred path.

Most strategy quietly rests on a single implicit forecast of the future, and then optimizes for that one future. Scenario planning refuses that bet. It constructs a small SET of alternative states of the external world the planner does not control - regulation, technology adoption, demand, geopolitics - organized by the two axes of uncertainty that most change the strategic choice, and then judges the strategy against the whole set instead of against any one prediction. The durable move is not drawing the grid. It is holding several divergent futures in parallel and asking which moves survive all of them. The output is a scenario set: 2-4 contrasting, internally consistent short narratives of alternative external futures, plus a robustness read of the strategy across them. It is explicitly not a prediction and not a single preferred path. The dominant packaging is the 2x2, because two high-impact and high-uncertainty axes cross into four contrasting worlds - enough variety to break single-future thinking without overwhelming a group.

When to Use

• The planning horizon is long, and the forces that most affect the choice are outside the planner’s control and not reliably predictable (regulation, technology adoption, demand, competitive structure, macro and geopolitics).
• A strategy is quietly riding on a single implicit forecast, and it would be worth knowing which moves hold up if that forecast is wrong.
• A high-stakes, hard-to-reverse bet needs to be stress-tested against more than one plausible future before it is committed.
• The team needs a shared, legitimate way to talk about futures that contradict the house view.

When NOT to Use

• Do not use it as forecasting. The quadrants are structured speculation, not probabilities. Assigning likelihoods to the worlds, or acting on “the most likely quadrant,” reintroduces exactly the single-future thinking the method exists to break. This is the most common and most damaging misuse, and even sophisticated users slip here.
• Do not use it to validate or path to one desired endpoint. That is think-backcasting (fix one desired future, derive the path back). Scenario planning refuses to pick a single future and derives no path.
• Do not use it to trace the ripples of one decision being made. That is think-futures-wheel (one consequence map radiating outward from one change), not a set of alternative external worlds.
• Do not use it to imagine one specified failure. That is think-premortem (assume one plan failed, reason back to causes). Scenario planning is multi-future and not failure-anchored.
• Do not reduce a rich force field to two axes for neatness. Forcing two orthogonal axes can discard the very interactions that matter and produce tidy quadrants with little content (Ramirez and Wilkinson, 2014). If the two best candidates are not independent, say so and reselect.
• Do not produce four mild variations of the present. Four near-identical worlds, or one obvious utopia and dystopia pair, give the comfort of “having done scenarios” with none of the cognitive benefit.
• Do not run it as ritual with no strategy tested against the worlds. Narratives that no strategy is stress-tested against are theater. The payoff is the robustness read and the signal watch-list, not the stories.

Instructions

When asked to build scenarios or stress-test a strategy against an uncertain future, follow these steps:

1. Frame the focal decision and horizon. State, in one line, the strategic choice under pressure and the time horizon over which the external future matters. The scenario set exists to serve this decision; with no decision, stop (that is the ritual anti-pattern).
2. Scan the driving forces. List the forces shaping the domain the planner does not control - regulation, technology adoption, demand, competitive structure, macro and geopolitics, social. Span categories; do not stop at the obvious one.
3. Sort by impact and by uncertainty. Rate each force on how much it would change the strategic choice (impact) and how unpredictable it is (uncertainty). The target is the high-impact AND high-uncertainty corner: the critical uncertainties. Predictable forces are noted as predetermined elements, not axes.
4. Select two critical-uncertainty axes. Pick the two forces that are both high-impact and high-uncertainty and are genuinely independent of each other. Each axis is a spectrum with two contrasting poles. Resist collapsing a rich field to two axes for neatness; if the two best candidates are not independent, say so and reselect.
5. Cross the axes into the 2x2 and name the four worlds. Cross the two axes; each of the four quadrants is the seed of one plausible external future. Give each quadrant a short, evocative name drawn from its pole combination.
6. Construct each world as a divergent, internally consistent narrative. For each quadrant, write a short narrative of that future - how the two poles and the predetermined elements play out together. The worlds must genuinely contrast (not four mild variations of the present) and each must hang together internally. (2-4 worlds; the 2x2 yields four, a smaller set may be defensible when two quadrants collapse into implausibility.)
7. Test the strategy for robustness across all worlds. Run the focal decision and candidate moves against each world. Identify the robust moves that survive every world; the moves that win in one world but lose in another (the bets); and the gaps no current move covers.
8. Name the early signals and the options to keep open. For each world, name the leading indicators that would tell the planner that world is arriving (the watch-list). Name the options worth keeping open precisely because the worlds diverge.
9. Emit the scenario set artifact per references/TEMPLATE.md: the two named axes, the named worlds with their narratives, and the robustness read (robust moves, bets, signal watch-list, options to keep open). Frame the worlds as structured speculation, never as ranked probabilities.

Output Format

Use the template in references/TEMPLATE.md. The deliverable is the filled scenario set - two named axes, the 2x2 of named worlds with short narratives, and the robustness read (robust moves, bets, signal watch-list, options to keep open) - not a prose essay. Never rank the worlds by likelihood.

Quality Checklist

Before finalizing, verify:

• The focal decision and horizon are stated in one line, and the scenario set serves that decision.
• The two axes are both high-impact and high-uncertainty, genuinely independent, and each is a spectrum with two named poles - not a rich field collapsed to two for neatness.
• The four worlds genuinely contrast (not four mild variations of the present) and each is internally consistent.
• A strategy is actually tested against every world: robust moves, one-world bets, and uncovered gaps are all named.
• Each world has a signal watch-list, and the options worth keeping open are named.
• The worlds are framed as structured speculation, never ranked by likelihood or treated as probabilities.
• The output is the scenario set artifact, not prose.
• No overclaiming: the evidence is practitioner-grade and transferred; claim a divergence-and-robustness aid, not a predictor or a measured gain in decision outcomes (see evidence/dossier.md).

Evidence

Tier P (governing; honest read M-down-to-P). Scenario planning is a genuinely established, half-century-old practitioner method (Wack 1985; Schwartz 1991; Schoemaker 1995), with a coherent rationale - counter single-future anchoring and test for robustness. There is one reasonably supportive controlled study (Meissner and Wulf, 2013) finding reduced framing bias on 252 management students, but the field’s most-cited author calls the usefulness evidence “anecdotal” (Schoemaker, 2004), the strongest real-expert study finds scenarios shift judgment toward whichever scenario is shown rather than uniformly improving it (Phadnis et al., 2015), some judgmental-forecasting work finds scenarios can worsen accuracy, and the 2x2 itself is critiqued as an oversimplified off-the-shelf tool (Ramirez and Wilkinson, 2014). Per this library’s conservative rule the governing grade is the lower half, P. All evidence is transferred from human subjects in workshop, lab, and field settings; none studies an AI-produced scenario set, which independently caps the grade at P. The skill ships as a divergence-and-robustness aid with a hard “this is not forecasting” wall, never as a predictor. Full grading, sources, and caveats: evidence/dossier.md.

Examples

See references/EXAMPLE.md for a completed scenario set on a real decision.

Deep dive: worked example

A full worked run (the shared Northwind scenario)

Scenario Set (2x2) - Worked Example

A completed run of the scenario-planning skill on a real, consequential decision. This is the quality bar a generated scenario set should meet.

Uses the shared recurring scenario (Northwind, a B2B SaaS weighing a self-serve free-tier launch) so examples across skills read as one coherent product. Where think-futures-wheel traces the ripples of launching the free tier, and think-backcasting maps a path back from one chosen future, this skill builds several uncontrollable external futures and asks which free-tier moves survive all of them. See docs/internal/AUTHORING.md.

The four worlds below are structured speculation, not probabilities. They are not ranked by likelihood. The value is the robustness read at the end, not the stories.

---

Focal decision and horizon

• Focal decision: Commit to a self-serve free tier as Northwind’s primary growth motion over the next 3-4 years (the platform bet), versus staying sales-led.
• Horizon: 3-4 years - long enough that the external environment, not the launch mechanics, decides whether the bet pays off.
• Uncontrollable domain: how B2B software is bought and built around Northwind - the structure of its buyers and the pace of AI-native disruption. Northwind controls its product and pricing; it does not control either of these.

Driving forces and predetermined elements

• Predetermined elements (high impact, low uncertainty): continued shift to cloud and remote work; ongoing data-privacy regulation tightening; buyers’ growing expectation of “try before you buy.” These shape every world and are not in contention.
• Critical uncertainties (high impact, high uncertainty): (1) whether Northwind’s buyers stay fragmented (many independent teams choosing their own tools) or consolidate behind a few platform vendors and procurement gatekeepers; (2) whether AI-native entrants disrupt the category incrementally (AI as a feature on top of today’s workflows) or with a step-change (AI agents that collapse the workflow Northwind sells). Two macro forces (regulation tightening, demand growth) were high impact but more predictable, so they are predetermined elements, not axes. A third candidate (funding climate) was set aside as not independent of the AI-disruption axis.

The two axes

• Axis A: Buyer structure - spectrum from fragmented (bottom-up, many independent team buyers; the self-serve sweet spot) to consolidated (top-down platform buyers and procurement gatekeepers).
• Axis B: AI-native disruption pace - spectrum from incremental (AI augments the existing workflow Northwind owns) to step-change (AI-native agents collapse or replace that workflow).

These two are genuinely independent: how buying is organized does not determine how fast the technology frontier moves, and vice versa.

The 2x2 - four named worlds

	Axis B = incremental AI	Axis B = step-change AI
Axis A = fragmented buyers	World 1: Land Rush	World 2: Agent Swarm
Axis A = consolidated buyers	World 3: Platform Gatekeepers	World 4: Suite Eats the Workflow

• World 1: Land Rush (fragmented x incremental) - The friendly world. Teams keep choosing their own tools bottom-up, and AI stays an enhancement to the workflow Northwind already owns. Self-serve free tiers are the dominant motion; the winner is whoever lands the most individual users and converts them. Northwind’s free-tier bet is directly in its element, and the fight is on activation, breadth of integrations, and conversion craft.

• World 2: Agent Swarm (fragmented x step-change) - Buyers are still fragmented and self-serve, but AI-native agents now do much of the job Northwind’s workflow used to do. Individual users still sign up bottom-up, but they increasingly arrive expecting an agent to do the work, not a UI to do it in. A free tier still acquires users cheaply, but the product behind it has to become AI-native fast or the funnel fills with users who churn to an agent-first competitor. Distribution is intact; the product thesis is under attack.

• World 3: Platform Gatekeepers (consolidated x incremental) - The technology frontier is calm, but buying has moved up to procurement, platform standards, and a few preferred-vendor suites. Bottom-up free signups still happen but rarely convert, because the real purchase decision sits with a gatekeeper who wants SSO, governance, security review, and a master agreement. A free tier here is a top-of-funnel and advocacy tool, not a closing motion; the deal is won (or lost) on enterprise readiness and platform fit.

• World 4: Suite Eats the Workflow (consolidated x step-change) - The hardest world. Buying has consolidated AND a step-change in AI lets the big platform suites absorb Northwind’s workflow as a native AI capability inside a bundle the gatekeeper already pays for. Standalone self-serve has the weakest pull here: individual enthusiasm does not move a procurement decision, and the incumbent suite is “good enough and already approved.” Survival depends on a defensible wedge the suite cannot quickly copy and on being where the gatekeepers already are.

Robustness read

How Northwind’s free-tier and platform moves fare across the four worlds.

Move	World 1 Land Rush	World 2 Agent Swarm	World 3 Gatekeepers	World 4 Suite	Verdict
Ship a free tier generous on individual-value features	wins	survives (acquisition still cheap)	survives (funnel + advocacy, not closing)	weak (does not move procurement)	robust-ish - useful in 3 of 4, never actively harmful
Bet the whole company on self-serve as the primary closing motion	wins	risky	loses	loses	bet - wins only in Land Rush
Invest early in an AI-native rebuild of the core workflow	neutral cost	wins	neutral	partially wins	robust under any step-change risk
Build enterprise readiness (SSO, governance, security, admin)	low value	low value	wins	necessary-to-survive	bet that pays in the consolidated column
Keep the architecture modular / embeddable (be where buyers already are)	helps	helps	helps	wins	robust - the one move that survives every world

• Robust moves (survive every world): keep the product modular and embeddable so Northwind can be adopted bottom-up, integrated into a platform, or surfaced inside an AI agent; and ship the free tier generous on individual-value features (it helps in three worlds and never hurts). These are committed regardless of which future arrives.
• Bets (win in one world, lose in another): betting the company on self-serve as the primary closing motion is a bet on Land Rush (World 1); going heavy on enterprise readiness is a bet on the consolidated column (Worlds 3-4). Make each consciously, not by default.
• Gaps (no current move covers): none of the moves above gives Northwind a defensible wedge in World 4 (consolidated + step-change) if the suite copies the workflow as a native AI feature. That gap is the strategic exposure to watch and to fund an option against.

Signal watch-list (which world is arriving)

World	Leading indicators that this world is arriving
World 1: Land Rush	Self-serve conversion stays healthy; competitors compete on activation and integrations; procurement rarely appears in deals
World 2: Agent Swarm	Free signups stay strong but retention drops as users leave for agent-first tools; “can it just do it for me?” dominates feedback
World 3: Platform Gatekeepers	Win rate increasingly gated on SSO / security review / master agreements; more deals routed through procurement and preferred-vendor lists
World 4: Suite Eats the Workflow	The major suites ship the workflow as a bundled native AI feature; standalone evaluations shorten or vanish; “we already have it in [suite]” objections rise

Options to keep open

Because the worlds diverge, three options are worth holding even though committing to any one now would be premature: (1) an AI-native core rebuild kept ready to accelerate the moment step-change signals appear (Worlds 2 and 4); (2) an enterprise / platform-fit track (SSO, governance, embeddability) staged so it can be pulled forward if buying consolidates (Worlds 3 and 4); and (3) a partner / embed path to sit inside the very suites that could otherwise eat the workflow, as a hedge against World 4. Keeping these open is the payoff of having modeled four futures instead of betting on one.

---

Note how this differs from its neighbors on the same Northwind decision. The think-futures-wheel example traces the consequences of one move Northwind makes (launching the free tier ripples into support load, MRR, comp). The think-backcasting example fixes one desired future and maps the path back to today. This scenario set does neither: it constructs four alternative external worlds Northwind does not control, holds them in parallel without ranking them, and asks which free-tier moves survive all of them. The deliverable is robustness (and a watch-list), not a consequence map and not a route.

Grounding: the full evidence dossier

What the research does and does not show, with graded sources

Evidence Dossier: Scenario Planning (2x2)

The single source of truth for the scenario-planning skill. The SKILL.md, the sidecar (skill.meta.yml), and the eval cases all derive from this file. If a claim is not here, it does not belong in the skill. Promoted from frameworks/_proposed/scenario-planning/dossier.md and admitted as a Build at tier P (correcting the candidate’s stale M).

Skill	thinking-framework-skills.scenario-planning (installable name think-scenario-planning)
Family	strategy-and-opportunity
Evidence tier	P governing (honest read M-down-to-P, capped at P - see “What the evidence shows”)
Confidence	Moderate that multi-future construction breaks single-forecast anchoring and surfaces robust moves; low that any specific decision-outcome effect transfers to agents
Status	draft (admitted from the v0.5.0 catalog tranche; tier corrected M -> P on the field-expert and review evidence)

---

1. The mechanism (what actually does the work)

Scenario planning, in its dominant “intuitive logics” / 2x2-matrix form, builds a small SET of internally consistent, deliberately divergent stories about the external future an organization will have to operate in but cannot control. The procedure: scan the driving forces shaping the domain; sort them by impact and by uncertainty; pick the two forces that are both high-impact and high-uncertainty (the “critical uncertainties”); cross them as the axes of a 2x2 grid; and treat each of the four quadrants as the seed of one plausible future, fleshed out into a short narrative. Strategy is then stress-tested against all four worlds, and the planner looks for moves that are robust (survive every quadrant), the early signals that would tell you which world is arriving, and the options worth keeping open.

The durable cognitive move is not the grid drawing. It is constructing multiple alternative states of an uncontrollable environment at once, organized by the two axes of uncertainty that most change the strategic choice, and then judging strategy against the whole set rather than against a single forecast. Two things distinguish it from ordinary planning: the object is the external world the planner does not control (not the planner’s own plan or its consequences), and the output is a set of divergent futures held in parallel (not one prediction and not one preferred endpoint). The 2x2 is the most common packaging because two axes yield four contrasting worlds, which is enough variety to break single-future thinking without overwhelming a group.

The output is a scenario set: 2-4 contrasting, internally consistent short narratives of alternative external futures, named by the two critical-uncertainty axes, plus a robustness read of the strategy across them - which moves survive every world, which early signals indicate which world is arriving, and which options to keep open. The point is not the stories; it is the robustness read and the signal watch-list the stories make possible.

2. Lineage

Scenario planning’s modern lineage runs through the RAND / Hudson Institute work of Herman Kahn in the 1950s-60s and, decisively, through Royal Dutch / Shell, where Pierre Wack and Edward Newland built the corporate scenario method beginning in the late 1960s and early 1970s; Wack’s two 1985 Harvard Business Review articles, “Scenarios: Uncharted Waters Ahead” and “Scenarios: Shooting the Rapids,” are the most-quoted papers in the field. Peter Schwartz, who led Shell’s scenario team in the 1980s, founded the Global Business Network (GBN) in 1987 and popularized the practical method - including the 2x2 critical-uncertainties matrix - in his 1991 book The Art of the Long View. Paul Schoemaker’s 1995 Sloan Management Review article gives the canonical step-by-step build. Shell’s preparedness for the 1973 oil shock is the founding anecdote.

The terms “scenario planning,” “scenario analysis,” and “2x2 scenario matrix” are generic and descriptive; the durable move is named for what it does (multi-future construction and robustness testing), and the skill ships documented descriptively with the lineage credited here rather than branded. The attribution string credits Pierre Wack / Royal Dutch Shell.

3. What the evidence shows, and what it does NOT show

The honest grade is P (practitioner), and this dossier deliberately overturns the candidate M tag, which was too generous. The reason is that the controlled evidence is thin, contested, and partly cuts against the method’s own marketing.

What the record supports. Scenario planning is a genuinely established, half-century-old practitioner method, taught widely and used across business and government, with a coherent rationale (counter single-future anchoring; test for robustness). There is one reasonably supportive controlled study: Meissner and Wulf (2013), an experiment with 252 graduate management students, found that scenario planning “reduces the framing bias” and “has a more positive effect on decision quality than tools traditionally used in strategic planning.” That is real, nameable support for a cognitive (de-biasing) benefit - but the subjects were students, and the authors themselves frame the result cautiously, concluding only that such tools “may in fact alter biases and decision quality.”

What the record does NOT support. The field’s own most-cited author, Paul Schoemaker, calls the evidence of scenario planning’s usefulness “anecdotal” (2004). The strongest empirical examination on real experts - Phadnis, Caplice, Sheffi and Singh (2015), three field experiments with transportation-infrastructure professionals - reports that their “extensive literature review unearthed only three experimental studies - all conducted with student subjects,” that two of those (Kuhn and Sniezek 1996; Schoemaker 1993) “tested the effect of scenarios on subjects’ confidence and reached contrary conclusions,” and that “none of these studies definitively answer whether the use of scenarios affects managerial judgment in the ill-defined long-range planning problems faced in real-world situations.” Phadnis et al.’s own finding is cautionary rather than confirmatory: scenarios did not uniformly raise or lower experts’ confidence; instead “expert judgment changes in accordance with how an investment fares in a given scenario” - i.e. the scenario you show shifts the judgment toward itself - and experts ended up preferring “more flexible options.” That is a behavioral / framing effect, not a demonstrated improvement in decision quality or accuracy. A separate strand of judgmental-forecasting research has even found that providing scenarios can worsen forecast accuracy under some conditions. The 2023 review of reviews (Cordova-Pozo and Rouwette) concludes the field “suffers from several methodological shortcomings,” with no accepted definition and many divergent approaches; the realist-synthesis literature (Wright, Goodwin and Cairns) likewise finds academic evidence on the mechanisms and outcomes of scenario planning “notably lacking, despite a substantive practitioner evidence base.”

A related design caution: Ramirez and Wilkinson (2014) show the 2x2 has hardened into a “somewhat simplistic, off-the-shelf tool” in which planners “compile a list of uncertain factors, from which only the two most important are selected” - a step that can discard the very interactions that matter and produce tidy quadrants with little content. This is why the procedure insists the two axes be genuinely independent and the worlds genuinely divergent.

Netting it out: the honest split is “M on a single student-subject de-biasing experiment / P-or-weaker once you weight the field-expert evidence, the contrary studies, and the most-cited scholar’s ‘anecdotal’ verdict.” Per this library’s conservative rule, the governing grade is the lower half: P, not M. There is no robust, replicated S/M body on the actual move (multi-future construction improving real decisions) to launder upward; the one M-leaning study is outweighed by contrary and null findings on the population that matters.

4. Transferred-evidence flag (required honesty for this library)

Every study above is on human subjects - students or human experts - in workshop, lab, and field settings. None studies a scenario set produced by or with an AI agent, nor whether an agent-produced 2x2 improves a human’s decision. The evidence is transferred from human contexts and not validated for AI-augmented use, which independently caps the grade at P. The AI value is mechanical and modest: an agent makes the method cheap to run, forces the discipline (a real driving-force scan, genuinely uncertain and independent axes, divergent and internally consistent worlds, an explicit robustness test), and produces a durable, inspectable artifact - benefits that do not depend on any contested outcome claim. The skill ships honestly as a P-tier divergence-and-robustness aid with a hard “this is not forecasting” wall, never as a predictor.

5. When it works / when it fails (drives the eval negative cases and “When NOT to Use”)

Works best when:

• The planning horizon is long, the environment is genuinely turbulent, and the main forces (regulation, technology adoption, demand, geopolitics) are outside the planner’s control and not reliably predictable.
• A strategy is quietly riding on a single implicit forecast, and it is worth knowing which moves hold up if that forecast is wrong.
• A high-stakes, hard-to-reverse bet needs stress-testing against more than one plausible future before commitment.

Fails or misleads when (poor-fit / anti-patterns):

• It is treated as forecasting. The four narratives are structured speculation, not probabilities; assigning likelihoods or acting on “the most likely quadrant” reintroduces exactly the single-future thinking the method exists to break. Even sophisticated users slip here. This is the central wall.
• It is used to validate or path to one desired endpoint. That is backcasting (fix one desired future, derive the path back), not multi-future construction; scenario planning refuses to pick a single future and derives no path.
• It is used to trace one decision’s ripples. That is the futures-wheel (one consequence map radiating outward from one change), not a set of alternative external worlds.
• It is used to imagine one specified failure. That is the premortem (assume one plan failed, reason to causes); scenario planning is multi-future and not failure-anchored.
• The two axes are chosen badly or for neatness. Reducing a rich field of forces to two orthogonal axes can force a false structure (Ramirez and Wilkinson, 2014) and discard the interactions that matter.
• The scenarios are not divergent or not plausible. Four mild variations on the present, or one obvious utopia / dystopia pair, give the comfort of “done scenarios” with none of the benefit.
• It is run as ritual with no link to a decision. Producing narratives that no strategy is ever tested against is theater; the payoff is the robustness read and the watch-list of signals, not the stories.

6. Output artifact

The skill must emit a scenario set, not prose: the focal decision and horizon; the two named critical-uncertainty axes, each with its two contrasting poles; the 2x2 of named worlds (2-4), each a short, divergent, internally consistent narrative; and the robustness read - the robust moves that survive every world, the one-world bets, the gaps no move covers, the per-world signal watch-list, and the options worth keeping open. The worlds are framed as structured speculation, never ranked by likelihood. A short summary may sit above the set.

7. Sources

1. Pierre Wack, “Scenarios: Uncharted Waters Ahead,” Harvard Business Review 63(5):73-89 (1985), and “Scenarios: Shooting the Rapids,” Harvard Business Review 63(6):139-150 (1985). The foundational Shell account; most-cited in the field. Practitioner / foundational.
2. Peter Schwartz, The Art of the Long View (1991). Popularized the intuitive-logics / 2x2 critical-uncertainties method via GBN. Practitioner / foundational.
3. Paul J. H. Schoemaker, “Scenario Planning: A Tool for Strategic Thinking,” Sloan Management Review 36(2):25-40 (1995). The canonical step-by-step build. Practitioner reference. (P) (See also Schoemaker’s later “anecdotal”-evidence admission, 2004.)
4. Philip Meissner and Torsten Wulf, “Cognitive benefits of scenario planning: Its impact on biases and decision quality,” Technological Forecasting and Social Change 80(4):801-814 (2013). Experiment, 252 graduate management students; found reduced framing bias and a more positive effect on decision quality than traditional planning tools, with the authors’ own cautious hedge. The single most supportive controlled study; student subjects. (M-leaning experiment, student population.)
5. Shardul Phadnis, Chris Caplice, Yossi Sheffi and Mahender Singh, “Effect of scenario planning on field experts’ judgment of long-range investment decisions,” Strategic Management Journal 36(9):1401-1411 (2015). Three field experiments with real infrastructure experts; documents that prior empirical evidence was limited to three student-subject studies with contrary results, and finds scenarios shift judgment toward the scenario shown and toward flexible options rather than uniformly improving it. The strongest real-expert evidence, and cautionary. (Field experiments.)
6. Rafael Ramirez and Angela Wilkinson, “Rethinking the 2x2 scenario method: Grid or frames?,” Technological Forecasting and Social Change 86:254-264 (2014). Critique of how the 2x2 matrix has become an oversimplified off-the-shelf tool and what reducing to two axes loses. (Critical literature.)
7. Kathya Lorena Cordova-Pozo and Etienne A. J. A. Rouwette, “Types of scenario planning and their effectiveness: A review of reviews,” Futures 149:103153 (2023). Finds convergence on a definition but persistent methodological shortcomings and many divergent approaches; bounds claims of established effectiveness. (Review of reviews.)
8. George Wright, Paul Goodwin and George Cairns (and colleagues), realist-synthesis and critical work on scenario planning (e.g. “Opening the ‘black box’ of scenario planning through realist synthesis,” Technological Forecasting and Social Change, 2020). Finds academic evidence on mechanisms and outcomes “notably lacking” despite a substantive practitioner base. (Critical literature.)

Excluded on the evidence rule: no specific decision-quality or forecast-accuracy effect size for scenario planning is asserted as fact in this dossier, because no robust, replicated, primary-sourced figure exists on the actual move. The one nameable controlled effect (Meissner and Wulf’s de-biasing result) is reported with its student-subject limitation, and the field-expert and review evidence are weighted against it to set the conservative governing grade of P.