Numbers persuade only after a story makes the world legible. Scenario narratives turn architecture into coherent worlds that executives can remember, Finance can quantify, and Risk can instrument. The narrative is not fiction; it is a disciplined explanation of why this world holds together—who does what, under which constraints—and what it means for us. The evidence pack then shows your work: sources, assumptions, and traceability so challenge strengthens, not derails, the decision.
10.1 Narrative Structure, Names/Hooks/Visuals & One‑Page Brief—Template
A scenario narrative is not a story for its own sake; it is a decision tool. It makes clear how the world would work, what would change first, how economics shift, and which moves become rational—all in a format FP&A can quantify and Risk can instrument. This section gives you (A) the structure of a decision‑grade narrative, (B) rules for names and hooks, (C) visual grammar that earns space on the page, and (D) a one‑page brief template you can drop into any gate pack.
A) Structure of a decision‑grade narrative
Use the five‑sentence logic—tight, testable, and behavior‑laden:
- States. Declare the axis values (or morphological factors) that define the world. Be explicit: regulation centralized/guardrailed, supply indexed/spot, cost curve steep/flat.
- Incentives & constraints. Name how key actors (customers, competitors, regulators, suppliers) must behave under these states. This is the mechanism section—no vibes.
- Economics. Directional anchors for demand, price/mix, unit cost, capacity/lead times, and working capital. Use ranges or arrows, not adjectives.
- Sequencing. What happens first, then next; where the path could fork. State natural signposts (see 13.1) tied to these turns.
- Contrast. One sentence contrasting the nearest neighbor world—why this is not that.
Add two operational add‑ons:
- Directional anchors sheet. One line for each key economic dimension (e.g., “capacity utilization tends to 88–92% within 2–3 quarters”).
- Implications for choices. Plain language: which no‑regrets, options, hedges, and big bets this world makes attractive or dangerous—and why.
Keep each narrative under 180 words before anchors and implications; anything longer hides the mechanism.
B) Names & hooks that carry the mechanism
Great names act like handles—people can pick up the world and carry it into meetings without the slide.
Rules that work:
- Mechanism first. Tie the hook to the causal spine: “Guardrails & Gatekeepers” (enforcement) beats “Headwinds”.
- Neutral tone. Avoid normative or fear‑laden framing; readers should not infer your preference from the name.
- Pairs that travel. Use symmetric, memorable contrasts across the set (e.g., “Spot & Sprint” vs. “Indexed & Inertial”).
- One idea per name. No compound metaphors; no inside jokes.
- Sticky but specific. Two or three words; avoid generic weather terms (Storm, Cloudy).
Quick test: If a BU head can retell the mechanism and the first signpost without the slide, the hook is working.
C) Visual grammar (only what earns decision time)
Visuals exist to clarify mechanism and timing—not decorate.
- Axis badge. A simple icon row showing the axis states (or checked morphological factors) for this world.
- Early‑move strip. A mini time‑line of what moves first (e.g., backlog, price resets, queue times), with the candidate signpost noted and its draft threshold window.
- Directional anchors pictogram. Five small arrows/bands (demand, price/mix, unit cost, capacity/lead times, WC). Each icon gets a footnote tag [MB‑##] or [A‑###]/[DRV‑###].
- No chart junk. No 3D, gradients, or cumulative waterfalls. Every visual must have a unit, scope, and “as of <date>.”
If a visual cannot survive the question, “What decision does this make clearer?”, it does not belong.
D) TPL‑09 Scenario Brief (one‑page template)
Use this layout; keep each element brief and tagged for traceability.
Header
- Name & hook. Two to three words that encode the mechanism.
- Axis/factor states. The values that define this world (or the curated combination in a morphological approach).
- Horizon tags. Monitoring / Evaluation / Commitment (from 3.1) so readers know what timing the narrative speaks to.
Five‑sentence logic
- States → Incentives/constraints → Economics → Sequencing → Contrast. Keep it tight; include one clause that references the highest‑value assumption [A‑###] and one candidate driver [DRV‑###].
Directional anchors (bullets, not a table)
- Demand: ↑ / ↔ / ↓ with an order‑of‑magnitude note and time‑to‑effect.
- Price/mix: band or rule of thumb (e.g., “index‑linked ±X bps per ΔY”).
- Unit cost: driver of change (learning curve, input price, scale).
- Capacity & lead times: staging cadence; bottleneck named.
- Working capital: inventory, AR/AP posture under this world.
Tag each line to model blocks [MB‑##].
Signposts & thresholds (draft)
- Two leading indicators you could measure within the monitoring horizon; for each, note the draft threshold window and hysteresis idea (e.g., “≥X for 30 days; unwind ≤X−δ for 45 days”). Link to indicator specs (13.1) if already filed.
Implications for choices (one paragraph)
- No‑regrets to enact now; options to pre‑buy (with expiry sense); hedges to consider (coverage bounds); big bets that only advance with specific evidence. Use mechanism language: “Indexing supply makes quarterly pricing rational; fixed‑price promotions are punished.”
Assumptions & risks (short list)
- The two most decision‑moving assumptions [A‑###], each with a falsification test; one structural risk (physics/policy) and one execution risk (capacity, partner, compliance).
Evidence & lineage footer
- [CIT‑###] source snapshots; “calc as of <date>”; owners for the brief (Strategy) and numbers (FP&A); Challenger named.
E) Build sequence (30–60 minutes per brief)
- Copy the axis/factor states from the Architecture Canvas (9.1).
- Draft the five sentences—start with incentives/constraints; fill states and economics second; end with contrast.
- Add directional anchors only where they change a decision; avoid inventing precision before 11.2.
- Draft two signposts and their threshold windows; ensure detection lead time ≥ action lead time (13.1).
- Write the implications paragraph using the taxonomy (no‑regrets, options, hedges, big bets).
- Tag everything with [A]/[DRV]/[MB]/[CIT]; add owners and “calc as of” dates.
- Read aloud once: if a BU head can say what changes Monday, the brief is ready.
F) Quality gates (reject if any fail)
- Mechanism clarity. Incentives/constraints are explicit; no vibe words.
- Instrumentable. At least one leading signpost has a draft threshold window and owner.
- Quant‑ready. Directional anchors map to known model blocks; no orphan claims.
- Decision separation. The implications paragraph makes different moves rational than in the neighboring world.
- Traceable. Every assertion and figure has a tag; evidence has a date and source.
G) Anti‑patterns—and the counter‑move
- Mood boards. Pretty names, no mechanism → rewrite incentives/constraints and signposts first.
- Over‑narration. Paragraphs of scene‑setting → enforce 180‑word limit before anchors.
- Forecast cosplay. Point numbers with false precision → replace with ranges and rules; leave decimals to 11.2.
- Trigger vibes. “Red/amber/green” signposts → write numeric thresholds with windows and hysteresis.
Build briefs to this template and your scenarios become portable, testable, and fundable—easy to retell, ready to quantify, and already wired to the triggers and options that will move when the world does.
10.2 Evidence Requirements & Citations—Checklist
Scenarios become credible when their claims are auditable. Your evidence standard should let a skeptical CFO (or a regulator) trace any statement from the one‑page brief back to a dated source and, if numeric, to a reproducible model block. Treat evidence like financial controls: tiered quality, versioned, and falsifiable. The goal is not volume; it’s decision‑grade sufficiency—the smallest, strongest set that changes or defends a choice.
What every citation must answer (five questions)
- What claim? Quote or paraphrase the specific assertion being supported.
- From where? Identify the source (publication/dataset/regulatory text/expert note) with a stable locator in the repository.
- When? Record collection date and original publication date; stale evidence is a silent bias.
- How measured? Note method, scope, and any transforms you applied.
- Who owns it? Name the internal owner (person) accountable for correctness and re‑use rights.
Source quality tiers (and rules of use)
- Tier 1 – Primary authoritative: statutes/regulations, regulator dockets and guidance, audited financials, official statistics, trade/market price feeds, vendor contracts. Rule: anchor structural claims here wherever possible.
- Tier 2 – Technical/sector analyses: peer‑reviewed papers, reputable think tanks, trade associations, sell‑side primers, vendor technical notes. Rule: acceptable where Tier 1 is unavailable; must be cross‑checked.
- Tier 3 – Commentary/press/blogs: journalism, newsletters, opinion posts. Rule: hypothesis generators only; never the sole basis for a trigger or a top‑five assumption.
- Expert judgment: interviews or advisory memos. Rule: cite by name (or anonymized per NDA), capture credentials, date, questions asked, and areas of uncertainty; pair with at least one Tier 1/2 corroboration for decision‑critical items.
Recency standards (set once and enforce)
- Fast‑moving markets (prices, capacity, rate paths): ≤ 8 weeks old at Gate 3.
- Regulatory texts and enforcement: cite current text plus most recent milestone; include “as‑of” date.
- Structural datasets (demographics, installed base): ≤ 12 months unless explicitly justified.
Document exceptions in the assumptions log (TPL‑10) with an expiry date.
Traceability grammar (use consistently across decks)
- Assumptions: [A‑###]
- Drivers: [DRV‑###]
- Model blocks: [MB‑##] (with owner and last calculation date)
- Evidence items (citations): [CIT‑###]
In the scenario brief, every assertion carries an assumption/driver tag; every number carries a model block tag; the footer lists [CIT‑###] items. The evidence pack hosts the details.
Evidence pack (one page per scenario; keep it tight)
- Claim map: list the 5–7 decision‑critical claims from the brief with their tags.
- Citations list: one to three CIT entries per critical claim (Tier 1 preferred), each with title, source, date, and a one‑line “why this supports the claim.”
- Transform note: if you transformed data (indexing, deflation, smoothing), include the exact formula or link to the reproducible script.
- Disconfirming twin: at least one contrary item for the two most important claims—logged even if rejected, with rationale.
- Rights & sensitivity: mark any non‑public/NDAd items; note distribution rules.
Numeric evidence—reproducibility requirements
- Lineage: raw snapshot stored in the repository (date‑stamped), transformation script/notebook saved and versioned, and output table linked to [MB‑##].
- Footnotes: every figure in TPL‑12 lists [MB‑##] and the evidence [CIT‑###] it draws from; “calc as of <date>” must appear.
- Rebuild test: a second analyst should be able to recreate the figure in one sitting from the artifacts; PMO runs spot checks before Gate 3.
Regulatory and legal hygiene
- Quote the rule, not the headline. Cite specific sections/paragraphs and current enforcement guidance.
- Interpretation vs. text: separate verbatim excerpts from your interpretation; log interpretation as an assumption [A‑###] with Legal as Contributor.
- Disclosure safety: Comms/IR reviews any externally redistributed chart; NDAs and license terms are logged in [CIT‑###].
Bias controls applied to evidence
- Disconfirming search mandatory for any Tier‑1‑anchored claim that is decision‑critical; the Challenger files the strongest contrary item.
- Denominator discipline: when citing “increase/decrease,” show base and sample (e.g., “12/47 approvals vs. 5/50 prior quarter”).
- Geographic balance: if a driver spans jurisdictions, include at least one non‑home‑market source or explain why a single jurisdiction governs.
Acceptance thresholds (decision‑critical vs. supporting)
- Decision‑critical claim (changes an option, hedge, or trigger): minimum two independent sources, at least one Tier 1; disconfirming twin logged; recency standard met; owner named.
- Supporting color (context only): one reputable source; clearly labeled as context; not used to set thresholds.
Citations metadata (what goes into [CIT‑###])
- Title, author/organization, publication and retrieval dates
- Source type and tier (1/2/3/expert)
- Stable link or repository path; version ID or checksum for files
- Scope (geo/time/segment), method summary, known caveats/limitations
- Rights/use notes; internal owner and next review date
“Red lines” for evidence
- No trigger or threshold may be justified solely by Tier‑3 commentary.
- No forward‑looking claim may appear without an assumption tag or model block reference.
- No deck advances to a gate if any exhibit lacks footnotes or has stale evidence beyond the recency standard.
Gate checks (read aloud)
- Gate 2: every scenario brief carries assumption/driver tags and a provisional citation list; any regulatory claim quotes current text.
- Gate 3: all ranges footnote to [MB‑##] and [CIT‑###]; a second analyst has passed the rebuild test for at least two critical exhibits.
- Gate 4: options and triggers cite the evidence that sets premiums, expiries, and thresholds; disconfirming items and responses appear in TPL‑10.
Anti‑patterns to cut on sight
- Link rot: “blog says” without a repository copy—replace with a stored PDF or dataset snapshot.
- Chart without lineage: pretty but unfootnoted—remove or rebuild.
- One‑sided packets: only confirming sources—add disconfirming twin or demote the claim.
- Timeless citations: structural data from years ago with no caveat—annotate or replace.
Build to this checklist and your scenario materials will withstand challenges. Leaders see the world, Finance trusts the numbers, Risk can set thresholds—and the organization moves because the evidence chain is short, strong, and owned.
10.3 Review & Sign‑Off—Gate Checklist
Gate reviews are where narrative meets governance. The purpose is not to “socialize” but to certify that a scenario brief and its evidence are decision‑grade, traceable, and wired to action. Run gates with a short agenda, clear roles, and objective pass criteria. Decisions are recorded in the room—no “we’ll confirm by email.” Anything that cannot change a choice is deferred to the next refresh.
How to run a gate (always the same five moves)
- Pre‑read discipline. PMO circulates the pack 24 hours in advance with version stamps and tag hygiene: every assertion carries an assumption/driver tag; every number footnotes a model block and citation. Late materials slip, not the standard.
- Open with the spine. Strategy Lead speaks for five minutes in the order World → Implications → Actions; no detours.
- Challenger airtime. Red Team presents disconfirming items and flip‑tests first; owners respond in writing within 48 hours in the assumptions log.
- Decision and record. Decision Owner states Approve / Approve‑with‑conditions (bounded in time/scope) / Not‑approved. PMO logs the decision, owners, budgets, and effective dates on the sign‑off sheet.
- Next step and SLA. If conditions exist, they have named owners, evidence required, and an SLA (days). The gate does not “pend.”
Gate 2 — Architecture Locked (Scenarios & Narratives)
Pass only if all are true:
- Distinct worlds. Three or four scenarios; each changes at least one option, hedge, trigger threshold, or timing/exposure of a big bet.
- Five‑sentence logic. Each brief states driver states, actor incentives/constraints, economic effects, sequencing, and contrast to a neighbor.
- Instrumentability. Two to three leading signposts per scenario phrased as if/then rules that can be monitored inside 3–12 months.
- Traceability. Assumption tags and driver codes appear on every assertion; initial citation list exists; cartoon extremes removed or grounded in mechanisms.
- Coherence & coverage. Cross‑impact notes exist for any high‑tension pair; no physics/regulatory contradictions; no obvious gap that would blindside the decision.
- Early quant anchors. Direction and order‑of‑magnitude hooks for demand, price/mix, unit cost, capacity/lead times, and working capital are written and agreed with FP&A.
Fail fast if: scenarios are tinted versions of one logic; axes co‑move; signposts are vibes (“tightening sentiment”) rather than measurable thresholds.
Gate 3 — Ranges & Evidence (Quantification & Exposure)
Pass only if all are true:
- One page per scenario. Standard metric ranges (demand, price/mix, unit cost, capacity/lead time, revenue, gross margin, opex, EBITDA, capex, FCF, net leverage) fit on a single page; plan break points are explicit.
- Stress clarity. Stress‑average and worst‑draw are shown per scenario; no cross‑scenario averaging; covenant headroom visible where relevant.
- Model lineage. Every figure footnotes a model block with owner/date; a second analyst has passed a rebuild test on at least two critical exhibits.
- Evidence standard. Decision‑critical claims carry at least two sources (one Tier‑1 where available); recency rules are met; disconfirming twins logged with rationale.
- Assumptions managed. High‑impact assumptions are in the log with falsifiable statements, thresholds, confidence labels, and challenge plans; status is current.
- Trigger candidates. Draft numeric thresholds exist for the signposts named at Gate 2; detection lead time exceeds action lead time.
- Risk/Legal concurrence. Hedge usage bounds and regulatory interpretations are reviewed; exceptions are documented with expiries.
Fail fast if: shadow spreadsheets appear; numbers lack footnotes; evidence is Tier‑3 commentary only; “probability theater” is used to weight structurally different worlds.
Gate 4 — Commitments & Triggers (Portfolio & Governance)
Pass only if all are true:
- Portfolio to action. The options canvas lists no‑regrets, options (with premium and expiry), hedges (with limits and unwind rules), and big bets (with Learning/Commit Gates); at least one real kill criterion exists.
- Trigger register. Each trigger is a numeric rule with owner, data source, and SLA: “If indicator ≥ X for Y days, do Z within N business days.” The Monitoring Dashboard Spec is dated and owned.
- Budgets & rights. Option premium pool, hedge caps, and analysis budget are approved; Decision Owner authority thresholds are written; any committee approvals complete.
- Board/Capital pack alignment. Narrative, ranges, options, and triggers in TPL‑19 align with the working materials; disclosure‑safe language reviewed.
- Hand‑offs. OKRs include at least one signpost KR and one delivery KR per relevant objective; the roadmap shows branch points and LG/CG dates; FP&A guardrails (bands) replace single‑line plan targets.
- Documentation complete. All assumption changes are reflected; versions are stamped; decision and effective dates logged.
Fail fast if: options have no premiums or expiries; triggers are color codes without thresholds; actions depend on future “alignment.”
Sign‑Off Sheet (TPL‑17) — one page per gate
- Gate number and date; pack version; decision at risk.
- Approvals: Sponsor, Decision Owner, FP&A, Risk, Legal/Treasury (as required), Challenger concurrence or dissent note.
- Decision wording (Approve / Approve‑with‑conditions / Not‑approved) and the exact commitments taken.
- Conditions, owners, evidence required, SLA to closure.
- Links to updated artifacts (Scenario Briefs, Ranges pack, Options & Triggers, Assumptions log).
- PMO record of effective dates and next review.
Red flags that block sign‑off
- Missing tags or footnotes anywhere in the executive pack.
- New analysis requests without a decision‑impact statement.
- Scope changes after Gate 1 without Sponsor approval and a new timeline/budget.
- Trigger breaches that prompt debate rather than action—rewrite the trigger before proceeding.
Run gates with this checklist and you convert narrative and evidence into authorized action: scenarios that leaders can recognize, numbers Finance can defend, triggers Risk can instrument, and commitments that move capital on time.
