Quantification & Modeling

This chapter turns coherent worlds into decision‑grade numbers. The point is not to predict a single outcome, but to translate scenario logic into operating and financial ranges that expose where the current plan fails, how options are priced, and which triggers matter. We keep three rules: work inside the scenario structure, show ranges before points, and preserve traceability from every figure back to drivers, assumptions, and model blocks.

11.1 Translate Narratives to Drivers & Assumptions—Step‑by‑Step

Goal. Produce a tight, auditable bridge from each scenario brief to model inputs: a small list of drivers with discrete states, the assumptions that tie them to business mechanics, and the model blocks they move. The outputs are (a) a filled Modeling Specification (TPL‑11) and (b) the first cut of the Ranges Pack (TPL‑12).

1) Re‑anchor on the decision and metrics.
Restate the one‑sentence decision, alternatives, and the standard metric set you will show for each scenario: demand/volume, price/mix, unit cost, capacity/lead time, revenue, gross margin, opex, EBITDA, capex, free cash flow, net leverage. This prevents model sprawl and keeps Finance focused on what will be decided.

2) Extract drivers from each narrative.
Take the five‑sentence logic and underline the explicit driver states (policy stance, ecosystem openness, input cost regime, adoption curve shape, supplier capacity, etc.). Limit to the five to seven that move value‑at‑stake. If a narrative clause does not map to a driver in the library, rewrite it or cut it.

3) Write the translation—plain language first.
For each driver, write one line that states how it moves business mechanics in this scenario. Examples: “Guardrails enforcement adds a fixed compliance opex and slows launch cadence by one quarter.” “Tight capacity lifts unit cost via premium sourcing and extends lead times by six weeks.” Only then translate to parameters. This avoids jumping to math without mechanism.

4) Map drivers to model blocks.
Link each translation line to the block it moves. Typical blocks include demand intercept and slope (elasticity), price/mix rule, pass‑through timing, unit cost curve, capacity/throughput and yield, cycle/lead time, working‑capital turns, capex cadence and depreciation, FX/commodity links, and covenant headroom. Assign a stable tag for each block (e.g., [MB‑05 Demand], [MB‑09 Unit Cost]) and record it in TPL‑11.

5) Choose time resolution and unit conventions.
Pick quarterly by default; move to monthly only where launch ramps or supply timing truly require it. Decide now: real vs. nominal, currency, calendar vs. fiscal, treatment of seasonality. Add a short “guardrails” paragraph in TPL‑11 so later requests don’t re‑open scope.

6) Parameterize with ranges, not points.
For each driver→block link, specify a P10–P90 band within the scenario (not across scenarios) and the central tendency you will show for planning. Anchor bands to evidence: base rates, market‑implied signals, or engineered bands from your driver states. Footnote each band to [CIT‑###] entries and the assumption it relies on [A‑###].

7) Declare assumptions in falsifiable form.
Every material parameter gets an assumption card in TPL‑10: the statement (“If input price basis remains ≤ X for ≥ Y weeks, pass‑through can hold at Z%”), the observable and threshold, the monitoring window, and the action if the assumption fails (exercise/expire option, adjust hedge, slip a gate). Name an Owner and a Challenger.

8) Encode reaction functions and constraints.
Where behavior matters, write the rule. Examples: “Price follows a floor/ceiling with quarterly resets if volatility ≥ V.” “Capacity adds in 10% tranches with 12‑week lag once utilization ≥ U.” “Opex flex ratio shrinks to r when revenue growth ≤ g.” Hard‑code physical/legal limits (throughput caps, minimum service levels, covenant floors) so worst‑case draws cannot violate reality.

9) Calibrate with two anchors.
Use a reference class (history or peers) for structural parameters and a market‑implied or vendor‑firm quote for volatile inputs. Where both exist, show the pair and choose the conservative for decisions that raise exposure, the optimistic only to size upside bands.

10) Produce early‑quant “directional anchors.”
Before full calculation, fill a one‑page Directional Anchors sheet per scenario: arrows and brief causes for demand, price/mix, unit cost, capacity/lead time, and working capital. Have FP&A sign this; it prevents later flips of sign that violate the narrative.

11) Build the within‑scenario range sheets.
FP&A now runs the model conditional on the scenario to produce TPL‑12: one page per scenario with bands for the standard metrics, plus stress‑average and worst draw. Each figure carries a [MB‑##] footnote and “calc as of <date>.”

12) Expose break points and flip conditions.
Add a small “break‑point strip” to each scenario’s page: the few thresholds at which plan viability fails or the preferred alternative flips. These come straight from the assumption cards and inform triggers later.

13) Record coherence and challenge.
PMO logs a short coherence check: do parameter signs and magnitudes respect the narrative and cross‑impacts? The Challenger runs one rebuild from raw sources and files any disputes in TPL‑10; Owners respond within 48 hours.

TPL‑11 Modeling Specification—fields you actually need

Decision at risk and metric set; time resolution and unit conventions; list of drivers used and their scenario states; driver→block links with [MB‑##] tags; parameter bands (P10–P90) with citations and assumption IDs; reaction functions; constraints and caps; ownership (FP&A block owner; Strategy counterpart); version/date and change log.

Acceptance checklist (five minutes before Gate 3)

Every narrative driver maps to a model block with a directional rationale and a range footnoted to evidence.
High‑impact parameters have falsifiable assumptions with thresholds and owners in TPL‑10.
Reaction functions and physical/legal constraints are encoded; worst‑case draws cannot produce impossible states.
One page per scenario in TPL‑12 shows standard metric bands, stress‑average, worst draw, and plan break points.
Tags are present on every assertion [A‑###]/[DRV‑###] and every number [MB‑##]/[CIT‑###]; a second analyst has passed one rebuild test.

Do this translation with discipline and you’ll get numbers that behave like your worlds, expose the tails that matter, and tie directly to options and triggers—so Gate 4 decisions rest on something sturdier than elegant prose.

11.2 Financial Translation (P&L, Cash, Balance Sheet) & Operating KPIs—Guide

Quantification only becomes decision‑grade when scenario logic flows cleanly through the P&L, converts to cash, and lands on the balance sheet with covenants and liquidity intact. The aim here is one chain of custody from narrative → drivers → model blocks → P&L ranges → cash waterfall → leverage and headroom—plus a short set of operating KPIs that explain why the numbers move. Work inside each scenario; show bands before points; preserve traceability to TPL‑11 (spec), TPL‑12 (ranges), and TPL‑13 (exposure & stress).

1) Translate to the P&L—mechanism first, math second

Start with the revenue build that the scenario implies and move down the statement.

Revenue (volume × price/mix): Derive volume from the demand block (intercept, slope, adoption curve) and price/mix from the pricing rule or contracting terms (indexation, floors/ceilings, pass‑through lags). If FX matters, keep nominal/real conventions explicit and footnoted.
COGS: Tie unit cost to the cost‑curve block (input regime, yield, learning) and capacity utilization. In supply‑constrained worlds, encode premium sourcing and scrap/rework.
Gross margin: Present as a range band; annotate the primary driver (price power vs. input basis vs. utilization).
Operating expenses: Separate scalable from fixed. Encode reaction functions (“marketing flexes to r% of revenue if growth ≤ g”; “contractor pool shrinks with six‑week lag”). Compliance or remediation opex sits here and should be scenario‑state driven.
D&A and other non‑cash: Link depreciation to capex phasing and lives; isolate non‑cash items so FP&A and Treasury can reconcile to cash.

Keep the P&L at the cadence of decisions (quarterly by default; monthly only where ramps or gating events demand it). Every line carries a model‑block tag and the few assumptions that matter.

2) Convert to cash—show the waterfall, not just EBITDA

Many models die here. Cash is where option premiums, hedges, and covenant tests live; make it explicit.

EBITDA → EBIT → NOPAT. Keep taxes scenario‑consistent (effective rate bands; loss‑utilization rules).
Working capital: Model AR, AP, and inventory with explicit days (DSO/DPO/DIO) that respond to scenario conditions (term tightening, vendor stretch, safety‑stock policy, backlog). Include deferred revenue and prepayments if your business carries them.
Capex: Phase by stage‑gate; flag committed vs. discretionary; tie to capacity lags. Include capitalized software or intangibles where accounting requires.
Financing: Interest bands from the leverage path; lease payments; share‑based comp cash effects if material.
Hedges & premiums: Show cash premiums and realized hedge P&L separately; include unwind rules to avoid hidden “stickiness.”
FCF band: Present P10–P90, stress‑average (mean of two worst draws), and worst draw per scenario. This is the number that sets risk budgets and guardrails.

3) Land on the balance sheet—headroom and runway

Balance‑sheet dynamics translate uncertainty into solvency and flexibility.

Leverage path: Project net debt using FCF bands and planned financing; include undrawn RCF and covenant tests (net leverage, interest cover, minimum liquidity). Show headroom as a band with the month of minimum headroom marked.
Liquidity runway: Weeks of runway under stress; state the trigger that would activate an RCF draw or another play (asset sale, cost program).
Capex‑in‑progress and WIP: Expose the portion that is irreversible; tie to kill criteria and unwind costs.
Off‑balance exposures: Guarantees, take‑or‑pay, long purchase commitments—surface them where scenario structure makes them bite.

4) Operating KPI spine—few, causal, and scenario‑relevant

Pick a compact KPI set that explains the P&L/cash behavior and can be instrumented as signposts or LG evidence. Five to eight is plenty; choose by function and scenario logic.

Demand & commercial: bookings/GMV, conversion rate, churn/retention, attach/uptake of key features, price realization vs. list.
Supply & delivery: capacity utilization, yield, cycle/lead time, backlog days, OTIF/service level.
Unit economics: CAC vs. LTV bands (where applicable), COGS per unit, contribution margin per cohort or SKU.
Working capital: DSO/DPO/DIO, backlog aging, advance billings.
Risk: breach probabilities of covenant thresholds (from Monte Carlo within the scenario), hedge coverage ratio vs. caps.

Write each KPI with its data source, cadence, owner, and whether it is a signpost KR or a delivery KR in OKRs. If it does not change a decision, drop it.

5) Presentation standard—one page per scenario (TPL‑12)

Every scenario page should look the same so differences pop.

Left column: P&L bands for the standard metric set with directional arrows and one‑phrase drivers.
Center: Cash waterfall (EBITDA → NOPAT → ΔWC → capex → financing → FCF) with stress‑average and worst draw called out.
Right column: Leverage path and covenant headroom band; liquidity runway callout; option premiums and hedge cash shown separately.
Footer strip: plan break points (where plan fails), flip conditions (thresholds that switch the preferred alternative), and three operating KPIs with current vs. band.

All figures have [MB‑##] footnotes and “calc as of <date>”; all causal statements carry [A‑###]/[DRV‑###] tags.

6) Controls that keep you honest

Double‑entry sanity: EBITDA + ΔWC − capex − taxes − interest ± hedges = FCF; FCF reconciles to net‑debt change every period.
Unit consistency: Nominal vs. real, currency, and seasonality treatment documented once in TPL‑11; no silent switches mid‑pack.
No double counting: Input‑cost shocks should not be both in COGS and again in opex “contingencies.”
Constraint checks: Worst draws cannot violate physical or policy limits (throughput, service level, covenants) because those limits are encoded.
Variant discipline: Tactical sub‑states (e.g., faster GPU normalization) live as within‑scenario variants and sensitivities, not new scenarios.

7) Acceptance checklist (use before Gate 3)

P&L bands trace to scenario drivers and reaction functions; price/mix, volume, and unit cost have explicit causal notes.
Cash waterfall is explicit; ΔWC drivers and capex phasing are scenario‑consistent; FCF bands show stress‑average and worst draw.
Leverage path and covenant headroom are visible with the month of minimum headroom; liquidity runway is stated.
Option premiums and hedge cash are separated; unwind criteria documented.
Five to eight operating KPIs are named with owners, sources, and whether they are signposts or delivery KRs.
Reconciliation checks pass; every figure is footnoted to [MB‑##] and dated; every assertion has assumption/driver tags.

Follow this chain and each scenario produces numbers that behave like the world you described, expose downside credibly, and connect cleanly to triggers and portfolio moves—so capital can be reallocated with speed and confidence.

11.3 Modeling Toolkit & Integrity (with Evidence Standards)—How‑To

Modeling in this playbook is decision plumbing, not a search for perfect forecasts. You are quantifying bands within a named scenario, exposing break points (e.g., month of minimum headroom), and showing how policies (pricing cadence, contracting, capacity staging, hedge coverage) change outcomes. This section explains when to model and how deep, how to choose the right tool, and the integrity standards that make every figure rebuildable and defensible.

A) Operating stance (what “good” looks like)

Within‑scenario, range‑first. Show P10–P90 bands, stress‑average (mean of two worst draws), and worst draw. Avoid cross‑scenario averaging; use scenarios to separate choices, not assign probabilities.
Policy‑aware. Encode reaction functions (what the firm would do under the scenario) before you simulate.
Headroom‑centric. Always report the month of minimum headroom and breach probabilities against guardrails.
Traceable. Every number footnotes a model block [MB‑##]; every parameter links to an assumption [A‑###] or driver [DRV‑###] and cites evidence [CIT‑###].

B) Narrative vs. Quant—decision rules

Use narrative only when: (1) the mechanism is primarily institutional/behavioral and already binds choices; (2) the decision is small and reversible; or (3) data are sparse and the immediate need is to define signposts and rules to act.
Go quant when: (1) multiple drivers interact non‑linearly; (2) guardrails (liquidity, leverage, SLAs) can breach; (3) you must price options/hedges/staging; or (4) the board requires numerical ranges to release capital.
Hybrid is the norm: lock the narrative’s five‑sentence logic (10.1), then quantify the elements that move headroom or breach probability.

C) Toolkit by question (pick the lightest tool that answers it)

What‑If (policy toggles).
Use for: pricing cadence, contract mix, inventory posture, tranche sizing.
How: deterministic runs with reaction functions on; report deltas to headroom/FCF and time‑to‑effect.
Output grammar: “Δheadroom at stress point,” “Δbreach probability within 12 months,” “time to effect.”
Sensitivity (one‑factor robustness).
Use for: understanding which parameters dominate risk (elasticities, pass‑through, supplier yields).
How: vary one input across its within‑scenario band; hold governance policies constant.
Output: ranked tornado; show usage bounds where the model remains valid.
Monte Carlo (tail exposure within a scenario).
Use for: breach probabilities, capital‑at‑risk pre‑trigger, hedge sizing.
How: sample uncertain exogenous drivers from your within‑scenario distributions; keep policy functions active.
Report: P10/P50/P90 for EBITDA/FCF/headroom; breach probability for guardrails; stress‑average and worst draw.
System Dynamics / causal loop (when feedback matter).
Use for: backlog/lead‑time cycles, adoption/learning curves, price‑capacity spirals.
How: small stock‑and‑flow modules calibrated to history; validate on shape and lead‑lag, not point fit.
Output: qualitative regime behavior + quantitative anchors to feed Monte Carlo or what‑ifs.

Heuristic: If changing a policy lever changes the distribution’s shape, you need at least Monte Carlo; if past shocks created oscillations, add a light SD module.

D) Parameterization & ranges (make uncertainty explicit)

Directional anchors first. From the scenario brief, set signs/order‑of‑magnitude for demand, price/mix, unit cost, capacity/lead times, and working capital (10.1).
Within‑scenario bands. For each key parameter, define a min/typical/max or distribution with notes: source, recency, and usage bounds (where the relationship likely breaks).
Reaction functions. Encode how pricing resets, how contracts index, how capacity tranches stage, and how inventory targets adapt under the scenario. Policies change the math more than parameter tweaks.
Constraints. Hard caps (plant, headcount, covenant) and floors (SLAs, compliance dates) must be explicit.
Time resolution. Choose the lowest resolution that preserves lead‑lag (often monthly for 12–24 months, quarterly beyond).

E) Integrity & evidence standards (non‑negotiable)

Evidence tiers & recency (link with 13.2/16.1).

Tier 1: filings, official series, ERP/CRM/SCM/Treasury data, auditable price feeds.
Tier 2: reputable research, audited vendor data.
Tier 3: commentary/estimates.
Model parameters draw ≥1 Tier 1 (or Tier1+Tier2); no Tier 3 alone. Record recency windows (e.g., quotes ≤30 days). Cite with [CIT‑###] and store snapshots.

Rebuildability.

Version TPL‑11 (Modeling Spec) and each [MB‑##].
Nightly rebuild tests for two exhibits (ranges page and breach probability) from stored snapshots.
Code and spreadsheets under version control; no calc logic in the visualization layer.

Reconciliation.

P&L → cash → balance sheet reconcile each run; headroom derived from cash and debt schedules; hedge cash flows separated.
Monthly aggregation matches accounting cadence; footnote any alignment assumptions.

Validation.

Backtest 12–24 months: show hit/false‑alarm rates for breach flags; show lead‑lag integrity for SD modules.
Challenger sign‑off on usage bounds and the top three assumptions [A‑###] that drive bands.
Cost‑of‑error note for any tuned threshold (why a false alarm is cheaper than a miss, or vice versa).

F) How to run the modeling cycle (7 steps)

Bind to a scenario. Name S‑## and paste the five‑sentence logic on the model cover.
Parameterize. Fill within‑scenario bands and reaction functions; tag sources and recency.
Choose the tool. What‑if → Sensitivity → Monte Carlo → SD, escalating only if the decision demands it.
Simulate & stress. Produce bands, stress‑average, and worst draw; locate the month of minimum headroom.
Summarize deltas. For each candidate policy/option, report Δheadroom at stress point, Δbreach probability, ΔFCF, and time‑to‑effect.
Write triggers. Convert sensitive parameters into indicator specs and trigger rules (13.1/12.4) with thresholds, windows, hysteresis, and unwinds.
File integrity notes. Rebuild test passed, reconciliation checked, backtests attached, change log updated.

G) Output grammar (what belongs in the pack)

Ranges strip: P10/P50/P90 bands for EBITDA, FCF, headroom with month of minimum headroom marked.
Exposure panel: breach probability vs. guardrails; capital‑at‑risk pre‑trigger.
Policy deltas: side‑by‑side what‑if results or option cards with Δheadroom, Δbreach probability, stress‑avg uplift vs. total cost, and time‑to‑effect.
Assumption call‑outs: the two [A‑###] that explain most of the band width and their falsification tests.
Traceability footer: [MB‑##] [A‑###] [DRV‑###] [CIT‑###], recency dates, usage bounds.

H) Anti‑patterns—and the counter‑move

Probability theater across scenarios. Counter: quantify within scenarios; use coverage/regret across worlds.
Point forecasts as promised. Counter: show bands, stress‑avg, worst draw; mark headroom minimum.
Sensitivity without policy. Counter: write reaction functions first; then sensitize.
Opaque spreadsheets. Counter: block‑structured [MB‑##], versioned scripts, rebuild tests.
Dirty evidence. Counter: enforce tier/recency; snapshots or it didn’t happen.
Trigger vibes. Counter: convert to numeric rules with windows and hysteresis or drop.

Build and run models this way and your numbers become actionable ranges: transparent, reproducible, and wired to the triggers and options that move capital when the world changes.

Scenario Planning Playbook >