Communicating Analytical Recommendations

The analysis was right and the decision still failed

An analytics team delivered a 47-slide deck on warehouse network optimization. Slide 38 showed the recommendation. The COO asked on slide 3: "What do you want me to do Monday?" The team had trees, scenarios, and a weighted scorecard (Lessons 1–4) but no decision sentence up front. The board deferred. Three months later a competitor moved. The model was not ignored because it was wrong. It was unusable under time pressure.

Communicating analytical recommendations is the capstone skill of Unit 6: translate EMV (expected monetary value), sensitivity, value of information, and multi-criteria scores into language executives can act on. A recommendation is not a chart. It is a claim (what to do), evidence (why now), assumptions (what must stay true), and risks (what breaks the plan), sized for the audience.

Managers fail here when they confuse precision with clarity, bury the lead in methodology, or hide uncertainty until the appendix. Finance wants thresholds; operations wants triggers; legal wants tail risks. One memo rarely serves all without layered structure. This lesson teaches pyramid communication, assumption ledgers, and pre-mortems that connect analysis to governance.

Unit 6 built analytical machinery: trees for sequential uncertainty, sensitivity for leverage, VOI for learning spend, MCDM for multi-goal choices. None of that value arrives if the recommendation memo reads like a methods textbook. Communication is not packaging. It is decision design: what gets decided, by whom, on what evidence, with what monitoring after the vote. A recommendation without triggers is a opinion; a recommendation with triggers is a policy.

Lead with the decision, not the model

Executives read top-down under stress. Use the pyramid principle: state the recommendation first, then the three supporting pillars, then methods for skeptics who want depth.

Template opening (adapt to context):

Recommend: Launch Region B in Q3 with a $2M cap and a 90-day conversion gate.
Why: EMV +$1.1M vs delay; tornado shows decision hinges on CAC payback, which pilot data already satisfies.
Risk: Downside scenario −$0.4M NPV if churn rises 200 bps; trigger pause if churn > 6.5%.
Ask: Approve budget and pre-register kill rule.

Notice what is absent from paragraph one: rollback notation, solver settings, regression tables. Those belong in backup.

Match precision to decision materiality (whether an item is large enough to influence a reasonable decision). A $50k marketing test needs one page. A $50M plant needs a memo plus appendix. Materiality guides depth.

Define acronyms once per document even if the audience "should know." New board members arrive every quarter.

The ask must be actionable: dollars, dates, owners, and gates. "We recommend further analysis" is not a decision sentence. "Approve $250k pilot with stop/go at 8% conversion by October 15, owned by VP Product" is. Committees defer when the ask is vague because deferral feels safer than committing to ambiguous scope.

One slide rule: if only one slide survives a meeting, it should contain recommendation, EMV or NPV headline, top tornado driver, and one risk trigger. Everything else is appendix. Analysts resist this because they fear oversimplification. Executives experience oversimplification as clarity.

Assumption ledgers, scenarios, and dissent

Every recommendation carries an assumption ledger: a table of beliefs that must hold for the action to succeed. Link each assumption to owner, evidence, and sensitivity rank from Lesson 2.

Scenarios (Lesson 2) become narrative bullets in the memo, not only Excel tabs. Downside should answer: "Do we survive?"

When EMV (Lesson 1) conflicts with risk policy (litigation tail, covenant), state dual view:

• EMV-optimal: Fight
• Policy-optimal: Settle unless reserve funded

Hiding conflict destroys trust. Boards prefer explicit trade-offs to false unity.

Document dissent and minority views when MCDM (Lesson 4) scores split functions. "Finance ranks Vendor A; Operations ranks Vendor B; weights make A win by 0.05 points, fragile to uptime weight."

VOI (Lesson 3) belongs in communication when recommending pilots: "Pilot EVSI $400k vs cost $250k; pre-defined stop/go at 8% conversion."

Assumption ledgers should include update cadence: monthly for CAC payback, quarterly for macro scenario, event-driven for competitor price moves. Stale assumptions are how good decisions decay into folklore six months later.

When two scenarios disagree on survival, show cash bridge not only NPV. Downside NPV −$0.4M might coexist with 14 months runway if spend is staged. Cash narrative prevents panic misread of discounted figures.

Visuals, numbers, and ethical framing

Good visuals reduce cognitive load:

• One decision tree snapshot with pruned branches labeled.
• One tornado for top five drivers.
• One scenario table with three rows, not twelve.

Bad visuals perform complexity: 3D charts, dual y-axes without labels, heat maps with unreadable fonts.

Number hygiene:

• Show one check line per critical calculation.
• Round for slides; keep exact backup.
• Never present false precision (NPV $3,247,891.33 from guesswork inputs).

Ethical communication:

• Disclose conflicts of interest (vendor paid demo).
• Do not cherry-pick scenarios that justify predetermined outcomes.
• Separate fact, interpretation, and judgment in prose.

For regulated contexts, archive the memo, model version, and assumption ledger together.

Tree visuals: label pruned branches with rolled-back EMV so viewers see why alternatives lost. Tornado visuals: sort by swing width, not alphabetical input names. Scenario tables: color is optional; words must stand alone in black and white PDF.

When presenting MCDM, show weights and top three scores, not a 12-criterion matrix. Invite skeptics to appendix for full scorecard.

Pre-mortems, triggers, and after-action reviews

A pre-mortem asks: "It is one year later and this decision failed. Why?" List failure modes before approval. Convert top failures to monitoring triggers tied to sensitivity break-evens.

Example triggers:

• Pause expansion if CAC payback > 13 months two quarters running.
• Re-open vendor choice if uptime < 94% in first 90 days.

Triggers make the recommendation durable beyond the meeting. They connect Lesson 2 break-evens to operating dashboards.

After-action review when outcomes arrive: compare realized to assumptions, update probabilities for future trees, do not blame individuals for honest forecast error. Learning loop improves organizational EMV over time.

For Unit 7 optimization memos, preview: report objective value, binding constraints, and shadow prices (marginal value of relaxing a constraint one unit) in plain language. Executives rarely want dual variables; they want "one extra machine hour worth $18 contribution."

Pre-mortems work best silent-then-share: each leader writes two failure reasons privately, facilitator clusters themes. Common clusters (demand miss, competitor response, integration delay) map directly to assumption ledger owners.

Triggers need threshold, measure, owner, and action. "Watch churn" is not a trigger. "VP CS pauses paid acquisition if logo churn rolling 90-day average exceeds 6.5%" is.

Audience layering: board, operators, and skeptics

Board layer: recommendation, EMV/NPV, Downside survival, explicit ask, dual view if policy diverges from EMV.

Operator layer: triggers, break-even thresholds, weekly metrics, who recalculates model when commodity prices move.

Skeptic layer (appendix): full tree rollback table, sensitivity tabs, Solver sensitivity report, MCDM score sheet with weights frozen before demos.

Send layers as one PDF with bookmarks rather than three conflicting decks. Conflicting versions breed rumor.

For litigation or HR decisions, legal review before circulation. EMV-optimal language can sound like recommending risky fight; frame as risk policy choice with quantified tails.

Optimization recommendations (Unit 7 preview)

When presenting Solver output, open with feasible mix and total contribution, then name binding constraints and top shadow prices. Example sentence: "Hours bind at 160; one more hour worth $5 contribution at margin; overtime at $4/hour is justified up to 10 hours per sensitivity allowable range."

Stress-test sentence mandatory for capital tied to mix: "Downside demand −15% shifts 40 Entree hours to minimum-contract SKUs; contribution falls 12%."

Pair optimization memo with Unit 6 scenario discipline: coherent Downside moves demand and margin together, not demand alone.

Memo templates by decision type

Capital expansion memo: Recommend / NPV / tornado drivers / Downside survival / ask dollars and date.

Vendor selection memo: Recommend / weighted score and fragile weights / veto status / EMV or TCO basis / ask contract signature authority.

Litigation memo: EMV-optimal / policy-optimal / tail outcome table / ask reserve or settle authority.

Weekly mix memo: Recommend quantities / contribution delta vs last week / binding constraint and shadow price / ask ops publish schedule.

Each template keeps pyramid order: action first.

Revising after board feedback

When the board sends you back for "more analysis," clarify whether they want new data (VOI on pilot), new weights (MCDM), or new risk policy (stress survival). Do not add fifty spreadsheet tabs without a decision question. Return with a one-sentence reframed ask.

Archiving and version control

Store Memo_v3_2026-07-01.pdf with Model_v3.xlsx and Assumptions_v3.csv in one folder. Version mismatch (memo cites v2 tornado, model is v3) undermines trust in future recommendations.

Oral presentation discipline

In live meetings, speak the recommendation in 30 seconds before sharing screen. Screen-first presentations lose executives who multitask. Tree and tornado slides support speech; they do not replace the opening sentence.

Handle Q&A with layered answers: short direct reply, offer appendix depth. "Shadow price is $5 per hour; details in sensitivity tab" respects time.

When analysis recommends "do not proceed"

Negative recommendations deserve the same pyramid structure. Recommend: Do not launch EU Phase 2. Why: Downside NPV −$1.2M with 25% scenario weight; tornado shows CAC payback fragile. Risk: proceeding burns $2M irreversible. Ask: Reallocate budget to core market test with EVSI justification.

Killing projects is a success mode for analytics, not a failure. Communicate it with the same rigor as approval.

Decision record for compliance

Regulated industries (utilities, healthcare) may require written basis for major commitments. Assumption ledger, tree rollback, and Solver sensitivity satisfy audit if archived together with approver names and dates.

Full sample memo (CloudLedger extended)

Recommend: Proceed DACH Phase 1 only, cap $6.0M through December.

Why: Base NPV +$3.2M; probability-weighted scenarios +$1.43M; EMV close to staging alternative. Tornado prioritizes Year-1 ARR and CAC payback; pilot achieved 11-month payback vs 14.8-month zero NPV threshold.

Risk: Downside scenario −$0.9M NPV and cash trough Month 4; trigger pause paid spend if CAC payback >13 months two quarters.

Ask: Approve Phase 1 budget; pre-register kill rule at 6.5% logo churn; assign CFO to publish assumption ledger monthly.

Appendix pointer: Decision tree rollback (Lesson 1), EVSI on optional DACH survey (Lesson 3, negative EMV gap), FreshPack-style capacity not applicable until Year 2 logistics model.

Practice problem 3

Write dual-view paragraph for NovaWear when EMV says Launch Now but marketing requests Study First for brand protection.

Solution: EMV-optimal: Launch Now (+$0.10M vs study path). Policy-optimal: Study First if management assigns ≥$200k value to avoiding weak launch visibility (non-EMV). Recommend board choose explicitly; finance shows EMV gap, marketing documents reputational override with trigger to launch if study favorable regardless (prevents infinite delay).

Stakeholder-specific one-pagers

CFO one-pager emphasizes NPV, tornado, covenant. COO one-pager emphasizes triggers, ops metrics, capacity shadow prices from Unit 7. GC one-pager emphasizes tail outcomes and policy-optimal branch. Same recommendation, different leading sentence, one archived bundle.

Q&A preparation sheet

List ten hostile questions ("What if churn doubles?" "Why not delay?" "Vendor B scored higher with uptime weight 0.35?") with two-sentence answers and appendix pointers. Preparation reduces meeting derailment.

Executive summary word limits

200 words for email approval, 600 words for board memo, unlimited appendix. Practice cutting methodology adjectives until each sentence carries decision content.

Part A: Header recommendation

Recommend: Proceed with EU launch Phase 1 (DACH only), budget $6.0M, not full EU-27 Day 1.

Part B: Evidence pillars

1. EMV/NPV: Base NPV +$3.2M; weighted scenario NPV +$1.4M with probabilities.
2. Sensitivity: Tornado ranks Year-1 ARR, CAC payback, gross margin as top three (Lesson 2).
3. Risk: Downside NPV −$0.9M without staging; staging limits burn to $2.5M before gate.

Part C: Assumption ledger excerpt

CAC payback 11 mo (pilot); break-even 14.8 mo. Trigger pause paid spend if >13 mo.

Part D: Managerial read / board ask

Approve Phase 1 with written kill rule at 6.5% logo churn rolling average. Full EU rollout only after gate memo.

Part A: Header recommendation

Recommend: Run Solver mix S=400, E=120, D=30 weekly; do not add overtime until hours shadow price exceeds $4/hour internal transfer rate.

Part B: Evidence pillars

1. Objective: Contribution $1,920 vs last week's ad hoc mix $1,740 (+$180).
2. Binding: Salad demand cap; hours slack 11 (non-binding).
3. Shadow price preview: If hours bind next week, marginal hour worth ~$5 (Lesson 5); compare to overtime $4.

Part C: Assumption ledger excerpt

Contributions from finance 7/10; demand caps from sales forecast; contract minimum 350 units. Trigger re-solve if chicken commodity +8% (cuts Salad contribution $0.40).

Part D: Managerial read

Ops accepts mix; sales warned Entree capped by capacity not market. Check: hours 149 ≤ 160 ✓; units 550 ≥ 350 ✓.

EMV-optimal: Fight (EMV −$5.2M vs settle −$8.0M).

Policy-optimal: Settle unless $25M litigation reserve funded; 40% chance −$22M branch breaches covenant.

Ask: Choose risk policy explicitly; do not hide behind average EMV.

Includes tree summary figure in appendix; memo body ≤2 pages.

Draft opening four sentences for a memo recommending Lease over Buy using Lesson 1 Ridge Manufacturing numbers (Lease net $390k vs Buy EMV $50k). Include one risk caveat.

Solution

Recommend: Lease equipment for three years rather than purchase the $500k machine. Why: Expected monetary value favors lease by $340k ($390k certain vs $50k EMV on buy). Risk: Lease foregoes upside if high-demand probability exceeds ~99%; if demand signals strengthen above 65% sustained, re-evaluate purchase. Ask: Approve lease and set review trigger at 12-month demand update.

List five rows for an assumption ledger for a price increase recommendation. Explain why "competitors do not react" needs an owner.

Solution

Example rows: (1) Price elasticity −1.2 (Marketing, conjoint study); (2) Retention unchanged (CS, pilot cohort); (3) COGS flat (Ops, supplier contract); (4) Competitor hold price (Strategy, intel owner required); (5) Implementation date July 1 (Legal, notice period). Competitor reaction assumption needs an owner because it is high leverage and often wrong; without ownership nobody monitors competitive response (Lesson 2 sensitivity spirit).

Treat recommendations as versioned products: memo, model, assumption ledger, sensitivity, tree figure, Solver report. Version ID in footer. Email subject line includes decision and date.

Train analysts to read recommendation paragraph aloud in under 20 seconds. If breath runs out, sentence is too long.

Negative recommendations deserve celebration in analytics culture; kills save capital.

Unit 7 shadow price sentences belong in ops memos: "Hour worth $5; overtime $4; take 10 hours max per sensitivity range."

After-action reviews update assumption ledger, not blame individuals.

Board education: one annual 30-minute session on reading tornado and dual recommendations reduces recurring confusion.

Extended practice: write full one-page memo (solution sketch)

Recommend: Settle Harbor litigation for $8M within 30 days.

Why: Policy-optimal given 40% chance −$22M tail breaches covenant; EMV-optimal fight saves $2.8M on average but risk committee rejects unbounded tail.

Risk: Settlement closes case; opportunity cost if win probability revised above 60% on new evidence.

Ask: Approve settlement authority and $25M reserve for future IP disputes.

Include tree appendix; memo ≤400 words.

Deep dive: translating regression to memo language

Instead of "coefficient −1.2 significant," write "10% price rise implies 12% volume drop in model, moving NPV from +$800k to +$200k; tornado ranks price second."

Deep dive: escalation triggers in writing

Every recommendation with positive NPV and Downside negative survival should include escalation trigger to CFO if cash months < X.

Closing integration across Unit 6 and Unit 7

Decision memos cite trees and scenarios. Mix memos cite Solver objective, binding constraints, shadow prices, and stress Downside mix. One analytics team serves both templates; version control covers both.

Oral delivery checklist

Stand up, state recommendation without slides, then show one tornado and one tree prune. Sit for Q&A with appendix open.

Communication is where analytical labor becomes organizational action. A correct model with a muddy memo loses to a rough model with a crisp ask. Unit 6 ends when you can write the ask in four sentences and defend every number in the appendix without hiding uncertainty.

Supplemental narrative: the 48-hour board packet

Imagine a board packet due in 48 hours for a $6M EU expansion. The analytics lead owns four pages maximum plus appendix. Page 1: recommendation, ask, triggers. Page 2: EMV or NPV headline, scenario table Base/Upside/Downside, survival sentence. Page 3: tornado top five with one-line managerial implication each. Page 4: assumption ledger with owners and update cadence. Appendix: tree rollback table, sensitivity tabs, MCDM score sheet with frozen weights, FreshPack-style capacity note if supply constrained.

The lead resists inserting regression diagnostics on page 1. The GC wants litigation-style dual view if EMV and policy diverge. The CFO wants break-even thresholds as triggers, not footnotes. Operations wants Monday actions: pause spend if CAC payback > 13 months. Each stakeholder receives clarity without a separate deck because layers live in one PDF with bookmarks.

Forty-eight hours is enough when templates exist from Lessons 1 through 5. Without templates, the team rebuilds methodology under stress and the COO asks on slide 3 what to do Monday. Templates are not bureaucracy; they are compassion for future you.

Supplemental narrative: communicating Solver results without jargon

When Unit 7 mix hits the memo, translate shadow prices: "One additional line hour worth $5 contribution at the margin; overtime at $4 is justified up to ten hours this week per sensitivity allowable range." Avoid dual variable vocabulary in the board body. Attach Sensitivity Report in appendix with version ID matching workbook FreshPack_v1.3.

Pair with Unit 6 Downside scenario: "Demand −15% and margin −10% together cut contribution 12% and shift 18 Entree hours to contract-minimum SKUs." Coherent scenario language prevents the board from assuming demand alone was stressed.

Supplemental practice walkthrough (Ridge memo, full)

Recommend: Lease three years. Why: Lease net $390k versus Buy EMV $50k; EMV favors lease by $340k. Risk: Buy wins if High demand probability exceeds ~99%; set 12-month demand review trigger. Ask: Approve lease signature this week; operations owns trigger dashboard. Check: EMV lease 390 vs buy 50 ✓. Dual view not needed; risk is upside foregone not tail loss.

Closing standards

Every recommendation memo ends with ask, triggers, and version ID. Every appendix ends with tree, tornado, or sensitivity evidence. Unit 6 communication skill is repeatable infrastructure, not presentation talent.

When in doubt, shorten the opening paragraph and strengthen the ask line; boards approve actions, not models.

Strong communication is iterative: draft memo, read aloud, cut methodology, add trigger, verify check lines, archive with model version, send once.

Pyramid structure, assumption ledger, and triggers are the minimum viable memo for any capital decision.

Communicating analytical recommendations completes Unit 6: trees, sensitivity, VOI, and MCDM only matter when the decision sentence is clear, assumptions owned, and monitoring hooks survive the meeting.

Practice the four-sentence opening until it feels automatic: recommend, why, risk, ask. That habit saves more capital than another tornado tab.

Unit 6 closes when analytical work produces decisions operators can execute Monday morning with triggers they can audit Friday afternoon.

Clear writing is part of the model, not packaging around it.

Key takeaways

• Open with recommendation, rationale, risk, and specific ask; methodology follows.
• Maintain an assumption ledger linked to sensitivity and scenarios.
• State EMV and policy views when they diverge; document fragile MCDM rankings.
• Use pre-mortems and triggers to connect analysis to operations after the meeting.
• Match document depth to decision materiality; keep backup auditable.

1. Rewrite a past deck opening as a four-sentence pyramid memo.
2. Add three trigger metrics to a live initiative using break-evens from your models.
3. Return to the unit page for Decision Analysis assessments, then begin Unit 7: Optimization and Managerial Modeling with Lesson 1: Optimization Problems, Objectives, and Decision Variables.