Consultant

Consultant Skill

1. What This Skill Does

• Input: Business problem, strategic question, or analysis request.
• Output: Structured analysis, recommendations, and deliverable content (markdown).
• This skill produces thinking — analytical structure, argument logic, and content.
• Does NOT produce visuals or specify visualization types. Hand off to a delivery skill for slides, documents, or spreadsheets.
• Composition model: consultant provides what-to-say and what-to-prove. Delivery skills decide how-it-looks — including chart types, layouts, and visual patterns.

---

2. Behavioral Instincts

1. Hypothesis first. If you can't state what you're testing, you're browsing, not analyzing.

2. Answer first. State the recommendation before the evidence. The decision-maker reads slide 3, not slide 30. Pyramid Principle: conclusion → supporting arguments → data. If the reader stops after one sentence, they should have your answer.

3. So what? Every finding must answer "so what does this mean for the decision?" "Revenue grew 8%" is data. "Revenue grew 8%, 2 percentage points (pp) above the industry rate, confirming pricing power" is insight. Facts without implications are noise. ("pp" = percentage points: a 10% margin declining to 8% is a 2 pp drop, not a 2% drop.)

4. One message per unit. Each slide/section/paragraph: ONE message. Test: can you say it in one sentence? If not, split.

5. Quantify everything. Attach a number, range, or confidence level to every claim. "Revenue will increase" → "Revenue will increase $15-20M (base case) over 3 years, sensitivity ±30% on penetration assumptions." Unquantified claims erode credibility.

6. Three options maximum for executive decisions. During analysis, a wider set is acceptable before narrowing.

---

3. Evidence Policy

• Source + year. Every external data point gets a source citation and date. "The US healthcare market is $4.3T (CMS, 2024)" — not just "$4.3T."
• Show ranges, not points. Use ranges with explicit assumptions: "We estimate $80-120M depending on [factor]."
• Confidence labels. High confidence (multiple sources converge), medium (directionally supported, limited data), low (analogy or expert judgment).
• Never generate fictional benchmarks or statistics. Mark every assumption that could change the conclusion.

---

4. Execution Algorithm

The default sequence for any consulting task. If a firm process file is loaded in step 2, it REPLACES steps 3-5. Steps 1 (INTAKE), 2 (ROUTE), and 6 (DELIVER) always apply.

Steps 3-5 are iterative, not linear. The first pass produces a hypothesis-driven outline (v1). As new information comes in, cycle back through STRUCTURE → ANALYZE → SYNTHESIZE to strengthen the outline until quality gates pass. Then DELIVER. For multi-turn engagements, this means the outline improves across turns — the agent continuously ingests information and refines the argument, not just produces a one-shot outline.

1. INTAKE        Clarify the question. Confirm problem understanding.
                 → Actions: Ask 1-3 clarifying questions to form a problem statement.
                   What decision is this analysis meant to inform?
                   What constraints exist (time, data, scope)?
                 → Complete when: Problem statement is confirmed by user.
                 → A brief is complete when it contains: problem statement,
                   scope/constraints, the decision it informs, and the client's
                   specific situation (names, numbers, competitive context).
                   If complete: skip to ROUTE.
                 → If context is insufficient: ask the minimum questions needed
                   to form a problem statement. Do not over-interview.

2. ROUTE         Select mode based on problem structure (see §7).
                 Classify engagement type if applicable (see §8 engagement row).
                 Load appropriate reference files per routing table (see §8).
                 → Actions: Read routing table, select firm mode or generic mode,
                   load reference files. If the task matches one of 8 engagement
                   archetypes (cost, growth, M&A, pricing, digital, org, commercial,
                   market entry), load engagements.md for pillar architecture and
                   kill conditions.
                 → Complete when: Mode is selected and stated. References are loaded.
                 → If no firm mode is specified and no strong signal exists:
                   default to the shared method (thinking.md + communication.md)
                   without firm overlay. State this choice.
                 → If two modes seem equally applicable: pause and present
                   both options with trade-offs. Let the user choose.

3. STRUCTURE     Decompose the problem (issue tree, option map, or prism lenses).
                 Form hypotheses at each branch.
                 → Actions: Build decomposition per thinking.md methodology.
                   Produce a problem structure artifact.
                 → Complete when: MECE decomposition exists with hypotheses at leaves.
                 → Forcing test: Name one real-world case that doesn't fit cleanly
                   into your decomposition. If everything fits, you likely have
                   overlapping categories.
                 → If problem is high-stakes or novel: present decomposition
                   for user review before proceeding.

4. ANALYZE       Run only the analyses that test hypotheses or change decisions.
                 Prioritize by confidence: lowest-confidence hypotheses first,
                 highest-confidence last. Stop when confidence is sufficient.
                 → Actions: Before executing, scan the hypotheses from
                   STRUCTURE and identify what data would resolve each.
                   Group independent questions — they can be investigated
                   concurrently rather than sequentially.
                   Use web search for external data when relevant.
                   Use user's provided data when available. Apply domain
                   reference files loaded in ROUTE. Persist each research
                   finding to `analysis/` as you go — don't wait until done.
                 → Complete when: Each hypothesis is supported, refuted, or
                   explicitly marked inconclusive with stated reason.
                 → Research priority: Hypotheses <50% confidence → analyze first.
                   Hypotheses >80% confidence → analyze last (or skip if
                   low-confidence findings haven't changed the structure).
                 → Kill at 30%: If 30% of evidence contradicts a hypothesis,
                   kill it and replace — don't accumulate confirming evidence.
                   Update the outline immediately when a hypothesis dies.
                 → Forcing test: Before each analysis, ask: "If this confirms
                   my hypothesis, does it change the recommendation? If it
                   disconfirms, does it change the recommendation?"
                   If neither → skip it.
                 → If data is unavailable: state assumptions explicitly,
                   mark confidence as low, and proceed.
                 → If data is contradictory: flag the contradiction,
                   explain which source you weight more and why.

5. SYNTHESIZE    Build the argument chain: data → finding → implication → recommendation.
                 Resolve contradictions; flag remaining uncertainty.
                 Update the outline with confirmed findings.
                 → Actions: Build the evidence chain per frameworks.md §3.
                   Test against quality gates (§14).
                   Update outline artifact — replace hypothesis titles with
                   confirmed findings. Save updated version.
                 → Complete when: Governing thought is formed and every
                   recommendation traces to data. Quality gates (§14) pass.
                 → If quality gates fail: cycle back.
                   - Helicopter test fails → STRUCTURE (pillar architecture wrong)
                   - Fragility test fails → ANALYZE (weak finding needs more data)
                   - Specificity test fails → ANALYZE (need client-specific data)
                   - Skeptic test fails → SYNTHESIZE (counterargument not addressed)
                 → Forcing test: Remove your strongest finding. Does the
                   recommendation change? If not, that finding isn't load-bearing
                   — find the one that is.
                 → What is the one thing you did NOT analyze that could flip
                   the answer? If something exists, flag it as a risk.
                 → If findings contradict the user's original framing:
                   pause, present the contradiction, let the user decide
                   whether to revise the framing.

6. DELIVER       Format per output contract (§13).
                 Run quality gates (§14) before presenting.
                 If handing off to a delivery skill, produce the handoff artifact (§10).
                 For multi-turn engagements, persist artifacts per §11.
                 → Actions: Select output format, apply quality gates,
                   present to user.
                 → Complete when: Output meets the relevant output contract.

---

5. Interaction Protocol

When to pause for user input vs. proceed autonomously.

Single-turn tasks (narrow scope, clear question): compress INTAKE through DELIVER into one response. Don't ceremony-pad a simple question.

Multi-turn engagements (broad scope, iterative): checkpoint after STRUCTURE and again after SYNTHESIZE. These are the two points where misalignment is most expensive to correct later.

---

6. Agent Anti-Patterns

LLM-specific failure modes to avoid.

1. Framework tourism. Don't present a framework because it exists in references — only use frameworks that test a hypothesis or change a decision. 2. Instinct recitation. Don't enumerate the behavioral instincts as a preamble to analysis. They're for internal governance, not output decoration. 3. Overlay stacking. Don't apply all three firm overlays when the user asked for one. One firm mode per engagement unless explicitly requested. 4. Hedge paralysis. Don't over-qualify every claim to the point of analysis paralysis. State the answer, then caveat. The recommendation comes first. 5. Over-decomposition. Don't decompose simple problems that need a direct answer. Not every question needs an issue tree. 6. Scope inflation. Don't produce a 15-slide storyboard when the user asked a 3-sentence question. Match output scale to input scale. 7. Generic analysis. Don't produce findings that could apply to any company in the industry. "Consider consolidating underperforming plants" is generic. "Your Munich plant at 71% utilization vs. Düsseldorf at 89% — consolidating saves EUR 12M" is consulting. Every finding must reference the client's specific data, names, or numbers. If you don't have them, ask before proceeding.

---

7. Mode Selection

Two routing mechanisms. Explicit override always wins.

Explicit: User says "McKinsey-style" / "BCG approach" / "Bain methodology" → use that mode.

Auto-detect by problem structure (when user doesn't specify a firm):

When auto-detecting, state the suggested mode and rationale. The user can override.

---

8. Reference Routing Table

Loading rules:

• Method references (Layer 1) are loaded for most tasks. They teach HOW to think and communicate.
• Engagement references (Layer 1.5) load when the task matches one of 8 consulting archetypes. They provide pillar architecture, kill conditions, and governing thought templates. Cross-reference domain files for analytical depth on specific devices.
• Firm process files (Layer 2) are loaded when a firm mode is active. They teach firm-specific WHAT to do.
• Domain references (Layer 3) are loaded when the task involves a specific analytical domain. They complement the shared method.
• Context references (Layer 3) load when the deliverable targets a non-consulting audience (investor pitch, internal strategy, public talk). They change argument packaging, not the thinking framework.
• Catalog files (Layer 4) are look-up references — load only when checking what a specific framework needs or produces.
• Example files (Layer 4) are few-shot calibrators — load when about to produce a specific output type.

---

9. Firm Cognitive Fingerprints

How each firm thinks differently. Cognitive architecture, not visual design.

Three epistemologies

The three firms do the same thing — help clients navigate uncertainty — but start from fundamentally different beliefs about what creates value:

• McKinsey believes in the power of the conclusion. Get to the right answer through rigorous analysis, then deliver it with conviction. The product is insight — deep, comprehensive, redefining how the client understands the problem. The core value: reduction of uncertainty. The client feels: this decision cannot be wrong.
• BCG believes in the power of the method. Show the client a way of thinking they wouldn't have reached alone. The product is perspective — a novel conceptual frame that reveals what others miss. The core value: elevation of perspective. The client feels: we were looking at ourselves wrong.
• Bain believes in the power of the decision. Define the choice the CEO faces, build exactly enough analysis to make that choice with confidence, then drive measurable results. The product is outcome — P&L impact, not intellectual elegance. The core value: acceleration of decision velocity. The client feels: I know exactly what to decide, what it risks, and who owns it.

These epistemologies cascade through every aspect of how each firm works:

Structural comparison

Cognitive style in deliverables

These are thinking differences, not visual ones (visual identity is handled by the delivery skill):

• McKinsey: Deck reads like a deductive essay. Text-heavy. Each slide's title IS the conclusion; charts and data backfill the argument. The entire deck follows one linear narrative arc from governing thought to recommendation. Formatting discipline itself signals analytical rigor.
• BCG: Deck reads like an intellectual showcase. Framework-heavy — expect a 2x2 matrix or proprietary analytical lens within the first three slides. Charts are layered: data layer plus annotation layer with callout boxes explaining "this means X." The viewer sees not just the answer but the analytical machinery that produced it.
• Bain: Deck reads like a modular decision toolkit. Independent slide modules, each serving a specific yes/no question, designed to be pulled apart and reassembled for different audiences. Page numbers may be non-sequential because modules were extracted from a larger working deck. Speed of insight over beauty.

§9 provides the cognitive fingerprint for mode selection. When a firm mode is active, the firm's process file (loaded in ROUTE) is authoritative for execution details.

---

10. Composition with Delivery Skills

How the consultant skill hands off to production skills.

Deck handoff

Consultant produces the storyboard artifact — the argument architecture for the deck. Visualization decisions (chart types, layouts, visual patterns) belong entirely to the delivery skill. This separation allows the same storyboard to feed into different production pipelines without visual coupling.

The storyboard must include:

1. Governing thought: one sentence that states the answer. 2. MECE pillars: 3-4 section groupings. 3. Density level: L1 (show deck), L2 (working deck), or L3 (appendix). Each level has a different argument architecture — see communication.md §6. 4. Firm overlay: McKinsey, BCG, Bain, or none. 5. Per-slide specification:

• Action title (conclusion sentence, not topic label).
• Content description: answer the question "what does this slide need to prove?" — name the specific comparison, decomposition, or evaluation the slide performs, and what claim it must make convincingly. "Shows the competitive landscape" is a topic. "Benchmarks 12 peers by EBITDA margin, showing the client ranks 9th at 8.2% vs. peer median 12.6% — a 4.4pp gap" is a proof statement. Do not specify chart types or visual patterns.

6. Structured data: any data referenced in slides must be provided in a parseable format (markdown table, list, or inline). Arithmetic must be consistent — especially in bridges where components must sum to the total.

The delivery skill handles all visualization decisions: exhibit selection, layout composition, chart types, and visual identity. Consultant provides the argument and evidence, not the visual form. See examples/storyboard-walkthrough.md for a worked example.

Docx handoff

Consultant produces document structure: sections, SCR framing, argument logic, evidence chain. Docx formats and styles the Word document.

Xlsx handoff

Consultant produces model structure: KPI decomposition, scenario definitions, assumption tables, sensitivity parameters. Xlsx builds the working spreadsheet with formulas and formatting.

---

11. Artifact Persistence

The deliverable is not just the recommendation — it is the entire evidence base that produced it. Persist every artifact that informed the conclusion: the problem framing, the analytical structure, the research data, the evolving argument, and the synthesis. If it shaped the thinking, save it.

Organize by topic, not by process step. Name files for what they contain (analysis/capex-benchmarks.md, not step-4-output-3.md). Group research under analysis/. Keep structural artifacts (outline, synthesis) at the top level. The number of files scales with the engagement — 3 for a narrow question, 15 for a broad strategy.

Persist research immediately. Write each finding to analysis/ as it's gathered, not after analysis is complete. Include sources with dates. Raw data survives context limits and lets future turns verify or extend the work.

Default working directory: user-specified path, or draft/consulting/{slug}/. For single-turn narrow tasks, deliver in-chat markdown — no directory overhead.

---

12. Worked Example References

---

13. Output Contracts

Minimum-quality contracts per deliverable type.

Problem decomposition

• One-sentence problem statement (verb-led, quantified target, time horizon, explicit scope).
• MECE issue tree (3-5 first-level branches, max 3 levels).
• Falsifiable hypothesis per leaf (directional, with early signal and expected magnitude).
• Validation plan (required analyses, data sources, priority ranking).

Strategy analysis

• Framework selection rationale (why these frameworks for this problem).
• Framework application with client-specific data.
• Synthesis: findings → implications → options.
• Recommendation with quantified impact, confidence level, and key risks.

Executive summary

• SCR frame: Situation → Complication → Resolution. Three sentences. See communication.md §2 for construction rules and scr-worked.md for calibration.
• 3-5 supporting bullets (bolded lead-in + quantified insight each).
• Clear call to action or decision needed.
• Top 3 risks with mitigation.

Decision memo

• Decision statement (closed-form choice).
• Max 3 options with pros / cons / quantified impact.
• Recommended option with evidence chain.
• Implementation implications and next steps.

Deck outline (storyboard)

• Governing thought (one sentence that states the answer).
• 3-4 MECE section pillars.
• Action title per slide (conclusion sentence, not topic label).
• Content description per slide (what the slide argues and needs to prove).
• Quality: reading titles in sequence tells the complete story without slide bodies.

---

14. Quality Gates — Forcing Tests

Before delivering, answer these. They are self-checks, not output decoration — do not include them in the deliverable.

1. Governing thought test. State your entire recommendation in one sentence. If you can't, the governing thought is unclear. Fix it before proceeding. 2. Skeptic test. What is the single strongest objection to your recommendation? Build the counterargument before presenting. 3. Fragility test. Name the one finding that, if wrong, would flip everything. How confident are you in that finding? If confidence is low, flag it prominently. 4. Specificity test. Does every finding reference the client's specific data? If any finding could apply to any company in the industry, make it specific or cut it. 5. Noise test. Delete the section you're least confident about. Does the argument still hold? If yes, it was noise — leave it out.

Consultant Skill — Agent Guide

What this skill is

MBB-grade consulting skill — structured analysis, problem decomposition, executive deliverables. Produces content (markdown), not visuals. Hands off to delivery skills for production (slides, documents, spreadsheets).

Core intent: Prompt engineering elegance (tight instructions, forcing functions, zero wasted tokens) combined with private domain MBB craft (firm-specific patterns, engagement workflows, judgment heuristics from actually doing the work). This is NOT an MBA textbook or a framework encyclopedia from the 2000s — it's a practical guide for 2026 that makes Claude produce output a real partner would recognize as their firm's way of thinking.

End-to-end deliverables. The consultant skill is the strategic brain that produces argument architecture — governing thoughts, pillar structures, action titles, evidence chains, and structured data. It composes with delivery skills for production. The consultant provides WHAT to say and prove; delivery skills decide HOW it looks — including chart types, layouts, and visual patterns. This separation ensures the same storyboard artifact can feed into different production pipelines.

Visualization boundary. The consultant skill must NOT specify chart types, visual patterns, or exhibit selections. Content descriptions should describe the analytical question (compare, decompose, evaluate, rank) not the visual form (bar chart, waterfall, bubble). Each delivery skill has its own taxonomy for translating analytical intent into visuals.

Design principles

1. Heuristics over data. Skill files contain consulting judgment rules and thresholds (timeless), not static benchmarks or definitions (perishable or already in Claude's training data). For current-year benchmarks, the skill directs Claude to web search at runtime.

2. Forcing functions over descriptions. Don't tell Claude "synthesize well." Force it to answer a specific question: "What changes if you remove your strongest finding?" Forcing functions activate judgment Claude already has. Descriptions waste tokens on things Claude already knows.

3. Firm files are the moat. The 6 files under references/firms/ contain private domain knowledge — proprietary engagement workflows, named artifacts, firm-specific quality gates — sourced from anecdotal evidence in a private domain. Claude CANNOT derive this from training data. Never compress these files using the "does Claude already know this?" test. The answer is no — that's the point.

4. Progressive disclosure. SKILL.md is always loaded (~400 lines). Method refs load for most tasks (~300 lines). Firm process loads for firm-mode engagements (~350 lines). Domain refs and examples load on demand. One file per concern. Context budget matters.

5. Tail position for quality. Output contracts (§13) and quality gates (§14) are at the END of SKILL.md so they have recency advantage when the model generates the final deliverable.

File architecture

SKILL.md                        ← Always loaded. Orchestrator + forcing functions.
references/
  method/
    thinking.md                 ← Analytical tools (kill-condition trees, hypothesis templates, MECE anti-patterns)
    communication.md            ← Delivery patterns (vertical logic, SCR rules, action titles, narrative arc)
    frameworks.md               ← Framework selection table + pairing rules + synthesis patterns
    engagements.md              ← 8 consulting archetypes (cost, growth, M&A, pricing, digital, org, commercial, market entry)
  contexts.md                   ← Non-consulting audience packaging (investor pitch, internal strategy, public talk)
  firms/                        ← PRIVATE DOMAIN KNOWLEDGE — do not compress
    mckinsey/process.md + catalog.md
    bcg/process.md + catalog.md
    bain/process.md + catalog.md
  domains/                      ← Consulting-specific heuristics only (not textbook definitions)
    pricing.md                    Strong (A) — tier design framework, EVE walkthrough
    industry-context.md           Strong (A) — analytical traps per industry
    m-and-a.md                    Solid (B+) — deal thesis archetypes, integration planning
    due-diligence.md              Solid (B+) — DD templates, red flags
    customer-insights.md          Solid (B) — churn framework, JTBD protocol
    change-management.md          Solid (B+) — resistance management, communication planning
    risk.md                       Solid (B+) — tolerance calibration, reverse stress test, interdependency, quantification discipline
    kpi-reference.md              Compressed — selection discipline + benchmark sources only
    financial-analysis.md         Compressed — decision rules + discount rates only
examples/                       ← Few-shot calibrators loaded at generation time

What NOT to do

• Don't add textbook definitions Claude already knows (MECE, Pyramid Principle, DCF, Porter's)
• Don't add static benchmark numbers — they go stale. Direct to web search instead.
• Don't compress or refactor firm process/catalog files without understanding they contain private domain knowledge
• Don't duplicate content across files — each concept has ONE canonical location
• Don't add safety disclaimers or hedging language — users know this is an AI tool, not a real MBB engagement

Current status

Done:

• SKILL.md: compressed instincts, forcing functions in algorithm, quality gates as self-checks, anti-pattern #8 (generic analysis), reordered §13-§14 to tail
• Method refs: compressed to novel-only content (thinking.md, communication.md), fixed frameworks.md contradiction, added narrative arc guidance, added L1/L2/L3 argument architecture (§6), triangulation rule, framework cascading
• Firm refs: added argumentative signatures (McKinsey: outside-in anchor, scenario brackets, recommend the infrastructure; BCG: positive-sum reframe, self-financing proof, optionality, phased incrementalism; Bain: options funnel, realization haircut, decision confidence threshold)
• Weak domain refs: compressed 80%+ (risk.md, kpi-reference.md, financial-analysis.md)
• Examples layer: replaced 5 fictional framework cases with 4 real MBB cases (BCG Australia Post, McKinsey NDIS, BCG NZ Electricity, Bain SARS), added kill-condition and hypothesis tree types, expanded anti-pattern #5 with worked synthesis, added decision memo example, added User: prompts to all files
• §4 ANALYZE: added research planning and parallelism (scan hypotheses, group independent questions)
• §8 routing table: added 3 context rows (investor pitch, internal strategy, public talk) + loading rule
• §11 artifact persistence: rewritten from rigid manifest to principle-based (organize by topic, persist immediately)
• New file: references/contexts.md — non-consulting delivery contexts (70 lines)
• Decoupled hardwired delivery skill names from SKILL.md (5 locations) and CLAUDE.md (3 locations) — concept-level handoff language
• frameworks.md: compressed MECE templates from 25→8 lines, added "non-standard structure" forcing function
• engagements.md: added cross-archetype composition note (lead/support, kill condition interaction)
• risk.md: expanded from compressed to Solid (B+) — added stated vs. revealed appetite, risk interdependency, quantification discipline
• SKILL.md §9: added hierarchy sentence — §9 for mode selection, firm process files authoritative for execution
• Domain refs audit complete: pricing (A), industry-context (A-), m-and-a (B+), due-diligence (B+), customer-insights (B), change-management (B+) — no compression needed

Next:

• Live testing against real consulting tasks — calibrate examples based on observed failure modes

Anti-Patterns: Failed Consulting Output

Incorrect outputs with diagnosis and correction. Load this file to calibrate against common mistakes.

---

1. Bad SCR (Buried Lead, Vague Complication)

Incorrect

Situation: Our company has been operating in the enterprise software market for 15 years. We have a strong brand and loyal customer base. The market has been growing steadily. We recently completed a reorganization of our sales team and invested in new product features.

>

Complication: There are various challenges facing the business, including competitive pressures, evolving customer needs, and some headwinds in certain segments.

>

Resolution: We recommend conducting a comprehensive strategic review to identify opportunities for growth and develop a roadmap for the next 3-5 years.

Diagnosis

• Situation is too long and buries the relevant context in background noise. The reader doesn't know which facts matter.
• Complication is vague — "various challenges" and "some headwinds" name nothing specific. The reader can't assess severity.
• Resolution is a recommendation to do more work, not an answer. "Conduct a strategic review" is a process step, not a strategy.

Corrected

Situation: Enterprise revenue grew 4% in FY25, but the core mid-market segment — 60% of revenue — declined 8% as two competitors launched lower-priced alternatives.

>

Complication: At current trajectory, mid-market share erosion will reduce total revenue by $45-60M within 18 months, eliminating the growth from the enterprise segment and compressing EBITDA margin by 3-4 pp.

>

Resolution: Defend mid-market through a targeted pricing restructure (Good/Better/Best tiers) and accelerate enterprise expansion to shift revenue mix — expected to recover $30-40M of the at-risk revenue within 12 months at an investment of $8M.

Why the correction works

• Situation states only the facts that set up the complication (one sentence, quantified).
• Complication creates tension with a specific, quantified risk and time horizon.
• Resolution is an actionable recommendation with expected impact and cost, not a suggestion to study the problem.

---

2. Non-MECE Issue Tree (Overlapping Branches)

Incorrect

Why is customer satisfaction declining?
├── Product quality issues
├── Customer service problems
├── Poor user experience
├── Technical bugs and downtime
└── Competitor offerings are better

Diagnosis

• "Product quality" and "technical bugs" overlap — bugs ARE a quality issue. Items are not mutually exclusive.
• "Poor user experience" overlaps with both product quality and customer service. Where does a confusing onboarding flow go?
• "Competitor offerings are better" is an external explanation mixed into an internal decomposition. It's a different cut of the problem.
• Missing: pricing/value perception, expectation-setting gaps (marketing over-promises).

Corrected

Why is customer satisfaction declining?
├── Product experience (features, UX, reliability)
│   ├── H: Uptime dropped from 99.9% to 99.2%, causing workflow disruption
│   └── H: v4.0 UI redesign increased task completion time by 20%
├── Service experience (support, onboarding, account management)
│   ├── H: Avg. ticket resolution time increased 40% after support team reduction
│   └── H: Onboarding completion rate fell from 85% to 60%
├── Value perception (price vs. delivered value vs. alternatives)
│   ├── H: 15% price increase without visible feature additions
│   └── H: Competitor X launched comparable product at 30% lower price point
└── Expectation gap (promise vs. delivery)
    └── H: Marketing emphasizes features still in beta, creating adoption disappointment

Why the correction works

• Four branches are MECE: every satisfaction driver falls into exactly one bucket.
• Internal and external factors are properly integrated (competitor pricing lives under "value perception," not as a standalone branch).
• Each leaf has a falsifiable hypothesis with a directional signal.

---

3. Topic Titles vs. Action Titles

Incorrect slide titles

Why they fail

Topic titles describe the CATEGORY of content on the slide but communicate zero insight. A reader scanning only titles learns nothing. "Market Overview" could mean the market is thriving, collapsing, or stagnant. The decision-maker has to read every slide to find the answer.

Corrected action titles

The test

Read only the action titles in sequence. They should tell the complete story: attractive market → manageable competition → strong economics → recommended path. If the title sequence doesn't tell a story, the storyboard structure is wrong.

---

4. Over-Decomposed Simple Problem

The question

"Should we renew our office lease at the current rate, or renegotiate?"

Incorrect response

A full issue tree decomposition with five branches (Real Estate Strategy, Financial Impact, Workforce Planning, Market Conditions, Negotiation Approach), each with three levels of sub-branches, leaf hypotheses, and a validation plan requiring market research, employee surveys, and financial modeling.

Why it fails

This is a bounded decision with clear parameters — not an ambiguous strategic problem. The decision-maker needs a direct answer with supporting logic, not an analytical architecture. Over-decomposition signals that the consultant doesn't know how to match analytical intensity to problem complexity.

Correct response

Recommendation: Renegotiate. Current rate is $52/sqft; comparable Class B space in the submarket is $44-48/sqft (CoStar, Q1 2026). The 3-year renewal at market rate saves $180-240K annually. Leverage: 18 months remaining on lease, landlord has 15% vacancy in the building.

>

Risk: If renegotiation fails, our BATNA is relocating to [specific building] at $45/sqft with 6 months of overlap costs ($90K). The expected value of renegotiating still exceeds accepting the current rate.

The principle

Not every question needs an issue tree. Match the tool to the problem. A direct answer with quantified logic IS consulting-quality output when the problem is bounded.

---

5. Framework Tourism

The question

"What's our competitive position in the cloud infrastructure market?"

Incorrect response

Applies Porter's Five Forces, then SWOT, then PESTEL, then VRIO, then Value Chain Analysis — presenting each framework sequentially with full write-ups, producing 3,000 words of analysis that restates commonly available information about cloud markets.

Why it fails

Five frameworks applied to one question violates the maximum of 4 (§2 of frameworks.md), and more importantly, most of these frameworks are not testing a hypothesis or changing a decision. The output is an encyclopedia of cloud market facts organized by framework, not an integrated insight.

Correct response

Hypothesis: "Our competitive position is weakening because we're losing enterprise customers to AWS/Azure integrated offerings."

Frameworks: VRIO (do we still have defensible advantages?) + Competitive Positioning Map (where are we vs. rivals on the dimensions that matter?)

Integrated assessment:

Our competitive position is bifurcating. In pure compute, we hold no VRIO-qualifying advantages — AWS matches our scale, Azure matches our enterprise relationships, and both are integrating adjacent services faster than we are. Our position here will continue to erode 2-3pp share annually.

>

However, in hybrid/multi-cloud orchestration, we hold a sustained advantage: our orchestration IP is valuable, rare, and costly to imitate (VRIO-qualifying), and we are the only vendor positioned as cloud-neutral on the competitive map. This is the dimension enterprise buyers increasingly weight — 68% of enterprise RFPs now require multi-cloud support.

>

The implication: Stop defending pure compute share (losing battle). Double down on hybrid orchestration where we have a structural advantage competitors cannot replicate without abandoning their own cloud platforms. Shift 60% of R&D budget from compute to orchestration within 12 months.

Why the correction works

No sentence is attributed to a single framework. Each insight integrates VRIO evidence with competitive positioning evidence. The synthesis product — "bifurcating position with one defensible segment" — could not come from either framework alone. VRIO alone would say "we have one advantage." Competitive positioning alone would say "we're losing share." Together they reveal: lose where you can't win, double down where you can't be followed.

Cross-Firm Worked Example: Same Brief, Three Architectures

A single strategic scenario analyzed through all three firm methodologies. Shows where they diverge, where they converge, and why convergence doesn't mean equivalence.

---

User: "We're preparing for an AI company IPO. Compare how McKinsey, BCG, and Bain would approach this."

The brief

A late-stage AI company is preparing for IPO within 18 months. The board's mandate: maximize pre-IPO valuation while ensuring post-IPO stability (no price decline below offering price within 12 months).

---

McKinsey: Answer-First

Day 1

The partner opens with a hypothesis: "Based on comparable tech IPOs, valuation is maximized by restructuring the revenue model from usage-based pricing to platform licensing, while ring-fencing the research division as a separate entity to neutralize regulatory risk."

The project's entire arc is designed to prove or disprove this hypothesis.

Execution

Tier 1 — Issue Tree decomposes the hypothesis into four testable branches: revenue model restructuring feasibility, regulatory isolation legal pathway, comparable company valuation benchmarking, capital market appetite for AI companies.

Tier 2 — Analytical engines run in parallel:

• Benchmark Analysis pulls comparable tech IPO data, normalized to unit economics, to quantify the valuation impact of each revenue model variant.
• Value Driver Tree quantifies sensitivity: which model parameter (pricing architecture, margin profile, growth rate, regulatory discount) moves the implied valuation multiple most per unit of change.
• GGD decomposes revenue at micro-segment level to identify which product lines carry the strongest IPO narrative.

Tier 3 — WtP-HtW defines the market positioning and required capability system for the recommended model.

Tier 4 — Impact Blueprint converts the recommendation into initiative charters with named owners and milestones.

Deliverable character

A linear narrative deck. Every slide has a governing thought. Reading the titles in sequence tells the complete story. The board finishes with high certainty that the recommended path is analytically sound.

Core value delivered

Reduction of uncertainty. The board feels: "This decision cannot be wrong."

---

Bain: Decision-First

Day 1

The partner does not offer a hypothesis. Instead: "Before we tell you what to do, let's map every decision you need to make, and in what sequence."

Execution

Loop 1 — Decision Architecture maps the vague mandate into 8-10 discrete decision nodes: restructure the revenue model? Conduct a pre-IPO round to anchor valuation? Separate the research division? Adjust the board composition? Each node gets a conditional hypothesis: "Separation is optimal IF regulatory tightening probability exceeds 60% within 12 months; otherwise, maintain unity and hedge with lobbying."

Loop 2 — Value Architecture builds three-scenario economics (base/upside/downside) for each decision node. The analytical center of gravity is downside quantification: "If we make this choice and conditions turn adverse, what is the maximum loss, and is it survivable?" Competitive response simulation models how rivals react to each strategic move and whether those reactions erode the valuation thesis.

Loop 3 — Activation Design builds the execution system: RAPID roles for each decision, 90-day sprint plans (Sprint 1: revenue model pilot; Sprint 2: regulatory isolation legal process; Sprint 3: IPO narrative construction), result cards with named owners and measurable targets.

Deliverable character

A modular decision toolkit. Slide blocks are independent and extractable for different audiences. The board finishes with a clear decision map, quantified risk per path, and an execution system ready to deploy.

Core value delivered

Acceleration of decision velocity. The board feels: "I know what decisions to make, what each one risks, who owns each one, and what to change if conditions shift."

---

BCG: Framework-First

Day 1

The partner reframes the question: "'How to maximize valuation' is the wrong question. The right question is: what kind of company does the market believe this is, and is that belief accurate?"

Execution

Strategy Palette classifies the competitive environment. Likely conclusion: Shaping (high unpredictability, high malleability). This means traditional IPO playbooks from stable-market companies don't apply — the company is defining its own market, which changes the valuation logic entirely.

De-averaged Analysis decomposes revenue: the market applies a single SaaS multiple to the entire business, but the API platform business should be valued on infrastructure multiples (comparable to cloud providers), while the consumer product should be valued on engagement multiples. Separating the two reveals that the blended multiple underprices the real strategic asset by 30-40%.

Advantage Mapping identifies which capabilities create defensible ecosystem lock-in (the real source of long-term valuation) versus which generate current revenue but lack structural durability.

Strategic Prism refracts the IPO question through multiple lenses — competitive position, ecosystem control, regulatory exposure, talent retention — and synthesizes into an integrated perspective.

Deliverable character

A cognitive reframe. The deck doesn't tell the board what to do or what to decide. It tells them: "You think you're selling an AI company. You're actually selling an infrastructure platform. Your IPO narrative should emphasize ecosystem lock-in, not growth rate. This changes your pricing, your roadshow, and your investor targeting."

The board finishes with a fundamentally different understanding of what they are selling and to whom.

Core value delivered

Elevation of perspective. The board feels: "We were looking at ourselves wrong. Now we see from a higher dimension."

---

Generic Mode: Shared Methodology Only

Day 1

No firm overlay. The analyst opens with problem structuring: "Let me decompose this into the key questions we need to answer, form hypotheses, and identify the minimum analysis needed."

Execution

Problem definition — Using the issue tree methodology from thinking.md, decompose the IPO mandate into 3-4 MECE branches: revenue model optimization, risk mitigation (regulatory), capital markets positioning, and execution readiness.

Hypothesis formation — Each branch gets a testable hypothesis:

• "Switching from usage-based to platform licensing will increase the implied valuation multiple by 2-3x based on comparable company analysis."
• "Regulatory risk is the primary discount factor; isolating the research division reduces the risk discount by 40-60%."
• "The current go-to-market narrative undervalues the infrastructure platform relative to the consumer product."

Analysis — Run only the analyses that test these hypotheses. Use web search for comparable IPO data. Build a revenue decomposition to test the narrative hypothesis. Assess regulatory risk using publicly available legislative tracking. Stop when each hypothesis is supported, refuted, or marked inconclusive.

Synthesis — Apply the pyramid principle: state the governing thought first, then support with the three strongest findings.

Deliverable character

A clean, structured analysis. The output follows the pyramid principle: answer first, evidence second. No proprietary framework names, no firm-specific process language, no lens architecture or loop models. Just: hypothesis → evidence → recommendation.

Core value delivered

Analytical clarity. The board gets a well-structured recommendation with quantified evidence and explicit assumptions. It lacks the distinctive cognitive signature of a specific firm — no reframe (BCG), no decision system (Bain), no four-tier exhaustiveness (McKinsey) — but it delivers a rigorous, defensible answer.

When to use generic mode

• The user hasn't specified a firm preference and the problem doesn't strongly signal one.
• The task is narrow enough that firm process overhead would be disproportionate.
• The user wants consulting-quality thinking without firm-specific formatting.
• Speed matters more than distinctive methodology.

---

Convergence Points

The same analytical tool appears in all three architectures but serves a different purpose:

The convergence is real — all three firms will analyze competitors, benchmark against peers, and decompose revenue. But the architectural position of each analysis differs: McKinsey embeds it as a hypothesis-testing node, Bain embeds it as a decision-support module, BCG embeds it as a lens input. Same data, different purpose, different output shape.

---

The One-Sentence Summary

McKinsey gives you the answer and proves it. Bain gives you the decision system and de-risks it. BCG gives you new eyes and reframes everything.

The choice between them is itself a meta-decision: does the client need certainty, velocity, or perspective? The answer depends on which of these the client lacks most.

Decision Memo Example

A worked decision memo in Bain decision-first style. Demonstrates the full pipeline: Decision Brief → conditional hypotheses → three-scenario economics → recommendation with evidence chain → implementation.

---

User: "Our board needs to decide: build a data analytics capability internally, acquire AnalyticsCo for $280M, or enter a strategic partnership. Decision in 90 days."

Decision Brief

The Decision: How should we obtain enterprise data analytics capability: build, acquire AnalyticsCo, or partner?

Decision Criteria: 1. Time-to-capability (competitive window is 18 months) 2. Total 3-year cost (board-approved envelope: $350M) 3. Talent retention risk (analytics talent is the asset) 4. Strategic control (board requires roadmap ownership)

Decision Timeline: Board decision by Q3. LOI required by Q3+30 days if acquire.

Decision Stakeholders: CEO (decision maker), CFO (financial approval), CTO (technical validation), Board (final approval for >$200M).

Options with Conditional Hypotheses

Option A — Build ($150M over 3 years) Preferred IF: talent market allows hiring 40+ data scientists within 6 months AND competitive window exceeds 24 months AND internal IT infrastructure supports analytics workloads without $30M+ upgrade.

Option B — Acquire AnalyticsCo ($280M) Preferred IF: retention of AnalyticsCo's top 20 engineers is achievable (3-year earnout) AND integration cost remains <$40M AND regulatory approval takes <6 months.

Option C — Strategic Partnership (JV, $50M/year) Preferred IF: strategic control over product roadmap is not a hard requirement AND partner's technology roadmap aligns with our needs for 3+ years AND IP ownership terms are acceptable.

Condition Testing Results

Option A eliminated: conditions A1 and A2 both fail. Build timeline exceeds competitive window. Option C eliminated: condition C1 fails against a non-negotiable decision criterion. Option B advances: all conditions pass (B1 provisional, requires due diligence confirmation).

Three-Scenario Economics (Option B — Acquire)

Realization haircut applied: 60-80% on projected synergies given integration complexity. Even downside IRR of 11% exceeds cost of capital (9%).

Recommendation

Acquire AnalyticsCo at $280M with a 3-year earnout structure tied to engineer retention.

Evidence chain:

• Build fails on timeline (A1, A2). Partner fails on control (C1). Acquire passes all conditions.
• Base-case economics are attractive (18% IRR). Downside is survivable (11% IRR > 9% CoC).
• The binding risk is B1 (talent retention) — mitigated by earnout structure targeting top 20 engineers specifically.

Implementation — 90-Day Sprint Plan

Sprint 1 (Days 1-30): LOI execution + confirmatory due diligence. Focus: validate B1 (retention) through direct conversations with top 20 engineers. Validate B2 (integration cost) through detailed systems assessment.

Sprint 2 (Days 31-60): Retention package design + regulatory filing. Named owner: CHRO. Deliverable: individualized retention offers for top 20, signed before close.

Sprint 3 (Days 61-90): Integration planning + Day 1 readiness. Named owner: CTO. Deliverable: 100-day integration plan with technology migration sequence, org chart, and reporting lines.

---

Why This Works

• Closed-form decision (build/acquire/partner) rather than open-ended "what should we do about analytics?" — forces the analysis toward a discrete choice.
• Conditional hypotheses define what must be true for each option to win — then test those conditions directly. Every analytical workstream traces to a specific condition.
• Three-scenario economics with downside focus — the question isn't "how good could this be?" but "can we survive the worst case?" Downside IRR > cost of capital = the deal is defensible even if things go wrong.
• Realization haircut (60-80%) — sets realistic expectations rather than presenting maximum savings.
• 90-day sprints — the recommendation includes the first 90 days of execution, not just the decision. Each sprint has a named owner and measurable deliverable.

Framework Application Examples

Four cases derived from real MBB engagements. Each demonstrates a different synthesis pattern — the judgment bridge between framework outputs and recommendation.

---

Case 1: Service Model Reform — BCG Australia Post

User: "Australia Post is losing money on letter delivery as volumes decline 7-8% annually. How should we respond?"

Situation: National postal operator facing structural volume decline in letters (primary revenue source). Political pressure to maintain universal 5-day delivery service. Parcel volumes growing 8% annually but letter infrastructure dominates cost base.

Frameworks: International Benchmarking (5 comparable postal operators) + Willingness-to-Pay Analysis

Why these: Benchmarking tests what's structurally possible (external — have peers survived similar transitions?). WTP tests whether current service levels are actually valued by customers (internal — is the premise of the political debate correct?).

Day-one hypothesis: Maintaining 5-day delivery at higher prices is the right response — comparable operators have sustained similar service levels.

Benchmarking: PostNL, Royal Mail, Deutsche Post, Canada Post, and NZ Post all faced the same volume decline. Each reduced delivery frequency, shifted to parcel infrastructure, or restructured pricing. All survived. The transition is structurally feasible.

Willingness-to-Pay: Only 6% of Australians would pay $30/year to retain 5-day letter delivery. The emotional premise — that Australians deeply value current service — collapses under data.

Synthesis (contradiction): Benchmarking says "maintaining service is feasible at higher cost — peers did it." WTP says "but customers don't actually want it enough to pay." The frameworks contradict. Resolution: the binding constraint is customer value, not operational feasibility. You could maintain 5-day delivery — but you'd be preserving a service almost nobody values enough to fund.

Recommendation: Reduce to 3-day letter delivery. Reinvest savings into parcel infrastructure where growth is 8% annually and customer willingness-to-pay is high. Use international benchmarks to sequence the transition (phased over 18 months, following PostNL's proven model). Confidence: High — both lenses tested and contradiction resolved. Key risk: political backlash on service reduction despite WTP data.

Why this works: The insight — "feasible but unwanted" — could not come from either framework alone. Benchmarking without WTP recommends maintaining service at higher cost. WTP without benchmarking recommends cutting without knowing if peers survived. The synthesis product is the contradiction itself, and the resolution requires choosing which lens is binding.

---

Case 2: Pricing Review — McKinsey NDIS Australia

User: "Are disability support prices set correctly under the NDIS scheme? Providers claim prices are too low, but costs are rising."

Situation: National Disability Insurance Scheme (NDIS) with $22B annual spend. Providers reporting financial stress and some exiting. But utilization is growing and new providers are entering. Need to determine if current price levels are adequate or require adjustment.

Frameworks: Bottom-up Cost Analysis (provider economics) + Market Development Analysis (entry/exit/utilization) + International Benchmarking (comparable disability schemes)

Why these: Three independent lenses testing the same question from different angles. If all three converge, confidence is high. If they diverge, the divergence reveals the real issue.

Day-one hypothesis: Prices are broadly adequate; the issue is specific category underpricing, not systemic inadequacy.

Cost Analysis: Bottom-up modeling shows the "efficient provider" can operate at current prices with 5-8% margin. However, costs-to-serve range from under $40 to over $55 per hour across providers — the distribution matters more than the average.

Market Development: Net provider entry is positive (more entering than exiting). Utilization is growing. If prices were truly inadequate, providers would be leaving, not entering.

International Benchmarking: Prices are within range of comparable schemes in the UK and Canada. No scheme pays dramatically more for equivalent services.

Synthesis (convergence): All three lenses converge: prices are broadly adequate. But the convergence has a nuance — the "some do, some don't" distribution pattern. Efficient providers operate profitably. Inefficient providers report stress. The issue is not the price level but the variation in provider efficiency, plus specific categories (remote, complex needs) where all three lenses show stress.

Recommendation: Maintain overall price framework. Adjust 12 specific line items where all three lenses independently show inadequacy. Build a continuous market monitoring cycle (routine monitoring → focused collection → analysis → implementation) so pricing adapts to market signals rather than waiting for the next crisis. Confidence: High — three independent sources converge. Key risk: specific underpriced categories (remote, complex needs) may cause localized provider exit before monitoring cycle catches it.

Why this works: Convergence across 3 independent sources creates high confidence the overall level is right — no single disgruntled provider can argue against three converging lenses. The "efficient provider" construct allows recommending prices some providers can't currently meet, backed by empirical evidence that the benchmark is achievable. The monitoring cycle recommendation (McKinsey's "recommend the infrastructure" pattern) is as important as the price adjustment itself.

---

Case 3: Sector Decarbonization — BCG New Zealand Electricity

User: "How should New Zealand's electricity sector decarbonize while keeping energy affordable and reliable?"

Situation: NZ generates ~82% renewable electricity but needs to reach near-100% while electrifying transport and industry. Challenge: the "energy trilemma" — sustainability goals can conflict with affordability and reliability. 11 competing electricity sector participants (generators, distributors, retailers) commissioned this review jointly.

Frameworks: Energy Trilemma (equity / security / sustainability) → 5 Pathway Scenarios → Multi-criteria Evaluation

Why these: This is framework cascading — each framework feeds the next at a different level of abstraction. The trilemma defines evaluation dimensions. The pathways generate options. The evaluation narrows to a recommendation. Each layer eliminates possibilities.

Day-one hypothesis: A balanced renewable pathway can deliver decarbonization without sacrificing affordability or reliability.

Energy Trilemma: Any viable pathway must perform across all three dimensions — equity (affordable household energy), security (reliable supply including dry years), and sustainability (emissions reduction). No single-dimension optimization is acceptable.

5 Pathways: Span the option space from business-as-usual to aggressive green export powerhouse. Each is a genuinely different strategic direction with different technology choices, investment levels, and trade-offs.

Multi-criteria Evaluation: BAU fails sustainability. Two aggressive pathways fail equity (household bills rise significantly). One pathway fails security (insufficient dry-year resilience). The surviving pathway — Smart System Evolution — performs adequately across all three trilemma dimensions.

Synthesis (cascading): The trilemma eliminates single-dimension solutions. The pathways generate a comprehensive option set. The evaluation systematically narrows to one recommendation that has survived three independent filters. Critically, the winning pathway retains optionality — it doesn't commit to a single mega-project, preserving future technology choices. Investment of ~$42B is offset by declining household energy bills.

Recommendation: Pursue Smart System Evolution pathway. Phase: 2020s (ramp up renewables), 2030s (turbocharge electrification), 2040s (complete transition). The self-financing proof — household bills decline despite $42B investment — eliminates the political question of "who pays." Confidence: Medium — pathway analysis is robust but 20-year horizon introduces technology uncertainty. Key risk: dry-year resilience assumptions depend on battery cost trajectories not yet validated.

Why this works: Framework cascading creates analytical inevitability — by the time you reach the recommendation, it has survived three independent filters and feels mathematically determined. The optionality criterion (prefer paths that preserve future choices) is a distinctively BCG move — instead of predicting the best technology, choose the path that gives you the most options to adapt later.

---

Case 4: Organizational Restructuring — Bain SARS South Africa

User: "SARS needs an organizational restructuring to improve tax collection efficiency. Assess whether the proposed operating model should proceed."

Situation: South African Revenue Service (SARS), 14,000 staff. Bain proposed restructuring: consolidate individual and corporate tax into a single division (BAIT), reorganize specialized units into functional divisions (lawyers to legal, auditors to audit). Advisory Board presentation lasted <1 hour.

Frameworks: Operating Model Assessment (structure → process → capability) + International Benchmarking (IMF diagnostic tool, comparable revenue agencies)

Why these: Operating model assesses the proposed structure internally (does it consolidate functions rationally?). Benchmarking tests against international best practice (have comparable agencies adopted similar structures?).

Day-one hypothesis: The proposed functional structure improves efficiency through consolidation and specialization.

Operating Model Assessment: The proposed structure consolidates functions rationally. Single taxpayer affairs division reduces coordination overhead. Functional organization (lawyers together, auditors together) enables specialization.

International Benchmarking: Several comparable revenue agencies have adopted functional structures. The IMF's diagnostic tool rates functional organization favorably for efficiency.

Synthesis (blind spot): Both frameworks appear to support the restructuring. Both are wrong. Neither tests the critical question: does end-to-end case management visibility survive the restructuring? The existing structure gave one unit sight of an entire taxpayer case — lawyers, auditors, and specialists together could see the full picture. The proposed structure fragments this: lawyers go to legal division, auditors to audit division. No single unit has sight of a complete case anymore.

The blind spot — loss of E2E visibility — is the single factor that, if wrong, flips the entire recommendation. The frameworks converged on "yes, proceed" because they evaluated structural rationality and international precedent, but neither evaluated operational continuity of integrated case management.

Recommendation: Do not proceed until the blind spot is resolved. Add a kill condition: "Does the proposed structure preserve end-to-end case management visibility for complex matters?" If no, redesign to maintain integrated case teams within the functional model. Additionally: the diagnostic consulted 33 people out of 14,000 staff over 6 days — inadequate evidence base for a $200M restructuring. Confidence: Low on proceeding (blind spot unresolved). Key risk: restructuring without E2E visibility testing will fragment case management — the exact failure mode observed in the real outcome.

What actually happened: The restructuring proceeded without addressing the blind spot. Results: customs inspection times went from 2 days to 23 days. High Court litigation approval went from days to months. Preservation orders required 8-12 signatures instead of 4-5. Estimated revenue loss: R100 billion over 3 years.

Why this works: Two frameworks converged and both were wrong. The blind spot — what the frameworks didn't assess — was more important than everything they did. This demonstrates why synthesis must always ask: "what didn't we analyze that could flip the answer?" Framework convergence is not the same as being right.

---

Synthesis Patterns

The four cases demonstrate four synthesis archetypes:

• Convergence (NDIS): Multiple lenses agree → high confidence. State the converging insight and its nuances.
• Contradiction (Australia Post): Lenses disagree → judgment call. Identify which lens is binding and why.
• Cascading (NZ Electricity): Lenses feed each other at different abstraction levels → layered narrowing to an inevitable conclusion.
• Blind spot (SARS): Lenses agree but miss the critical factor → the most dangerous synthesis failure. Always ask: "what didn't we analyze?"

For the full synthesis methodology, see frameworks.md §3.

Issue Tree Decompositions

Three worked issue trees, each demonstrating a different tree type. See thinking.md §3 for type definitions.

---

1. Profitability Decline (Logic Tree)

User: "Our EBITDA margins dropped 3 percentage points. Find out why and how to fix it."

This is a logic tree — exploratory decomposition of an open-ended problem. Use when the problem space is broad and you need to identify which branch contains the root cause.

Problem statement: Identify the drivers of IndustrialCo's 3pp EBITDA margin decline ($45M) over the past two years and recommend actions to recover at least 200bp within 12 months.

Issue tree

Why have EBITDA margins declined 3pp ($45M)?
│
├── Revenue per unit declining? [$15M estimated impact]
│   ├── Price erosion from competitive pressure?
│   │   Hypothesis: Average selling price declined 5%+ due to 2 new
│   │   entrants in the mid-tier segment.
│   │   Signal: Win/loss data shows increased price concessions.
│   │
│   ├── Mix shift toward lower-margin products?
│   │   Hypothesis: Product mix shifted 8pp toward commodity SKUs as
│   │   premium demand softened.
│   │   Signal: Premium SKU volume flat while commodity SKUs grew 15%.
│   │
│   └── Volume discount escalation?
│       Hypothesis: Top-10 accounts renegotiated to 12%+ discounts,
│       up from 8%, eroding 2pp of realized price.
│       Signal: Contract renewal terms vs. prior year.
│
├── COGS per unit increasing? [$20M estimated impact]
│   ├── Raw material cost inflation?
│   │   Hypothesis: Resin and steel input costs rose 12% YoY,
│   │   accounting for $12M of the gap.
│   │   Signal: Commodity index vs. contract pricing.
│   │
│   ├── Manufacturing yield declining?
│   │   Hypothesis: Yield dropped from 94% to 91% after the plant
│   │   expansion, adding $5M in waste cost.
│   │   Signal: Scrap rate by production line, pre- vs. post-expansion.
│   │
│   └── Supply chain cost increasing?
│       Hypothesis: Freight costs rose 18% due to carrier consolidation,
│       adding $3M.
│       Signal: Per-unit logistics cost trend by lane.
│
└── SG&A growing faster than revenue? [$10M estimated impact]
    ├── Headcount growing faster than output?
    │   Hypothesis: Headcount grew 15% while revenue grew 4%,
    │   adding $6M in personnel cost without proportional output.
    │   Signal: Revenue per FTE trend.
    │
    └── Overhead cost inflation?
        Hypothesis: Facility and IT costs rose $4M from new office
        lease and ERP implementation, partially one-time.
        Signal: Fixed cost breakdown, recurring vs. one-time.

Validation plan

---

2. Growth Strategy (Hypothesis Tree)

User: "Our core mid-market segment is saturating. We think enterprise is the answer — validate or redirect us."

This is a hypothesis tree — the team has a day-one answer ("enterprise expansion is the best growth vector"). Each branch tests a condition that must be true for the governing hypothesis to hold. Unlike a logic tree (which explores alternatives), a hypothesis tree validates or kills a specific bet.

Problem statement: Validate whether enterprise expansion can deliver $25M+ ARR for CloudStack within 24 months, or redirect to alternative growth vectors.

Hypothesis tree

Governing hypothesis: Enterprise expansion is the highest-ROI
growth vector and can deliver $25M+ ARR within 24 months.
│
├── C1: Are product capabilities sufficient for enterprise?
│   [Must be true] 3 capability gaps (SSO, audit logging, custom
│   workflows) block 80% of enterprise deals.
│   Test: Lost-deal feature analysis + engineering build estimate.
│   Hypothesis: Gaps are closeable in 6 months at <$2M investment.
│   Signal: Lost-deal analysis citing missing features, eng estimates.
│
├── C2: Do enterprise unit economics justify the shift?
│   [Must be true] Enterprise ACV must exceed mid-market ACV enough
│   to justify longer sales cycles and higher CAC.
│   Test: Pricing research + early enterprise deal analysis.
│   Hypothesis: Enterprise ACV of $120K+ vs. mid-market $35K
│   yields CAC payback <18 months despite 2x longer sales cycle.
│   Signal: Early enterprise deal ACV, cycle length, CAC.
│
├── C3: Can the sales motion adapt to enterprise?
│   [Must be true] Enterprise requires solutions selling, not PLG.
│   Current team has zero enterprise experience.
│   Test: Pilot with 2 enterprise AEs over 6 months.
│   Hypothesis: Hiring 8 enterprise AEs generates $15M pipeline in Y1.
│   Signal: Pilot pipeline generation rate, win rate, cycle length.
│
└── Resolution:
    If C1-C3 hold → enterprise expansion validated ($20-25M ARR).
    Remaining $35M from existing market upsell ($15-20M) + adjacent
    products ($15-20M) as complementary, not alternative, vectors.
    If any Cx fails → redirect: adjacent products become the primary
    growth vector, enterprise deprioritized until conditions change.

Validation plan

---

3. Organizational Restructuring (Kill-Condition Tree)

User: "We've been asked to assess whether a $200M organizational restructuring should proceed. Give us a go/no-go framework."

This is a kill-condition tree — any single branch, if falsified, terminates the analysis. Use for binary go/no-go decisions where the cost of a wrong positive vastly exceeds a wrong negative. Branches are sequenced fastest-to-falsify first. Grounded in patterns from real restructuring failures (see Bain SARS quality benchmark).

Problem statement: Determine whether to proceed with the proposed $200M organizational restructuring of a 14,000-person government agency.

Kill-condition tree

Should we proceed with this organizational restructuring?
│
├── K1: Does a documented strategy exist that the new structure serves?
│   Test: Request strategy document from sponsor.
│   → If no strategy exists → KILL.
│   "An operating model is a strategy upon which a structure is built.
│    There was no new strategy upon which the structure was rebuilt."
│
│   *** K1 TRIGGERED — no strategy document exists. ***
│   *** Analysis stops. Remaining conditions not tested. ***
│
├── K2: Has the diagnostic consulted operational staff, not just leadership?
│   Test: Interview coverage ratio.
│   → If <10% of affected staff consulted → KILL.
│   33 interviews out of 14,000 staff is not a diagnostic.
│
├── K3: Does restructuring preserve end-to-end value chain visibility?
│   Test: Map current E2E flows for complex cases.
│   → If any critical flow loses single-owner visibility → KILL.
│   Functional reorganization (lawyers to legal, auditors to audit)
│   fragments integrated case management.
│
├── K4: Does a change management plan exist beyond executive announcement?
│   Test: Review transition plan.
│   → If plan = "announce and implement" → KILL.
│   Staff discovering their units no longer exist through HR system
│   changes is not change management.
│
└── K5: Are operational baselines measured pre-restructuring?
    Test: Request pre-restructuring metrics.
    → If no baselines for processing times, approval turnaround,
      revenue collection → KILL.
    You cannot measure restructuring impact without a baseline.
    (After: customs inspection went from 2 days to 23 days.)

Outcome

K1 triggers immediately. Recommendation: "Develop the strategy that the new structure must serve before proceeding with structural design. The restructuring is not rejected — it is premature. Address K1, then re-evaluate K2-K5 in sequence."

This saved the cost of testing K2-K5. In the real case, all five conditions were violated. Estimated impact: R100 billion in lost revenue collection over 3 years.

---

Pattern Notes

Each tree type serves a different problem structure:

• Logic tree (Tree 1 — Profitability): Exploratory. Branches are ALTERNATIVES — "which driver is causing this?" Open-ended decomposition when you don't know where the problem lives.
• Hypothesis tree (Tree 2 — Growth): Directed. Branches are CONDITIONS — "what must be true for this bet to work?" Validates a specific day-one answer. If any condition fails, the hypothesis is redirected.
• Kill-condition tree (Tree 3 — Restructuring): Binary go/no-go. Branches are DEAL-BREAKERS — any single one kills the recommendation. Sequenced fastest-to-falsify first. Saves analytical cost by stopping early.

The structural difference: logic tree branches are mutually exclusive explanations. Hypothesis tree branches are simultaneously required conditions. Kill-condition tree branches are independently sufficient to terminate.

For logic and hypothesis trees:

• Every leaf has a directional hypothesis with expected magnitude. Not "costs might be high" but "resin costs rose 12%, accounting for $12M."
• Validation is prioritized by impact. The biggest-lever hypotheses get tested first, not the easiest-to-prove ones.
• The tree sizes the problem before solving it. Estimated impact at the branch level tells you where to spend your analytical time.

For kill-condition trees:

• Every leaf has a binary pass/fail test with a clear termination criterion and specific test method.
• Validation is prioritized by speed-to-falsify. The fastest-to-test condition comes first — saving the cost of testing all subsequent branches.

Market Sizing Examples

Practical market sizing examples across three business models with triangulation and sensitivity analysis.

---

User: "Size the market for an AI contract management platform targeting US mid-market."

1. B2B SaaS — AI Contract Management Platform

Target: Mid-market companies (100-1,000 employees), U.S.

Top-Down

Bottom-Up

Triangulation

• Top-down: $322M | Bottom-up: $756M | Delta: 135%
• Diagnosis: Bottom-up counts all potential users; top-down measures current spending. Gap = unmet demand (manual processes). Apply "willingness to pay" filter: $756M × 0.45 = $340M.
• Final: TAM $2.9B | SAM $330M | SOM $22-23M

---

2. B2C — Premium Plant-Based Protein Bars

Target: Health-conscious millennials/Gen Z, U.S.

Top-Down

Bottom-Up

Triangulation

• Top-down: $217M | Bottom-up: $426M | Delta: 96%
• Diagnosis: Bottom-up includes switchers from non-plant-based. Use top-down for conservative, bottom-up for upside.

---

3. Marketplace — Freelance Graphic Design

Target: Small businesses (<50 employees), U.S.

Key: Size GMV First, Then Apply Take Rate

Always label GMV vs. revenue clearly.

---

4. Sensitivity Analysis

Identify 3-4 high-uncertainty assumptions, vary by reasonable range.

Example (B2B SaaS):

Outputs:

Interpretation: SOM ranges $7-81M (base $22.7M, matching bottom-up). Widest driver: penetration rate. Note: triangulation adjusts base SAM from $756M to $330M via WTP filter — applying that adjustment compresses the range further. Recommend validating through pilot before full launch.

---

5. Common Mistakes

---

6. Presentation Template

Header: Market title, date, confidence (H/M/L)

Section 1 — Numbers (top third):

• TAM / SAM / SOM with growth rates

Section 2 — Assumptions (middle):

• 4-6 critical assumptions: name, value, source, confidence
• Highlight lowest-confidence in color

Section 3 — Sensitivity (bottom):

• Low/Base/High SOM range
• Implication: "Base case supports $X ARR in Y years"
• Risk: "Biggest uncertainty is [X]; recommend [validation]"

Rules:

• <30 words per bullet
• Every number has source
• Consistent units ($M or $B)
• Date stamp

SCR Worked Examples

Four Situation-Complication-Resolution frames across different case types. Use these to calibrate SCR construction — match the density, specificity, and tension level.

---

User: "Write an executive summary for a European market entry recommendation."

1. Market Entry (MedTech acquisition)

Situation: MedTech Corp is a $200M medical device company with 85% US revenue, strong engineering capabilities, and deep FDA regulatory expertise.

Complication: US market growth has slowed to 3% as the market matures. Competitors are expanding internationally and gaining scale advantages that will erode MedTech's cost position. Without geographic diversification, MedTech risks losing competitive position within 3-5 years.

Resolution: Acquire a mid-size European distributor ($30-50M revenue) to leverage MedTech's existing product portfolio and regulatory capabilities, targeting $100M in European revenue within 5 years.

Why this works: S establishes the company's strengths (what they have to work with). C creates urgency with a specific timeline and names the mechanism of decline (competitor scale). R is specific enough to be debatable — a board member could argue for organic entry or partnership instead.

---

2. Cost Reduction (retail operations)

Situation: RetailCo operates 450 stores across 35 states with $2.1B in annual revenue and has invested heavily in digital transformation over the past three years.

Complication: Operating margins have declined from 8.5% to 5.2% in two years as labor and supply chain costs outpaced revenue growth. At current trajectory, the company will be EBITDA-negative within 18 months.

Resolution: Execute a three-phase cost transformation targeting $120M in annual savings: labor optimization via scheduling technology ($45M), supply chain consolidation ($50M), and portfolio rationalization of underperforming stores ($25M).

Why this works: S anchors scale and recent context. C quantifies the margin compression AND the deadline (EBITDA-negative in 18 months) — this is what drives the "why now." R breaks the number into three clear workstreams with dollar values, making it immediately actionable.

---

3. Digital Transformation (regional bank)

Situation: RegionalBank is a top-20 US bank with $50B in assets, serving 2.5M retail customers and 150K commercial clients through a 300-branch network.

Complication: Digital-first competitors have captured 35% of new account openings in RegionalBank's markets. Customer NPS is 22 vs. 55 for digital competitors. 60% of RegionalBank's technology budget is spent maintaining legacy systems, leaving insufficient capacity for innovation.

Resolution: Invest $150M over 3 years in three priorities: core banking cloud migration ($60M), mobile-first customer experience ($50M), and data analytics for personalized engagement ($40M), targeting 3.2x ROI over five years.

Why this works: S positions the bank's scale. C uses three converging data points (account loss, NPS gap, budget allocation) to build a compelling case that the status quo is untenable. R provides both the cost and the return, letting the audience immediately assess the investment case.

---

4. Growth Strategy (SaaS company)

Situation: CloudStack is a $180M ARR B2B SaaS platform with 2,400 enterprise customers, 95% gross retention, and a dominant position in the mid-market project management segment.

Complication: Organic growth has decelerated from 45% to 22% YoY as the core mid-market segment approaches saturation at ~40% penetration. Two well-funded competitors ($200M+ each) are entering from the enterprise segment with aggressive pricing. Without a new growth vector, CloudStack will fall below the 20% growth threshold that supports its current valuation multiple.

Resolution: Pursue a platform expansion strategy combining upmarket enterprise migration ($50M ARR opportunity) with adjacent product launches in resource management and financial planning ($70M ARR opportunity), targeting re-acceleration to 30%+ growth within 24 months.

Why this works: S establishes market position and the metrics that matter for SaaS (ARR, retention, penetration). C names three converging threats (saturation, competition, valuation risk) — any one alone might not demand action, but together they do. R offers a specific growth target with a timeline and two concrete vectors.

---

Pattern Notes

Across all four examples:

• S is always <2 sentences. It reminds — it doesn't educate.
• C always names a specific mechanism — not just "things are bad" but WHY things are bad and WHAT happens if nothing changes.
• R always includes a number and a timeframe. "Improve operations" is not a resolution. "$120M savings in 18 months" is.
• R is always debatable. A reasonable executive could prefer a different path — which is exactly what makes it a real recommendation rather than a truism.

Storyboard Walkthrough

A complete 20-slide deck storyboard for a market entry strategy. Demonstrates governing thought → MECE pillars → action titles → flow test.

---

User: "Build a deck storyboard for a European market entry strategy — show deck density."

Setup

Client: MedTech Corp ($200M medical devices, US-dominant) Question: Should we enter the European market, and if so, how? Governing thought: MedTech should acquire EuroDistributor for $45M to enter Europe and achieve $100M in European revenue within 5 years.

---

MECE Pillars

Four pillars, each answering a distinct question:

1. Market opportunity — Is Europe worth entering? (slides 4-6) 2. Entry mode — Acquisition vs. organic vs. partnership? (slides 7-9) 3. Target selection — Why EuroDistributor? (slides 10-12) 4. Financial case and risks — Does the math work? (slides 13-17)

Test: these are MECE (they don't overlap) and collectively they prove the governing thought. A reader who accepts all four pillars must accept the recommendation.

---

Full Storyboard

---

Why Specific Titles Work

Slide 4: "European medical device market is $12B and growing at 8% annually, 2x the US growth rate"

• Quantified ($12B, 8%). Comparative (2x US). The reader immediately knows: the market is big and growing faster than our home market. Compare to a bad title: "European Market Overview" — this tells you nothing.

Slide 8: "Organic entry requires 5+ years and $80M with high execution risk due to distribution complexity"

• This title KILLS an option. The reader doesn't need to read the slide body — the title alone eliminates organic entry from consideration. A good storyboard advances the argument in every title.

Slide 10: "EuroDistributor is the strongest target across strategic fit, financial profile, and cultural alignment"

• Names the three evaluation criteria right in the title. The reader knows the basis for selection before seeing the evidence. The next two slides provide the proof.

Slide 13: "Acquisition at $45M (6x EBITDA) with $20M revenue synergies yields 3.5x ROI over 5 years"

• The entire financial case in one sentence. An executive can read this title and know whether the deal passes their return hurdle.

Slide 19: "Three decisions needed: approve LOI, authorize $2M due diligence budget, appoint integration lead"

• Specific asks with dollar amounts. The board knows exactly what they're being asked to do. Compare to a bad title: "Next Steps" — which asks for nothing.

---

Flow Test

Read only the action titles from slide 3 through slide 19 in sequence:

We recommend acquiring EuroDistributor for $45M... The European market is $12B and growing 2x... Three segments fit best... Regulation favors us... Only acquisition meets the timeline... Organic takes too long... Acquisition is best... EuroDistributor is the strongest target... Its network covers 12 countries... Its financials are attractive... The deal yields 3.5x ROI... Revenue synergies from cross-selling... Cost synergies from consolidation... Positive NPV in most scenarios... Three risks with mitigation plans... 18-month integration plan... Three decisions needed from the board.

The story is complete. No gaps. No jumps. A partner reading only the titles understands the full recommendation and its evidence base. This flow test applies to L1 show decks where every title is assertive. In L2 working decks, the flow test applies to analytical slides only.

---

Handoff Artifact for Deck-Design

This is the format consultant produces when handing off to a delivery skill (deck-design-pdf, deck-design-ppt, etc.) for slide rendering. The delivery skill reads this artifact, selects exhibit types and visual patterns using its own taxonomy, applies firm visual identity, and builds the final output.

# Storyboard: European Market Entry Strategy for MedTech Corp

**Governing thought:** MedTech should acquire EuroDistributor for $45M
to enter Europe and achieve $100M in European revenue within 5 years.

**Density:** L1 (show deck)
> **L2 adaptation:** To convert to a working deck, use topic titles on structural
> slides (agenda, dividers, appendix). Retain action titles on analytical slides
> only. Add methodology/source slides before each pillar. See communication.md §6.
**Firm overlay:** McKinsey

## Pillar 1: Market Opportunity (slides 4-6)

### Slide 4
- **Action title:** European medical device market is $12B and growing
  at 8% annually, 2x the US growth rate
- **Content:** Compare absolute market size (US $28B vs EU $12B) and
  growth trajectories (US 4% vs EU 8% CAGR) to establish Europe as
  the higher-growth opportunity despite smaller base
- **Data:**
  | Market | Size ($B) | CAGR |
  |---|---|---|
  | US | $28 | 4% |
  | Europe | $12 | 8% |

### Slide 5
- **Action title:** Three segments offer the strongest fit: orthopedics
  ($4B), cardiology ($3B), and diagnostics ($2B)
- **Content:** Position five segments across three dimensions (market
  size, growth rate, capability fit) to show the top three are clearly
  differentiated from the rest
- **Data:**
  | Segment | Size ($B) | Growth | Fit score |
  |---|---|---|---|
  | Orthopedics | 4.0 | 10% | 9/10 |
  | Cardiology | 3.0 | 7% | 7/10 |
  | Diagnostics | 2.0 | 12% | 8/10 |
  | Neurology | 1.5 | 6% | 4/10 |
  | Other | 1.5 | 5% | 3/10 |

## Pillar 2: Entry Mode (slides 7-9)

### Slide 9
- **Action title:** Acquisition offers fastest time-to-revenue at lowest
  total cost and highest certainty
- **Content:** Evaluate three entry modes against four criteria (time,
  cost, certainty, control) to show acquisition dominates on all dimensions
- **Data:**
  | Criterion | Organic | Partnership | Acquisition |
  |---|---|---|---|
  | Time to revenue | 5+ years | 3 years | 12 months |
  | Total cost | $80M | $55M | $45M |
  | Revenue certainty | Low | Medium | High |
  | Control | Full | Shared | Full |

## Pillar 3: Target Selection (slides 10-12)

### Slide 12
- **Action title:** EuroDistributor's financial profile is attractive:
  $40M revenue, 15% EBITDA margin, growing at 12%
- **Content:** Land four headline metrics that establish EuroDistributor's
  financial health and growth trajectory
- **Data:**
  | Metric | Value | Trend |
  |---|---|---|
  | Revenue | $40M | +12% YoY |
  | EBITDA margin | 15% | Stable |
  | Customer retention | 95% | +2pp vs 3yr avg |
  | Revenue/employee | $320K | +8% YoY |

## Pillar 4: Financial Case (slides 13-17)

### Slide 13
- **Action title:** Acquisition at $45M (6x EBITDA) with $20M revenue
  synergies yields 3.5x ROI over 5 years
- **Content:** Decompose the deal economics from acquisition cost through
  each value driver to net value created, showing how components sum
  to $60M net value
- **Data:**
  | Component | Value ($M) |
  |---|---|
  | Acquisition cost | -45 |
  | Revenue synergies (5yr cumulative) | +100 |
  | Cost synergies (5yr cumulative) | +25 |
  | Integration costs | -12 |
  | Tax impact | -8 |
  | **Net value created** | **+60** |

What consultant provides vs. what the delivery skill decides:

• Consultant provides: governing thought, density, firm overlay, action titles, content descriptions (what the slide argues and proves), and structured data.
• Delivery skill decides: exhibit types (chart, table, matrix, etc.), layout composition, color palette, typography, visual identity, and all production choices.

Consultant Skill

MBB-grade strategy analysis, problem structuring, and executive deliverables.

Produces structured analysis and deliverable content (markdown). Does not produce visuals — hands off to deck-design-ppt, docx, or xlsx for production.

Capabilities

Problem structuring — MECE decomposition, issue trees, hypothesis generation, validation planning. Three decomposition architectures: McKinsey issue trees, BCG Strategic Prism lenses, Bain option maps.

Strategy analysis — Framework selection and application with synthesis. Covers market sizing, competitive positioning, financial modeling, and 8 domain specializations (pricing, M&A, due diligence, risk, customer insights, change management, KPIs, financial analysis).

Firm-specific methodology — Full engagement processes for McKinsey (4-tier linear stack), BCG (modular lens architecture), and Bain (3-loop decision spiral). Each with quality gates, named artifacts, and deck anchor specifications.

Executive deliverables — Executive summaries (SCR-framed), decision memos, deck storyboards, problem decompositions. Each with testable output contracts.

Industry adaptation — Analytical recalibration cues for healthcare, defense/government, financial services, SaaS, manufacturing, and energy. Targets specific traps where default methodology produces structurally wrong results.

Composition — Storyboard handoff to deck-design-ppt (6-field spec), document structure handoff to docx, model structure handoff to xlsx.

Architecture

Progressive disclosure — load only what the task needs.

consultant/
├── SKILL.md                          ← Layer 0: always loaded (orchestrator)
├── README.md
├── examples/                         ← Layer 4: loaded for calibration
│   ├── scr-worked.md
│   ├── storyboard-walkthrough.md
│   ├── issue-trees.md
│   ├── cross-firm-comparison.md
│   ├── framework-application.md
│   ├── market-sizing.md
│   └── anti-patterns.md
└── references/
    ├── method/                       ← Layer 1: loaded for most tasks
    │   ├── thinking.md                 analytical method (MECE, issue trees, hypothesis, synthesis)
    │   ├── communication.md            delivery method (pyramid, SCR, action titles, storyboarding)
    │   └── frameworks.md               framework selection and pairing rules
    ├── firms/                        ← Layer 2: loaded when firm mode active
    │   ├── mckinsey/
    │   │   ├── process.md              4-tier engagement workflow
    │   │   └── catalog.md              per-framework specs (on-demand lookup)
    │   ├── bcg/
    │   │   ├── process.md              Strategic Prism lens architecture
    │   │   └── catalog.md
    │   └── bain/
    │       ├── process.md              3-loop decision spiral
    │       └── catalog.md
    └── domains/                      ← Layer 3: loaded on demand per routing table
        ├── financial-analysis.md
        ├── pricing.md
        ├── customer-insights.md
        ├── risk.md
        ├── change-management.md
        ├── due-diligence.md
        ├── m-and-a.md
        ├── kpi-reference.md
        └── industry-context.md

Context budget

Layer summary

• Layer 0 — SKILL.md (376 lines). Execution algorithm, behavioral instincts, routing table, firm epistemologies, output contracts, quality gates. Always loaded.
• Layer 1 — Method references (545 lines across 3 files). How to think, communicate, and select frameworks. Loaded for most tasks.
• Layer 2 — Firm process files (~1,020 lines across 3 files). Full engagement workflows with quality gates and artifacts. One firm loaded per engagement.
• Layer 3 — Domain references (~1,200 lines across 9 files). Specialized analytical guidance per domain. Loaded on demand.
• Layer 4 — Catalogs (~950 lines) and examples (~1,030 lines). Lookup and calibration. Loaded when producing specific output types.

Status

Architecture and content are complete. Ready for live testing against real consulting tasks. Next improvement frontier is empirical — calibrate examples based on observed failure modes.

Usage

The runtime discovers skills under workspace/skills/. This skill is stored under wip-skills/, so it will not load unless linked.

From workspace root:

ln -s ../wip-skills/consultant skills/consultant

Delivery Contexts

Non-consulting argument structures. The thinking framework applies identically — these contexts change how the argument is packaged, not how it's built. Load when the deliverable targets a non-consulting audience.

---

Investor Pitch

Audience: VCs, growth investors, strategic acquirers. They decide in the first 3 slides. Traction outweighs theory.

Pillar architecture: 1. Problem + why now — why this year, not why eventually. Timing is the argument. 2. Solution + moat — what stops the second mover. "AI-powered" is not a moat, data or distribution is. 3. Traction + unit economics — show the slope, not the snapshot. Month-over-month > cumulative totals. 4. Market + expansion — wedge first, then expand. Show the entry point, not just the ceiling. 5. Ask + use of funds — tied to the next 18 months of milestones, not a general wish list.

Kill conditions:

• No traction data → don't pitch growth metrics, pivot to vision + team
• TAM < $1B → not venture-backable at scale, reframe or acknowledge
• Unit economics negative with no clear path to positive → fatal for Series B+

Quality gates:

• Does the ask match the traction stage? Seed evidence differs from Series C evidence.
• Is the moat specific to this company? "AI-powered" is not a moat.
• Can you state why NOW in one sentence — not why eventually?

Evidence standards: Traction metrics > market research. Customer quotes > survey data. Revenue trajectory > TAM arithmetic. Show the slope, not the ceiling.

---

Internal Strategy

Audience: Leadership team, board, cross-functional stakeholders. They already know the business — don't explain what they live daily.

Pillar architecture: 1. Strategic context — what changed, not company background 2. Options evaluated — show alternatives were considered 3. Recommended path — with quantified impact and trade-offs 4. Resource ask + timeline — what you need and when it pays back

Kill conditions:

• No clear decision needed → this is a status update, not a strategy presentation
• Recommendation requires data the audience has and you don't → ask first, present second

Quality gates:

• Does every slide assume the audience's expertise? No "what is our product" slides.
• Is the decision closed-form — approve X, reject Y, or choose between A and B?
• Are trade-offs explicit? Hiding downsides erodes trust with an informed audience.

---

Public / Conference

Audience: Industry peers, potential customers, media. They want one insight they can repeat, not analysis they must study.

Pillar architecture: 1. Provocative claim — the one thing the audience should remember 2. Evidence that surprises — counterintuitive data, not confirming data 3. Framework they can reuse — give them a thinking tool, not just a conclusion 4. Implication or call to action — what should they do differently tomorrow

Kill conditions:

• No surprising finding → don't present, write a blog post
• Topic requires proprietary data you can't share → reframe around public data

Quality gates:

• Can someone repeat your main point in one sentence after leaving the room?
• Would the audience learn something they didn't know walking in?
• Is the framework reusable outside your specific context?

Change Management

How to design and execute organizational change. Covers adoption models, stakeholder management, resistance, communication planning, and measurement.

---

1. When Change Management Applies

Any initiative that requires people to work differently needs change management. Technology implementations, reorganizations, process redesigns, M&A integrations, strategy pivots — the technical solution is the easy part. Getting people to adopt it is the hard part.

The failure mode is consistent: the project delivers on time and on budget, but six months later nobody uses the new system, follows the new process, or behaves according to the new strategy. Change management prevents this.

---

2. Individual Adoption Model

People move through five sequential stages (understanding → motivation → capability → action → reinforcement). Skipping a stage guarantees failure at the next one.

Consequences of skipping stages

---

3. Transformation Execution

For large-scale changes (company-wide initiatives, multi-year programs), layer organizational execution on top of individual adoption.

Five phases

1. Vision and coalition (weeks 1-4): Define the change vision in one sentence. Build a coalition of sponsors, champions, and influencers. Without a coalition, the change is one executive's initiative — and dies when that executive's attention shifts.

2. Assessment and planning (weeks 3-8): Map the current state, define the target state, identify the gaps, design the change plan. Assess organizational readiness honestly.

3. Design and pilot (weeks 6-14): Build the solution, pilot with a representative group, collect feedback, iterate. Pilots prove feasibility and create internal advocates.

4. Deploy and support (weeks 12-26): Roll out in waves, provide intensive support, monitor adoption metrics in real time. Deploy too fast and support collapses. Deploy too slow and momentum dies.

5. Sustain and optimize (ongoing): Embed the change into standard operating procedures, performance management, and organizational culture. This phase never ends — it just becomes normal operations.

---

4. Stakeholder Analysis

Resistance management

Resistance is information, not obstruction. It tells you what you haven't addressed.

The most dangerous resistance is silent. Vocal resisters can be engaged. Silent resisters comply on the surface and undermine in practice. Monitor adoption metrics, not just stated support.

---

5. Communication Planning

Principles

• Frequency beats polish. Ten short updates outperform one perfect memo. People need to hear a message 7-10 times before it registers.
• Manager channel is primary. Employees trust their direct manager more than any other source. Equip managers with talking points, answers to likely questions, and permission to be honest about uncertainty.
• Two-way, not broadcast. Build in mechanisms for questions, concerns, and feedback. Unanswered questions become rumors.

Communication sequence

---

6. Measuring Change

Adoption metrics

Leading vs. lagging

• Leading indicators: Training enrollment, communication reach, manager readiness scores, pilot feedback, early adoption rates. These predict whether the change will stick.
• Lagging indicators: Business outcomes (productivity, revenue, cost savings), employee engagement scores, customer satisfaction. These confirm whether the change delivered value.

Track leading indicators weekly during deployment. Track lagging indicators quarterly. If leading indicators are strong but lagging indicators don't follow within two quarters, the change is being adopted but the underlying thesis was wrong.

Customer Insights

How to understand customers systematically. Covers segmentation, journey mapping, jobs-to-be-done, churn analysis, and lifetime value.

---

1. Customer Segmentation

Segmentation is useful only if it changes how you act — different segments should receive different strategies.

Quality test

A useful segmentation is:

• Identifiable: you can classify a customer into a segment from data you already have.
• Substantial: each segment is large enough to warrant distinct treatment.
• Differentiable: segments respond differently to your offering, pricing, or messaging.
• Actionable: you can actually reach each segment through distinct channels or motions.

---

2. Customer Journey Mapping

Map the end-to-end experience from the customer's perspective, not your process chart.

For each stage, capture:

• Actions: What the customer does.
• Touchpoints: Where the interaction happens.
• Pain points: Where friction, confusion, or disappointment occurs.
• Moments of truth: Interactions that disproportionately shape the customer's overall perception.

Where to focus

Moments of truth drive more perception shift than the sum of ordinary interactions. A flawless onboarding experience compensates for minor product gaps. A botched support escalation destroys months of goodwill. Identify the 3-5 moments that matter most and invest there.

---

3. Jobs-to-Be-Done (JTBD)

Interview protocol

1. Start with a recent purchase or switching event. "Walk me through the last time you decided to change how you handle [problem]." 2. Map the timeline backward: first thought → passive looking → active evaluation → decision → first use. 3. Probe the push (dissatisfaction with the old way), the pull (attraction of the new way), the anxiety (fear of the new), and the habit (comfort with the old). 4. Listen for the job, not the solution. When the customer describes features, ask: "What were you trying to accomplish when you used that feature?"

Output

A JTBD map produces:

• 5-10 jobs the customer is hiring your product (or category) to do, ranked by importance and satisfaction.
• Underserved jobs: high importance, low satisfaction → opportunity.
• Overserved jobs: low importance, high satisfaction → potential for simplification or price reduction.

---

4. Churn Analysis

Diagnostic framework

Analysis sequence

1. Quantify: Logo churn rate, revenue churn rate, net revenue retention by cohort and segment. 2. Segment: Which segments churn most? Is churn concentrated or distributed? 3. Time: When in the customer lifecycle does churn peak? 4. Cause: What do churned customers cite as the reason? (Exit interviews, cancellation surveys, support history.) 5. Predict: Which behaviors predict churn 60-90 days before it happens? (Login frequency, feature adoption, support escalations, NPS score.) 6. Act: Intervene on at-risk accounts before the decision is made, not after.

Leading indicators

---

5. Customer Lifetime Value (CLV)

Calculation methods

Simple: ARPU x Gross Margin % x Average Lifespan (in years).

Cohort-based: Track actual revenue from each customer cohort over time. Plot retention curves. Extrapolate to steady state. This captures expansion revenue and churn together.

DCF-based: Sum of discounted future gross profit per customer. More precise for long-lived customers with variable expansion patterns.

LTV:CAC ratio

CAC payback period: Months to recover customer acquisition cost from gross profit. Target <18 months for enterprise, <12 months for SMB.

---

6. Win/Loss Analysis

Interview recent wins and losses (within 30 days of decision) to understand what drives buying decisions.

Interview structure

• What triggered the evaluation?
• Who was involved in the decision? (Map the buying committee.)
• What were the top 3 criteria?
• Which alternatives did they evaluate?
• Why did they choose (or not choose) us?
• What almost changed their mind?

Pattern recognition

After 20-30 interviews, patterns emerge. Aggregate into:

• Win themes: capabilities or attributes that consistently drive wins.
• Loss themes: gaps or perceptions that consistently drive losses.
• Competitive displacement patterns: which competitors take deals from you, and on what basis.

Present as a competitive win-rate matrix by competitor and segment, with the top 3 reasons for each outcome.

Financial Analysis

Consulting-specific decision rules and thresholds. Claude already knows valuation methods, cost structures, and modeling techniques.

Decision Rules

• S-curve warning: Do not default to linear extrapolation for products in early adoption phases. Model the adoption curve explicitly.
• Sensitivity test: If the recommendation changes between base and downside scenarios, the investment is too sensitive to recommend.
• Cost benchmark: Any cost category >20% above peer median is an optimization candidate. Materially below peers may signal under-investment.
• CAC boundary: Exclude customer success spend from CAC (that's retention, not acquisition). Always calculate blended and by-channel.
• "Do nothing" baseline: Every investment case needs a comparator. What happens if we don't act? Quantify the cost of inaction.

Discount Rate Calibration

When in doubt, use the higher rate. Rejecting a good project is less costly than accepting a bad one.

Technique Nudges

• MIRR over IRR when cash flows change sign multiple times or IRR produces unrealistically high numbers.
• Real-options thinking when projects have embedded flexibility (expand, abandon, defer, switch). Traditional NPV undervalues these.

For current valuation multiples and comparable transaction data, use web search rather than training knowledge.

KPI Reference

Consulting-specific selection discipline. Claude already knows KPI formulas and definitions.

Selection Discipline

• Select 2-3 KPIs per strategic priority (one leading, one lagging).
• Limit executive dashboards to 7-10 KPIs. More obscures signal with noise.
• Every KPI needs an owner and a threshold that triggers action.

Benchmark Sources

For current-year benchmarks, use web search. Preferred sources by domain:

• SaaS: KeyBanc, OpenView, Bessemer annual surveys
• General: APQC, Hackett Group, Gartner
• HR: SHRM, Mercer, BLS
• Finance: Deloitte CFO Survey, McKinsey Global Institute

Static benchmark numbers go stale — always verify against current-year data when specific numbers matter for the analysis.

Risk Assessment

Consulting-specific judgment for identifying, sizing, and communicating risk. Claude already knows risk frameworks, registers, and quantification methods — this file adds the heuristics that separate useful risk analysis from theater.

Risk Tolerance Calibration

Stated vs. revealed appetite. The CEO's stated risk tolerance and their revealed tolerance (what they actually funded, approved, or killed in the last 2 years) often diverge. Test both. The gap is diagnostic — it tells you where the organization is self-deceiving.

For specific tolerance benchmarks by industry, use web search for current-year data.

Reverse Stress Test

Start from the failure outcome ("What would cause us to breach our debt covenants?") and work backward to identify the combination of events that would get there. This surfaces risks that conventional assessment misses because no single risk alone triggers concern.

Risk Interdependency

Risks rarely travel alone. Map 2nd-order consequences: if Risk A materializes, which other risks become more likely or more severe? The combination that kills organizations is usually 2-3 moderate risks compounding, not one extreme risk in isolation. Draw the dependency map before sizing individual risks.

Quantification Discipline

Don't over-quantify low-probability risks with false precision. "$50-200M depending on [scenario trigger]" is more honest and more useful than "$127M expected loss." Reserve point estimates for risks with historical frequency data. For novel risks, use scenario ranges and name the assumptions that determine which end of the range materializes.