The Consulting Mind: Structure, Hypothesis, Answer-First

Consulting Track · Branch: Foundations (cn-foundations), ⬜ White belt cluster (Analyst rung: the belt itself arrives at the end of this two-node cluster) · ~6 focused hours · Gate: mastery quiz ≥85% + the placement diagnostic

Picture the moment before the moment. A client (or a case interviewer playing one) has just finished a two-minute problem statement: revenue is down, or a board wants to enter a new market, or a private-equity fund wants a view on a target company by Friday. The talking stops, and every eye in the room turns to you. You have no textbook to open, no dataset already cleaned, no professor waiting to confirm you're "getting warm." What you have is a messy, underspecified problem and the expectation that within sixty seconds you will start turning chaos into a plan. Most people's first instinct, trained by years of school and technical work, is to reach for more information: read everything, compute everything, rule nothing out until the evidence is overwhelming. That instinct, however well it served you in a classroom, is close to useless here, and understanding why is the entire point of what follows.

The claim you'll spend those hours testing fits in a sentence. Consulting is not a body of expert facts to memorize. It is a way of processing an unfamiliar, high-stakes problem in public, under time pressure, with incomplete information, and that way has a name, a small number of moves and a testable structure you can learn like a scale on an instrument.

There are three moves. Structure means breaking the mess into a clean, gap-free map before you touch a single number. Hypothesis means forming a specific, falsifiable best-guess answer on day one and letting it steer what you analyze, where curiosity would otherwise do the steering. Answer-first means saying your conclusion before your reasoning, every time, because that's how a busy decision-maker actually listens.

Most of the job hangs on those three.

None of the three is taught to mastery here. Each gets its own dedicated treatment later: issue trees and MECE in full, hypothesis-driven problem solving in its own right, and the complete answer-first synthesis at the crown of the foundations branch. What you get now is the mental model that makes all three make sense, so that everything downstream in the Consulting galaxy meets prepared ground and not raw confusion.

By the end you'll be able to state, from first principles, why "gather all the facts first" is backwards for this kind of problem. You'll be able to name and use the six-stage problem-solving cycle every real engagement runs on: define, structure, prioritize, analyze, synthesize, recommend. And you'll have taken a 20-item placement diagnostic that tells you which of the galaxy's early clusters (structuring, case math, frameworks, industry economics) most deserves your first hours.

A warning first, so that nothing here oversells itself. No course manufactures the judgment a partner earns from decades of live client work. That judgment compounds from experience no course can simulate. What these first hours and the 21 modules behind them build is the teachable floor: the structures and the math, the frameworks with their blind spots, the industry economics, plus the communication habits a great consultant stands on before judgment ever enters the room. Build the floor well, and judgment has somewhere solid to grow.

Learning objectives

By the end, you can:

1. State, in one sentence and then in mechanism, why an unfamiliar, high-stakes problem rewards structure-first, hypothesis-driven, answer-first thinking over exhaustive fact-gathering, and name the failure mode ("boiling the ocean") the opposite habit produces.
2. Name and correctly sequence the six-stage problem-solving cycle (define → structure → prioritize → analyze → synthesize → recommend), and state what each stage must produce and what it must not do.
3. Build a first, single-level MECE issue tree for a simple business problem, and run the two-part MECE test on it (no overlap between branches; no gap left uncovered), a preview of the full issue-tree treatment ahead.
4. State a specific, falsifiable day-one hypothesis for a business problem, and distinguish a genuine hypothesis from a hunch, an opinion, and an open question, a preview of the hypothesis-driven module ahead.
5. Explain answer-first (pyramid) communication in one paragraph (governing thought before supporting evidence), and rewrite a chronological, "bottom-up" narrative into an answer-first one, while correctly identifying that full storyline and deck construction are out of this node's scope (the consulting-specific synthesis logic gets the full treatment in this branch's crown module; deck mechanics belong to the future Presentation galaxy).
6. Distinguish strategy consulting from three things it is commonly confused with: an audit, a market-research service, and an unconstrained academic strategy exercise. Then name the MBB career ladder (Analyst → Associate/Consultant → Engagement Manager → Principal → Partner) this galaxy's belts are built on.
7. State the honest-scope boundary of the whole Consulting galaxy (what is teachable versus what only live-engagement judgment can build), and explain why that boundary does not diminish the value of drilling the teachable floor.
8. (Productivity objective.) Use an AI chat assistant to draft a first-pass issue tree, hypothesis, or answer-first check, and then personally red-team the output for MECE gaps, hidden hunches dressed as hypotheses, or a buried conclusion. AI proposes, you dispose.
9. Take the 20-item placement diagnostic, score each of its four slices, and correctly map any below-threshold slice to the specific module cluster that will close the gap.

A note on how this galaxy's gates work, stated once here. Every module in the Consulting galaxy carries the same completion rule: a mastery quiz at ≥85% (crowns at ≥90%), plus a mapped "challenge": a Case-Engine case, a drill set, or, for this one foundational node, the placement diagnostic below. Credits and entitlements (where the app enables them) only ever unlock access; they never move a threshold, and no galaxy gates another: you can start here, in the Excel track, or in Finance, on day one, in any order.

Prerequisites & connections

Builds on. Almost nothing. Like every galaxy's first node, CN0.01 unlocks on day one with no prerequisite from any other galaxy or branch. You need no prior consulting exposure and no finance background. You don't need to code either. Ordinary comfort with percentages and ratios, plus quick estimation, will help on the case-math slice of the placement diagnostic below. If you have already worked through any of the Finance galaxy's early material, such as M0.01's five-questions orientation, M4.01's business-model anatomy or M4.03's Porter's-five-forces treatment, some of the vocabulary will be familiar. That overlap is a bonus. It's never a requirement, and everything you need is re-derived here from first principles whatever you bring.

Feeds forward, within the branch. CN0.02 (The Engagement: Proposal to Trusted Advisor) is the cluster's belt node. It takes the mind installed here and applies it across a full live engagement, end to end: proposal and SOW basics, scope-creep management, team roles and delegation, and the trust equation that governs the client relationship. You cannot skip CN0.02 to reach the belt. CN0.01 only orients you.

Feeds forward, across the galaxy. CN1.01 (Issue Trees & MECE) gives the "structure" move sketched here its full depth: the four cut types (algebraic, process, segment, stakeholder) plus the habit of testing a tree until it has no gap left. CN1.02 (Hypothesis-Driven Problem Solving) does the same for "hypothesis", with driver trees and the minimum killing analysis, then the one-day answer and the "ghost deck." CN1.03, the branch's crown (The "So What": Synthesis & Answer-First Storyline), does it for "answer-first". The governing-thought-to-support pyramid gets built and graded there, and the short preview here hands off to it. Downstream of structuring, CN2.01 (Case Math Under Pressure) and CN2.02 (Market Sizing & Guesstimates) sharpen the quantitative slice the diagnostic samples. CN3.01/CN3.02 (strategy frameworks) sharpen the framework-recognition slice, and CN4.01/CN4.02 (industry knowledge) sharpen the industry-economics slice.

You may have already completed the Cycle-2 elective E11.03 (The Consulting Case-Interview Lab) and its case CL-11.3 ("Kite Foods Profitability"). If so, treat what follows as the doorway into the deep kitchen that taster promised. E11.03 compresses a six-beat arc into a 30-minute interview simulation. The Consulting galaxy gives the same underlying moves room to breathe across 22 modules and a full Case Engine.

Cross-galaxy: cited, never re-taught. The Finance galaxy owns material the Consulting galaxy leans on and does not repeat. M4.01 (Business Model Anatomy) and M4.02 (Unit Economics) underpin the "how does this business actually make money" instinct every case draws on. You'll see their vocabulary (contribution, CAC, LTV) reused directly in Worked Example 4 below. M4.03 (Industry Structure: Porter & Value Chains) and M4.04/M4.05 (the moat modules on Greenwald and 7 Powers) are the finance-side treatment of the same competitive-structure ideas CN3 will apply from the advisory seat. M3.09 (Modeling II: DCF, Comps, LBO & M&A Basics), M3.10 (Special Situations) and the elective E11.02 (M&A, Deals & Special Situations Lab) own the deal modeling that CN6 will apply from the deal-rationale and synergy side. The finance galaxy builds the model, and the consulting galaxy builds the advisory judgment on top of it. M10.03 (The Communication Toolkit: Making the Analysis Change a Decision) is the finance corpus's own Minto-pyramid module, and CN1.03 applies the identical answer-first logic to a consulting recommendation without re-deriving it from scratch. And M5.01–M5.10, the seventeen finance sector playbooks, are the depth CN4 will point to once you reach industry knowledge in earnest, so it does not duplicate them.

Placement diagnostic

This is a placement instrument. It isn't a gate, and it never blocks you from any module, because under the rule stated above a galaxy never gates itself shut against a learner who wants to jump ahead. It gives you an evidence-based answer to "where should my first ten hours in the Consulting galaxy go?" It has four slices, sized to the galaxy's own early structure. There are 6 items on structuring (maps to CN1) and 6 on case math and sizing (maps to CN2), then 4 on framework recognition (maps to CN3) and 4 on industry economics (maps to CN4). Take it closed-book, in one sitting, in roughly 25 minutes, before you read the Core teaching content below. The point is to measure what you already carry in, so nothing you are about to learn should help you. Then read the Core content whatever your scores. The diagnostic tells you where to spend extra time. It never tells you to skip the core.

Section A: Structuring (S1–S6)

S1. A retailer splits its revenue problem into three buckets: (a) revenue from returning customers, (b) revenue from new customers, and (c) revenue from customers who bought via a promotion. A colleague claims this split is MECE. Is it, and if not, what is wrong with it?

S2. A hospital is investigating total patient wait time. Which of the following is a process cut, as opposed to a segment or algebraic cut? (a) wait time by insurance type (b) wait time = time-to-triage + time-to-bed + time-to-physician + time-to-discharge (c) wait time by day of week (d) wait time by hospital wing

S3. A telecom buckets churned customers into: (1) churned due to price, (2) churned due to poor network quality, (3) churned due to a competitor's offer. A colleague claims this is MECE. Identify the flaw (there is more than one).

S4. A firm's profit fell 20% year over year. Which single equation below is the correct algebraic cut to open the decomposition, true by definition and resting on no assumption? (a) Profit = Revenue − Cost (b) Profit = Market share × Industry growth (c) Profit = Customer satisfaction × Price (d) Profit = Employee count × Productivity

S5. A university's declining-applications problem is broken into four views: prospective students, current students, alumni, and faculty. What kind of MECE cut is this, and what is it good for (and not good for)?

S6. A retailer's cost base is split into rent, wages, and cost of goods sold, and a consultant claims this collectively exhausts all costs. What, at minimum, is missing?

Section B: Case math & sizing (M1–M6)

M1. A product sells for ₹250 with a variable cost of ₹150 per unit. Fixed costs run ₹40,00,000 per year. What is the breakeven volume in units?

M2. Revenue grew from $100M to $161.1M over exactly 5 years, compounding steadily each year. Which of the following is closest to the annual growth rate (CAGR)? (a) 6% (b) 8% (c) 10% (d) 12%

M3. A portfolio holds 60% of assets earning 8% and 40% earning 14%. What is the weighted-average return?

M4. A case gives you: Revenue ₹12 crore, COGS ₹7.2 crore. What is the gross margin, as a percentage?

M5. Estimate 39 × 41 in your head, in under three seconds, using a shortcut rather than long multiplication. State the product and the shortcut you used.

M6. A market-sizing guesstimate needs the number of coffee-drinking adults in a country of 340 million people, of whom 78% are adults, of whom 62% drink coffee daily. To the nearest 10 million, how many coffee-drinking adults are there?

Section C: Framework recognition (F1–F4)

F1. A private-equity fund wants to know whether a niche industrial-parts market is structurally attractive before it bids on a target. Which framework is the most directly useful entry point? (a) BCG Growth-Share Matrix (b) Porter's Five Forces (c) McKinsey 7S (d) Blue Ocean Strategy

F2. A conglomerate must decide which of nine business units to fund, hold, or divest. Which framework is the natural first cut? (a) Porter's Five Forces (b) BCG Growth-Share Matrix (c) Jobs-to-Be-Done (d) Ansoff Matrix

F3. An associate opens every case, regardless of the actual question, with a full hour-long Five Forces workup, even when the client's real question is "why did our defect rate spike last quarter?" Name the meta-skill failure this illustrates, in a phrase.

F4. A subscription software company must decide between expanding into a new customer segment and deepening penetration in its current one. Which classic 2×2 is built for this question? (a) SWOT (b) Ansoff Matrix (c) McKinsey 7S (d) Porter's value chain

Section D: Industry economics (I1–I4)

I1. Which cost structure most predicts that a business will compete primarily on utilization and capacity rather than on per-unit input cost? (a) high fixed cost, low variable cost (b) low fixed cost, high variable cost (c) roughly equal fixed and variable cost (d) no meaningful fixed cost

I2. A regulated utility earns an allowed return on a regulated asset base. Which of the following most threatens its long-run economics? (a) a rate-setting regulator lowering the allowed return (b) a new entrant undercutting price in an unregulated adjacent market (c) currency depreciation (d) seasonal demand swings

I3. In a two-sided marketplace (for example, a ride-hailing platform), which KPI most directly signals that network effects are strengthening as the platform scales? (a) gross merchandise value alone (b) liquidity (the match rate or time-to-match) trending up on both sides (c) total employee headcount (d) average office-lease cost per employee

I4. Which pattern below is the classic "unbundling" disruption vector: a previously bundled full-service offering pulled apart by narrower, cheaper, digitally native entrants? (a) full-service brokerage versus discount and robo-advisory platforms (b) domestic steel versus import competition (c) commercial real estate versus interest-rate cycles (d) oil refining versus crude-price swings

Answer key & explanations

S1: Not MECE (overlap). A new customer's first purchase can also be a promotional purchase, so bucket (c) overlaps both (a) and (b). The categories are not mutually exclusive.

S2: (b). It decomposes the total along the sequential steps a patient actually passes through, which makes it a process (value-chain) cut. (a), (c) and (d) are all segment cuts. They slice the identical total by a different attribute without describing any sequence.

S3: Overlap and a gap. The overlap: a customer can leave for a competitor whose offer is both cheaper and has better network quality, so a single churn event can sit in more than one bucket. The gap: reasons such as poor customer service or relocation are not covered by any of the three buckets.

S4: (a). Profit = Revenue − Cost is true by definition for any accounting period. It is an identity and makes no claim about what drives the business. (b), (c) and (d) are all candidate drivers. Each is a plausible but unproven relationship, which is a different and later kind of statement (see Section B of Core teaching content, "the moment a cut becomes a hypothesis").

S5: A stakeholder cut. It segments the problem by who is affected or who holds a relevant view, where an algebraic or process cut would use a formula or a sequence. Stakeholder cuts are excellent for surfacing whose perspective is missing from a discussion. They are poor at proving a tree is quantitatively exhaustive, because "the four stakeholders" is an assertion that nobody derived.

S6: At minimum, an "other operating expenses" bucket. Marketing, utilities, depreciation, corporate overhead, interest and taxes are all missing, at least. A cost tree of nothing but rent and wages plus COGS is not collectively exhaustive. A MECE cost tree needs an explicit catch-all bucket or full enumeration, so no real cost can fall through.

M1: 40,000 units. Breakeven = Fixed costs ÷ contribution per unit = ₹40,00,000 ÷ (₹250 − ₹150) = ₹40,00,000 ÷ ₹100 = 40,000 units.

M2: (c) 10%. $100M growing at 10% a year for 5 years is $100M × 1.10⁵ = $100M × 1.61051 = $161.05M ≈ $161.1M. That's a clean, checkable compounding fact worth memorizing as a reference point. At 10% a sum doubles roughly every 7.2 years, the "rule of 72".

M3: 10.4%. Weighted average = (0.60 × 8%) + (0.40 × 14%) = 4.8% + 5.6% = 10.4%.

M4: 40%. Gross margin = (Revenue − COGS) ÷ Revenue = (₹12 cr − ₹7.2 cr) ÷ ₹12 cr = ₹4.8 cr ÷ ₹12 cr = 0.40 = 40%.

M5: 1,599, via difference of squares. 39 × 41 = (40 − 1)(40 + 1) = 40² − 1² = 1,600 − 1 = 1,599. This "round to the nearest ten, then correct" shortcut gets drilled to reflex in the case-math module ahead.

M6: Approximately 160 million (accept 150–170 million as a reasonable estimation band). 340M × 0.78 = 265.2M adults, and 265.2M × 0.62 = 164.4M coffee-drinking adults, which rounds to 160 million to the nearest 10 million.

F1: (b) Porter's Five Forces. It directly assesses structural industry attractiveness: rivalry, buyer and supplier power, threat of entry, threat of substitutes. BCG needs a portfolio of businesses to place on its matrix, and this is one single-market question. 7S diagnoses internal organizational alignment, which is a separate matter from market attractiveness. Blue Ocean asks about creating uncontested space, a different question entirely.

F2: (b) BCG Growth-Share Matrix. It is built for allocating capital across a portfolio of businesses by growth rate and relative market share: fund, hold, or divest.

F3: "Framework-as-theater" (also acceptable: template-forcing, or reciting instead of diagnosing). The associate is reaching for a famous framework by reflex before diagnosing what the client's question requires. That failure gets full treatment later, with refusal drilled to a reflex, in the meta-skill module on when frameworks mislead, built on Rumelt's diagnose-before-prescribe principle.

F4: (b) Ansoff Matrix. Its two axes (new versus existing markets and new versus existing products) are built for the "penetrate deeper or expand into a new segment" decision.

I1: (a) High fixed cost, low variable cost. When most cost is fixed, as it is for airlines or telecom towers or hospitals, profitability hinges on filling capacity, since each incremental unit costs little once the fixed base is covered. A low-fixed-cost business such as a staffing agency competes on controlling per-unit cost instead.

I2: (a) A rate-setting regulator lowering the allowed return. For a business whose entire return is defined by a regulatory formula, that formula is the dominant structural risk. It is far more central than competitive entry, since rate-regulated utilities are typically protected natural monopolies. Currency matters only if inputs or debt are foreign-denominated, and seasonality is an operational matter that leaves the structure untouched.

I3: (b) Liquidity, the match rate or time-to-match, trending up on both sides. Network effects in a two-sided marketplace show up as the marketplace getting better and faster at matching supply with demand as it scales. Raw GMV can rise from price alone. Headcount or lease cost say nothing about the network itself.

I4: (a) Full-service brokerage versus discount and robo-advisory platforms. This is the textbook unbundling vector, a bundled, full-service offering pulled apart by narrower, cheaper, digitally native entrants. The same vector recurs in media and telecom. (b) and (d) are cost/price-cycle dynamics and (c) is a macro-rate dynamic, so none of them is an unbundling vector.

Scoring & what to do next

Score each section separately. There's no single combined score, because the four slices measure four different things and a strength in one says nothing about the others.

Clearing the strong bar on all four slices does not exempt you from CN1–CN4, since a diagnostic never bypasses a gate. It does let you move through those clusters' early material briskly and spend your first real hours on the harder crowns instead. Falling short on two or more slices doesn't mean you're behind. The galaxy assumes nothing about your starting point, and the diagnostic has done its one job. With a key you can check yourself against, it has told you where your first hours will pay off most.

Core teaching content

The trap of "gather all the facts first"

Nearly everyone arrives at their first real business problem carrying an instinct built by school and, often, by technical training: understand everything before you conclude anything. It's a good instinct in a domain with no clock and no cost to being thorough, such as a doctoral thesis or a from-scratch engineering proof. A forensic audit with a mandate to be exhaustive is another. It's close to disastrous in a consulting engagement, and the reason is structural. Taste has nothing to do with it. A real engagement runs on a fixed, short clock of days or weeks, almost never the months an academic study would take. It runs on a fixed budget, because every hour you spend is a real cost someone is paying. And it answers to an audience that must act on your answer, which is more than admiring your rigor. Under those three constraints, "analyze everything before concluding anything" only looks careful. It guarantees that you either miss the deadline or deliver a technically complete answer too late to change the decision it was meant to inform.

Consultants have a blunt name for the failure this produces: boiling the ocean. You chase every possible thread and build every model that could be built, because no thread has yet been ruled out. On day twelve of a three-week engagement, the client's question is still sitting there, un-analyzed.

You've probably seen it happen. Maybe you've done it.

The consulting mind inverts the instinct. Of all the habits the Consulting galaxy asks you to build, this inversion is the hardest, because the first few times it feels like intellectual recklessness. Where the old instinct runs analyze first, conclude later, the consulting mind runs form a specific best-guess answer immediately, then analyze only what would change it. That's neither guessing nor a shortcut around rigor. It points rigor at the one question that matters under a deadline: which two or three pieces of evidence, if I had them today, would most change my recommendation? Everything else is waste, because it doesn't move the answer. It can be as interesting as you like, and it can make the analysis feel more "complete". It's still waste.

Conn and McLean's Bulletproof Problem Solving names this the difference between an "exhaustive" and a "focused" approach. Barbara Minto's entire Pyramid Principle exists because the inversion has a second half. Once you have the answer, you must also deliver it in reverse order from how you found it, conclusion first.

Hold that discomfort for now. The rest of the core content is there to dissolve it.

The six-stage problem-solving cycle

Every real engagement, a $2M six-week strategy project and a 25-minute case interview alike, runs on the same underlying cycle. It just gets compressed or stretched to fit the clock available. The Consulting galaxy names it in six stages. Learn them cold, because every module ahead assumes you can place any activity you're doing onto this map:

Before you meet each stage in depth, notice two things about the cycle. First, it's iterative and not strictly linear. Analysis at stage 4 routinely kills your hypothesis and sends you back to stage 3 or even stage 2. A good consultant treats a killed hypothesis as progress, because eliminating a wrong answer is how the true one gets found faster.

Second, if you skip a stage or reorder it, you get a specific failure you can name. Skip "define" and you solve the wrong problem elegantly. Skip "prioritize" and you boil the ocean. Skip "synthesize" and you produce a data dump dressed up as an answer. Every item in "Common mistakes" below is, underneath its wording, one of these six stages done out of order or skipped outright.

Move 1: Structure: MECE and the issue tree, a first taste

The first move, structure, answers a single question. Before you analyze anything, can you draw a map of every piece this problem breaks into, with no piece counted twice and no piece left out?

The test for a good map has a name borrowed from McKinsey's internal vocabulary and popularized by Barbara Minto: MECE, Mutually Exclusive, Collectively Exhaustive.

Mutually exclusive means no two branches of your tree can both be true of the same fact. A single dollar of cost belongs in one branch and never two. So does a single churned customer, or a single unit sold. Collectively exhaustive means the branches, taken together, cover every possibility, so nothing real can fall outside all of them. Section A of the placement diagnostic above tested this. S1 was an overlap failure, S3 was both an overlap and a gap, and S6 was a pure gap.

The simplest and most reliable starting tree in business is an algebraic cut, a decomposition that is true by definition and needs no assumption. Profit = Revenue − Cost is the canonical example. It can't be false, because it's how profit is defined, and you can keep decomposing along the same logic (Revenue = Price × Volume, Cost = Fixed + Variable) for as many levels as the problem needs. Three other cut types round out the toolkit ahead. A process cut breaks a problem along the sequential steps something actually passes through, as in S2's hospital wait-time example, where a patient goes from triage to a bed. From there it's a physician and finally discharge. A segment cut slices an existing total by an attribute such as geography or product line or customer type. It implies no sequence or formula. A stakeholder cut breaks a problem by whose perspective is affected, as in S5's university example.

None of the four is universally "correct." The skill, graded in full later, is picking the cut that matches the shape of the question in front of you. Then you test any tree you produce against the two-part MECE check before you trust it with a single hour of analysis.

{
  "skill": "Getting a fast first-pass MECE issue tree before you've fully thought a new problem through",
  "use": "Have a chat assistant draft a candidate issue tree for a business problem in seconds, so your own time goes to the higher-value work: red-teaming the draft for overlap and gaps, and re-ordering its branches by what is actually worth investigating first.",
  "tools": ["Chat assistants — Claude, ChatGPT, Gemini"],
  "prompt": "I'm structuring this business problem: <PROBLEM>. Draft a single-level MECE issue tree with 3-5 branches, using an algebraic, process, segment, or stakeholder cut. Name which cut you used and why. Do not analyze or recommend anything yet — structure only, no numbers, no conclusions.",
  "verify": "An AI-drafted tree can look clean and still hide a real overlap or a real gap — before you trust it, run the two-part MECE test yourself: could a single real fact belong to two branches at once? Is there a plausible case that fits none of them? Also check that each branch is genuinely a category (an identity or a value-neutral slice), not a disguised hypothesis about the cause dressed up as a category.",
  "diy": "You must be able to build and MECE-test a tree by hand with no tool at all — that is the actual skill CN1.01 grades, and it is what lets you sketch a defensible tree on a blank page in a live client meeting where no AI assistant is in the room.",
  "market": "IN"
}

Move 2: Hypothesis: the day-one answer that steers the work

The second move answers a different question. Given your structure, where should you actually spend your scarce analysis hours? The consulting answer is hypothesis-driven problem solving. On day one, before you have proof, you form your most specific best guess at the answer. Then you use that guess to decide what to analyze next.

A hypothesis is a very particular kind of statement, and it's worth separating sharply from three things people routinely mistake for one. A hunch ("something feels off about our marketing spend") is too vague to test. It names a feeling and makes no claim. An opinion ("we should just cut costs") is a recommendation wearing a hypothesis's clothes, because it has skipped straight to stage 6 without passing through stages 2–5. A question ("why did margins fall?") is the thing a hypothesis exists to answer. A real hypothesis is a specific, falsifiable, testable claim, such as "margins fell because input-cost inflation was only partially passed through in price." You can imagine a specific piece of evidence proving it wrong, and it names that evidence in advance.

Hypothesis-driven work also inverts the ordinary meaning of "rigor." An exhaustive analyst treats rigor as coverage. A hypothesis-driven consultant treats rigor as discrimination. Did the analysis have the power to confirm or kill your specific claim, and did you stop the instant the answer was clear? Here is the seed of what the hypothesis work ahead calls the minimum killing analysis. Ask "what is the cheapest, fastest test that could prove this false?" and run that first. A hypothesis that survives its own best attempt at disproof is worth far more than one propped up by a mountain of non-discriminating analysis. The same later material introduces the one-day answer. That's a best guess labeled as such, defensible on day one and refined as evidence arrives. It also introduces the ghost deck, slide titles sketched before the analysis exists to force clarity about what each piece is for. Both get full teaching treatment there. For now, hold only the principle: a specific, falsifiable guess, tested by the cheapest disconfirming evidence you can find, beats an open-ended search for the truth every time the clock is real.

{
  "skill": "Stress-testing a day-one hypothesis for the cheapest disconfirming test",
  "use": "Ask a chat assistant to propose the fastest, cheapest piece of evidence that would kill your hypothesis if it's wrong — sharpening the 'minimum killing analysis' habit before you spend real time or client goodwill chasing exhaustive proof.",
  "tools": ["Chat assistants — Claude, ChatGPT, Gemini", "Research/agentic tools — for a first pass at reference industry benchmarks, always verified"],
  "prompt": "My day-one hypothesis is: <HYPOTHESIS>. What is the single cheapest, fastest piece of evidence I could gather that would most convincingly DISPROVE this hypothesis if it's wrong? List two candidates, ranked by cost/speed to obtain.",
  "verify": "Sanity-check any suggestion against what you actually know about the client's data — an AI has no idea what your engagement can realistically pull this week; you choose the real minimum-killing analysis, and verify any external benchmark it cites against a primary source before it enters a client deliverable (Guardrail 1-5, AI0.01).",
  "diy": "You must be able to state your own hypothesis and design your own kill-test unaided — CN1.02 drills this to a reflex, and a real client meeting will not pause for you to open a chatbot.",
  "market": "IN"
}

Move 3: Answer-first: pyramid logic, and why it is not arrogance

The third move governs communication, and it's the one most people resist hardest, because at first glance it looks like presumption: say your conclusion before you explain how you got there. The instinct to reason chronologically ("first we looked at X, then we found Y, which led us to consider Z, which is why we now believe...") is a natural way to think. It's also the wrong way to communicate to a senior, time-poor audience.

Barbara Minto's Pyramid Principle is the foundational text behind the whole move, and its full application arrives at the crown of the foundations branch. Minto makes the case from how listening actually works. A senior audience's patience for an unlabeled build-up is measured in tens of seconds, not minutes. A listener who doesn't yet know your conclusion has to hold every fact you mention in working memory, guessing at its relevance, until you finally reveal why it mattered. The build-up costs the speaker time, and it costs the listener more.

The alternative is the pyramid. State the governing thought, your single answer-first conclusion, in the very first sentence, so it gives every following fact its meaning before you say it. Beneath the governing thought sit two or three supporting points. Each is independently defensible, and each is a reason the governing thought is true, which is more than being "related" to it. Beneath each support sits the specific evidence. Read top to bottom, the pyramid answers the question immediately. Read bottom to top, it shows how the conclusion was earned. The content is the same, organized for the reader's benefit in place of the writer's process.

Some of this is deliberately left for later. Building a full multi-level pyramid from a messy set of findings comes afterwards. So does storylining a full deck, along with the visual/design craft of a slide. That full construction and its grading rubric belong to the crown of the foundations branch, and deck craft belongs to the future Presentation galaxy. For now, hold only the reflex: conclusion first, then the reasons, then risks and a next step. You'll recognize it the moment you meet it built in full.

{
  "skill": "Checking whether a draft recommendation is genuinely answer-first before you send it",
  "use": "Paste a drafted recommendation paragraph into a chat assistant and ask it to identify whether the very first sentence states the conclusion, or whether the conclusion is buried after background and reasoning — a fast outside read of a habit that is genuinely hard to catch in your own writing.",
  "tools": ["Chat assistants — Claude, ChatGPT, Gemini"],
  "prompt": "Here is a recommendation paragraph: <TEXT>. Does the FIRST sentence state the actual recommendation, or does the conclusion only appear later, after background and reasoning? Quote the sentence you would move to the front, and say in one line why a busy reader would get lost before reaching it as currently written.",
  "verify": "The model's read of 'answer-first' is a useful first opinion, not a certified grade — the real test is Minto's own pyramid check (does the governing thought alone answer the question, with every support genuinely and only defending it, nothing more, nothing less?), and you must apply that test yourself; do not let an AI's approval substitute for your own structural check.",
  "diy": "You must be able to write a governing thought cold, from your own analysis, before you ask anything to check it — CN1.03 grades this by hand, with no tool allowed in the room, because a real client meeting will not pause for you to open a chat window.",
  "market": "US"
}

What strategy consulting is: and three things it is often confused with

With the three moves on the table, it's time to define the activity itself. A surprising amount of confusion about the Consulting galaxy traces back to an unclear picture of the job.

Strategy consulting is the practice of helping an organization make a high-stakes, ambiguous decision it cannot confidently make alone. It structures the problem and then forms and tests a hypothesis about the answer. Then it delivers an answer-first, actionable recommendation, on a fixed timeline, for a fee, all without the consultant then doing the organization's ongoing operating work. Every clause in that definition matters, and each rules out one common confusion.

It is not an audit. An audit's objective is conformance. Looking backward, did the numbers and the controls correctly follow an agreed standard? A consulting engagement's objective is a forward decision. Given an uncertain future, what should the organization do? An auditor who found everything compliant has succeeded. A consultant who found "everything is fine, no change needed" has usually failed to find the lever the client is paying to discover.

It is not market research. A market-research deliverable is typically data and description, such as survey results or a market-size estimate or a competitive landscape, handed over for someone else to interpret. A consulting deliverable is a recommendation with an owner's name on it, carried all the way through synthesis to an answer-first "here is what you should do, and why." Market research is frequently an input to a consulting engagement, and Worked Example 2 leans on this kind of data. It's rarely what the engagement delivers.

It is not an unconstrained academic exercise. An academic case study is judged by the completeness and rigor of the analysis, on a timeline of months with no real capital at stake. A consulting engagement is judged by whether the client made a better decision than they would have without you, on a timeline of days or weeks. Real money and jobs ride on the answer. The same frameworks (Porter, BCG, Ansoff) appear in both worlds. What separates them is where you stop. Consulting stops at "good enough, on time, defensible", and academia keeps going to "exhaustive, eventually".

The job runs on a small number of teachable moves, where you might have expected an ever-growing body of facts. So the industry organizes its people around a career ladder that measures growing responsibility for those moves, and the galaxy's belts are built directly on it:

• Analyst, the White belt: structure and analyze what you're told to.
• Associate/Consultant, Orange and then Green: own a full workstream and cast frameworks correctly.
• Engagement Manager, Purple: run the whole cycle across a team.
• Principal, Brown: own the deal-and-portfolio judgment.
• Partner, Black: own the client relationship and the engagement itself, end to end.

The very next node begins that ladder's substance with the engagement itself, and the galaxy's 22 modules walk you up every rung.

The consulting mind versus other minds

To feel the difference, watch three trained, capable minds meet the identical problem ("why did Q3 revenue miss its target?") and diverge immediately. None of them is smarter. Each is optimizing for something different.

None of the three minds is wrong in its own habitat. A rushed, unreplicated academic finding is a real failure, and an engineering fix shipped without proving root cause is a real liability. The point is narrower.

Each mind is calibrated to its own constraints.

Import the academic mind's "coverage" standard, or the engineering mind's "prove root cause exhaustively" standard, into a three-week engagement with a real and fixed clock. You get the boiling-the-ocean failure from the opening of this page. Recognizing which mind a situation calls for, and consciously switching into the consulting mind when the clock and budget say so, is itself a skill the galaxy builds one module at a time.

What the Consulting galaxy can and cannot teach. A great consultant's judgment is compounded from hundreds of live engagements with real clients and real stakes. Add years of pattern recognition across industries. No course can manufacture that experience, and the galaxy doesn't pretend otherwise anywhere in its 22 modules. Three things are teachable, and the galaxy commits to teaching them to the deepest bar the medium allows. The first is the thinking. It covers structure and hypothesis and answer-first, plus the meta-skill of knowing when a famous framework would mislead you. The second is the mechanics: case math, market sizing, the standard case formats, business and industry economics. Deal and portfolio logic belong here too, and so do operating-model and change design. The third is the craft of leading an engagement and a client relationship. That covers proposals and scope and team roles. It also covers the trusted-advisor relationship and how you run a working session or a steering committee. All of it is drilled with relentless repetition against realistic, India-and-US, synthetic-by-default cases. What you build here is the skill floor, the scaffolding that real-world judgment later compounds on top of. It doesn't replace the compounding itself. Keep that boundary in view, and every hour you spend here is worth what it claims to be worth. No more, and no less.

Where this leaves you

Hold the cycle and the three moves together and the rest of the galaxy becomes legible before you have opened a single downstream module.

You now know that an unfamiliar, high-stakes problem is processed in six named stages (define, structure, prioritize, analyze, synthesize, recommend). Each has its own job and its own failure mode when skipped. "Structure" means a MECE issue tree tested for overlap and gaps, which is more than a list of categories that felt natural. "Hypothesis" means a specific, falsifiable day-one guess tested by the cheapest disconfirming evidence, and a hunch or a pre-baked opinion doesn't qualify. "Answer-first" means the governing thought comes in the first sentence, every time, because that's how a real audience actually listens.

You also know what the galaxy can and cannot give you. It gives you the floor, and the judgment that stands on it has to come from experience. Everything from here forward applies these ideas in full depth. You'll take one move at a time, then one industry and one deal-type at a time.

Common mistakes & how experts think differently

1. "I need to see all the data before I can say anything." That's the exhaustive-analysis instinct in its purest form, and it runs backwards under a real clock. The expert forms a specific and falsifiable day-one hypothesis, then analyzes only what would change it. A defensible directional answer delivered on time beats a complete one delivered too late to matter.
2. "A good structure is whatever categories come to mind first." The categories that come to mind first frequently overlap. They're often incomplete too, and sometimes both. S1 and S3 in the placement diagnostic above are real-shaped examples. The expert never builds on a tree without running the two-part MECE test first. Could one fact sit in two branches? Is there a case that fits none?
3. "I'll walk them through my reasoning chronologically so they can follow how I got there." Among newcomers this is the commonest presentation mistake of all. It optimizes for the speaker's comfort at the listener's expense. The expert states the governing thought first, because a senior audience's patience for an unlabeled build-up is measured in tens of seconds.
4. "Reciting a famous framework proves I know strategy." A memorized Five Forces workup, produced on reflex whatever the question (F3 in the diagnostic), is framework-as-theater. It isn't analysis. Rumelt's core objection in Good Strategy/Bad Strategy is this very habit, and his remedy is diagnosis before prescription. The expert diagnoses the question first, sometimes building a bespoke structure instead of forcing a template. The meta-skill gets full treatment later.
5. "If my structure and hypothesis are good, the math doesn't need to be exact." Sloppy arithmetic under time pressure destroys credibility fast, however good the structure was. A client who catches one wrong breakeven calculation discounts everything that followed it. The expert treats mental math as a drilled reflex, honed in the case-math work ahead, and never as an afterthought.
6. "Consulting is basically an audit, or market research with better slides." Both comparisons erase what defines the job: a forward-looking, owned recommendation reached through hypothesis-driven structuring. A backward-looking conformance check doesn't qualify, and neither does a handed-over dataset.
7. "The best answer is the most complete one." Completeness is an academic and engineering virtue. Under a real deadline, the best answer is the most defensible and actionable one reachable in the time available, which is the 80/20 rule at work. The expert would rather deliver a well-hedged directional answer on Friday, clearly labeled as directional, than a "complete" one the following Friday. Friday's decision was the one that needed making.

Run your eye down all seven and you'll see one thread. A consulting problem isn't a research problem wearing a suit. It's a decision problem under a real clock, and each of the three moves (structure, hypothesis, answer-first) exists to produce a defensible, actionable answer inside that clock. How thorough the process felt was never the measure.

Worked examples

Worked example 1: Define: the presenting problem is not the real problem (India, ₹)

The board of Grihalaxmi Appliances (a synthetic mid-sized Pune appliance retail chain; all figures illustrative and internally consistent, not a claim about any real company) calls an emergency meeting. Revenue has held flat at ₹450 crore for the year, but profit has fallen from ₹36 crore to ₹18 crore, a 50% collapse. The chairman opens with: "This is Amazon and Flipkart eating us alive. I want a plan to match their online prices, and I've already asked finance to model a ₹40 crore digital-transformation capex ask for the next board meeting." Everyone in the room nods. You're handed the brief to "confirm the online-competition diagnosis and finalize the capex case."

Solution. Step 1: write the situation-complication-question, and notice what's missing. Start with the situation. Revenue is flat at ₹450 crore, and profit fell from ₹36 crore to ₹18 crore, so margin fell from 8% to 4% (₹36cr ÷ ₹450cr = 8%, ₹18cr ÷ ₹450cr = 4%, a clean, internally consistent pair). Now the complication. The board has already supplied a cause, online competition, and a cure, ₹40 crore of capex, before anyone has done any structuring or hypothesis-testing. The brief you've been handed skips stages 2 through 5 of the cycle entirely and starts at stage 6.

Step 2: test the board's causal story against one cheap, available fact, before accepting the brief as written. Suppose, for illustration, that Grihalaxmi's category mix shows the profit decline concentrated in large appliances such as refrigerators and washing machines. In this market the great majority of unit volume in those categories is still bought in physical stores, because of installation and financing needs, and only a small share has historically migrated online. If the "Amazon problem" story were true, the pain should show up disproportionately in the categories most exposed to online substitution. If it shows up in the least exposed ones instead, the presenting diagnosis doesn't survive contact with the company's own sales mix.

Step 3: redefine the decision-making question. Nobody is really asking "how do we match online prices?" The question is: "Is Grihalaxmi's profit collapse driven by competitive price pressure (which would justify the capex ask), or by something the ₹40 crore capex cannot fix at all, and which is it, given the mix evidence above?" That's a very different brief. The board decides to adopt it. You don't get to reinterpret your instructions on the quiet.

Read it. A client's presenting problem is often a hypothesis about the cause, delivered with the confidence of a fact. Your job at the define stage is to notice that gap before a single hour of analysis goes into solving the stated problem elegantly while the real one waits. Worked Example 5 returns to Grihalaxmi once the analysis is in.

Worked example 2: Structure: the MECE profit tree (US, $)

Northlight Analytics (a synthetic Boston-based B2B SaaS company) reports $80M of annual revenue against $92M of total cost, a $12M net loss. The CFO's brief is blunt: "Build me a plan to fix profitability." Before analyzing anything, structure the problem.

Solution. Step 1: the algebraic cut. Profit = Revenue − Cost. Revenue = (annual contract value per account) × (number of accounts) = $40,000 × 2,000 accounts = $80,000,000, which matches the given figure. Cost = COGS + Sales & Marketing + R&D + G&A = $20M + $36M + $20M + $16M = $92M, which matches too. This is a synthetic, internally consistent illustration, and no real filing stands behind it.

Step 2: test the tree for MECE. Is it mutually exclusive? Yes, by construction. Under any standard chart of accounts, a given dollar of cost is booked to one of COGS, S&M, R&D or G&A, and never two. Is it collectively exhaustive? Also yes, by construction, because the four categories are defined to sum to total operating cost with no residual left over. A real engagement would confirm this against the actual trial balance. Here it holds because of the way the illustration was built.

Step 3: let the tree localize the highest-priority branch, without yet analyzing it. Express each cost bucket as a percentage of revenue. COGS is 25% ($20M ÷ $80M), S&M 45% ($36M ÷ $80M), R&D 25% ($20M ÷ $80M) and G&A 20% ($16M ÷ $80M). Take an illustrative benchmark for a company of this profile: S&M typically runs at 25–35% of revenue at this growth stage. That's a stated, illustrative reference point, and nobody is citing it as a real-world statistic. Against it, the 45% S&M figure stands out furthest from a plausible norm, so it's the natural first candidate for prioritization in Worked Example 3.

Read it. The tree did no analysis at all. It only located, cleanly and without double-counting, where $92M of cost lives, and let one simple ratio suggest where to look next. The structure stage hands you a map. The verdict comes later.

Worked example 3: Prioritize: ranking hypotheses by impact against cost-to-test (India, ₹)

A telecom engagement has survived the structure stage with three live hypotheses about a margin decline, and the team has three weeks. Hypothesis H1, network-maintenance cost inflation, could move the answer by an estimated ₹8 crore and would take about 2 days to test with data already on hand. Hypothesis H2, a competitor's pricing war, could move the answer by an estimated ₹15 crore but would take about 10 days to test properly, because it needs a market survey. Hypothesis H3, a billing-system error under-charging a customer segment, could move the answer by an estimated ₹3 crore and would take about 1 day to test with a single data pull.

Solution. Step 1: compute a simple prioritization score, impact ÷ cost-to-test. H1: ₹8 crore ÷ 2 days = 4.0 crore/day. H2: ₹15 crore ÷ 10 days = 1.5 crore/day. H3: ₹3 crore ÷ 1 day = 3.0 crore/day.

Step 2: rank by score, not by raw impact. Ranked: H1 (4.0) > H3 (3.0) > H2 (1.5). Raw impact alone would have put H2 first at ₹15 crore, so this ranking is a different one. The prioritize stage trades impact against the cost of finding out. A three-week engagement can't afford to spend a third of its clock chasing the largest hypothesis first when two smaller, much cheaper tests can be run, and possibly resolved, before it.

Step 3: sanity-check the ranking against sequencing and dependency as well as the score. A pure score can mislead. Say H3's billing-error test would also generate data you need to scope H1 properly. Then running H3 first has value beyond its own score. Treat the score as a starting point for a fast, defensible ranking. It can't replace your judgment about how the pieces of work relate to each other.

Read it. Novices skip prioritization more often than any other stage, treating every surviving hypothesis as equally worth chasing. That's the boiling-the-ocean failure from earlier, under another name. A simple impact-over-cost-to-test ratio, sanity-checked for sequencing, turns "everything matters" into an ordered, defensible plan for the next few days.

Worked example 4: Analyze: the minimum killing analysis (US, $)

Returning to Northlight Analytics (Worked Example 2), the prioritized hypothesis is: "S&M efficiency has collapsed because customer acquisition cost (CAC) has risen faster than customer lifetime value (LTV)." Design and run the minimum analysis that would confirm or kill this claim.

Solution. Step 1: identify the two numbers that would settle the question, and nothing more. CAC = S&M spend ÷ new customers acquired in the period = $36,000,000 ÷ 200 = $180,000 per new customer. LTV = (annual contract value × gross margin %) ÷ annual churn rate = ($40,000 × 0.70) ÷ 0.20 = $28,000 ÷ 0.20 = $140,000 per customer. Both formulas come straight from the contribution/CAC/LTV machinery in the Finance galaxy's unit-economics material, so they aren't re-derived here.

Step 2: compute the ratio that actually tests the hypothesis. LTV ÷ CAC = $140,000 ÷ $180,000 ≈ 0.78. An illustrative rule of thumb for a healthy subscription business puts a sustainable LTV:CAC ratio meaningfully above 1, and it's often cited informally at around 3. Treat that as a reference figure. Nobody here is citing a real statistic. Either way, 0.78 is well below even the loosest healthy threshold.

Step 3: state what this does and does not prove. It confirms the hypothesis. At current spend efficiency, each new customer costs more to acquire than that customer will ever return, so growth itself is currently destroying value. It's worse than merely expensive. What it does not tell you yet is why CAC is so high. Is it channel mix, sales-cycle length or win-rate? Answering that takes a second, more granular hypothesis and a second minimum analysis, and stage 4 reserves those for after this first, coarser hypothesis is confirmed.

Read it. Two numbers, correctly defined and divided, confirmed a specific claim in minutes. There was no dashboard, no cohort study and no "check everything else while we're in the data" detour. That's what "minimum killing analysis" means, and the hypothesis work ahead drills it against harder, messier cases.

Worked example 5: Synthesize: turning findings into one governing thought (India, ₹)

Return to Grihalaxmi Appliances (Worked Example 1). The analysis, illustrative and synthetic, is complete, and three findings have survived. Gross-margin compression from unrecovered input-cost inflation cost the business ₹10 crore. A wave of store-lease renewals at sharply higher rents cost ₹6 crore. A one-off inventory write-off cost ₹2 crore. Online competition, the board's original theory, turned out not to be a material direct driver of the decline at all.

Solution. Step 1: check that the findings actually reconcile with the problem as defined. ₹10 crore + ₹6 crore + ₹2 crore = ₹18 crore, which matches the profit fall stated in Worked Example 1, from ₹36 crore to ₹18 crore. Don't treat this check as a flourish you can skip. If you can't show that a synthesis adds up to the original gap, it isn't trustworthy yet.

Step 2: separate the findings (what happened) from the insight (so what). A finding is "gross margin fell 3 points" or "lease costs rose." An insight is the one sentence that turns those findings into an answer to the question Worked Example 1 redefined: "The profit decline is structural cost, not competitive, and it is fixable within a year, without new capex, by renegotiating supplier contracts and the recently repriced store leases."

Step 3: notice what the insight does to the board's original ask. It answers the redefined question ("is this competitive or something the capex can't fix?") with "something the capex can't fix." And it recasts the ₹40 crore digital-transformation ask as most likely solving the wrong problem. The board couldn't have reached that conclusion from the findings alone without this synthesis step.

Read it. Synthesis turns three separate, reconciled findings into one governing thought that answers the question actually asked. A longer list of true things wouldn't do that. Jump straight from findings to a recommendation and you get the "data dump dressed as an answer" failure from the cycle table above.

Worked example 6: Recommend: the answer-first paragraph, built and contrasted (US, $)

Using the Northlight Analytics findings (Worked Examples 2 and 4), write the recommendation two ways: chronologically, then answer-first.

Solution. Step 1: the chronological version, which is what most people write first and what you shouldn't lead with. "We began by building a profit tree and found that S&M was 45% of revenue against a benchmark range of 25–35%. We then tested whether this reflected an acquisition-efficiency problem by computing CAC and LTV, and found CAC of $180,000 against LTV of $140,000, an unfavorable ratio of 0.78. Given this, we recommend Northlight reduce new-logo spend outside its top three verticals for two quarters..." The recommendation itself doesn't arrive until the third sentence, after the reader has been walked through the whole analytical path.

Step 2: the answer-first version. "Northlight should freeze new-logo sales-and-marketing spend outside its top three verticals for two quarters and redirect that budget to net-revenue retention, because at current spend efficiency the company is acquiring customers for more than they are worth: blended CAC of $180,000 exceeds LTV of $140,000 (a 0.78 ratio, well below a healthy threshold). This is driven overwhelmingly by S&M running at 45% of revenue against a 25–35% benchmark range for a company at this stage, concentrated in low-conversion verticals outside the top three. The risk is a near-term slowdown in new-logo bookings, which we would mitigate by ring-fencing the top three verticals' budgets untouched and reviewing the freeze after two quarters against a recomputed LTV:CAC ratio. Next step: finance re-cuts the S&M budget by vertical for sign-off this week."

Step 3: name what changed between the two versions, and what did not. The underlying facts and numbers are identical, and so is their order of discovery. The difference is which sentence a time-pressed reader hits first. In the second version, a CFO who reads only the opening sentence already has the whole actionable conclusion, the reason behind it and enough of the number to sanity-check it. Nothing downstream is needed to act, though all of it stays there to defend the ask under questioning.

Read it. For now, that paragraph is all "answer-first" asks of you: governing thought, supports, risk, next step, in prose. It's deliberately not a full pyramid built from a large, messy set of findings, and it isn't a storyline across ten exhibits. That construction, its rubric and its crown-level grading bar belong at the crown of the foundations branch. Deck and visual craft belong to the future Presentation galaxy. Get comfortable with the shape here and the deeper work will feel like a continuation. You won't be starting a new subject.

Practice set

Work each problem fully before you read its solution. Company names are synthetic. All figures are illustrative and internally consistent unless a real, publicly reported figure is explicitly cited, and none are cited in the problems below.

P1 (guided). The thesis, briefly. State in a single sentence what the consulting mind is at bottom, and in one more sentence explain why "gather all the facts first" fails under a real engagement's constraints.

Solution. The consulting mind is a structured way of processing an unfamiliar, high-stakes problem (structure first, hypothesis-driven, answer-first). It isn't a body of facts to memorize. "Gather all the facts first" fails because a real engagement runs on a fixed clock, a fixed budget, and an audience that must act on the answer, so exhaustive analysis either misses the deadline or arrives too late to change the decision it was meant to inform.

P2 (guided). Situation-Complication-Question. A regional airline's board says: "Our on-time performance is terrible and it's costing us customers. Build us a plan to fix scheduling." Write a one-line situation, a complication and the question that matters, without accepting the board's implied cause.

Solution. Situation: on-time performance has declined and customer complaints have risen. Complication: the board has already assumed the cause (scheduling) and the fix (a scheduling overhaul) before any structuring has happened. On-time failures could equally stem from maintenance turnaround, crew scheduling, air-traffic congestion at specific hubs, or weather exposure on specific routes. Question: "What is actually driving the on-time decline, and does the answer point to a scheduling fix, a maintenance fix, or something else entirely?"

P3. MECE test: spot the flaw. A retailer segments its cost base into: cost of goods sold, employee costs, and marketing costs, and claims the split is MECE. Evaluate.

Solution. It isn't collectively exhaustive. Rent, utilities, depreciation, corporate overhead, interest and taxes are all real costs that fit none of the three named buckets. A MECE cost tree needs either full enumeration of every cost category or an explicit "other operating expenses" bucket, so nothing gets dropped without anyone noticing.

P4. Build an algebraic cut. A restaurant's profit fell. Build a two-level algebraic issue tree starting from Profit = Revenue − Cost.

Solution. Level 1: Profit = Revenue − Cost. Level 2: Revenue = (covers served) × (average check size); Cost = Food cost + Labor cost + Rent + Other fixed costs. Each level-2 branch is true by definition of how revenue and cost are constructed. That satisfies mutual exclusivity, since no cost or revenue driver is counted twice, and collective exhaustiveness, since nothing is left out once "other fixed costs" is the catch-all.

P5. Case math: breakeven. The restaurant from P4 has fixed costs of $18,000/month. Average check size is $28, and the variable (food) cost per check is $10. What is the monthly breakeven number of covers?

Solution. Contribution per cover = $28 − $10 = $18. Breakeven covers = $18,000 ÷ $18 = 1,000 covers per month.

P6. Case math: CAGR. A company's revenue grew from ₹50 crore to ₹72.6 crore over exactly 3 years, compounding steadily. What is the approximate annual growth rate? (Hint: check whether 20% fits cleanly.)

Solution. ₹50 crore × 1.20³ = ₹50 crore × 1.728 = ₹86.4 crore. That's too high, so 20% doesn't fit. Try 13%: 1.13³ ≈ 1.4429, and ₹50cr × 1.4429 ≈ ₹72.15cr, which is close. The exact rate solves 72.6/50 = 1.452 = (1+r)³, so 1+r = 1.452^(1/3) ≈ 1.1325, i.e., ≈13.25% per year. Check it: 50 × 1.1325³ ≈ 50 × 1.4523 ≈ 72.6, so the numbers reconcile.

P7. Case math: weighted average. A firm's total headcount is 70% based in India at an average fully-loaded cost of $18,000/year and 30% based in the US at an average fully-loaded cost of $95,000/year. What is the blended average fully-loaded cost per head?

Solution. (0.70 × $18,000) + (0.30 × $95,000) = $12,600 + $28,500 = $41,100 per head.

P8. Case math: smart rounding. Estimate 62 × 58 in your head in under three seconds using the difference-of-squares shortcut, and show the shortcut.

Solution. 62 × 58 = (60 + 2)(60 − 2) = 60² − 2² = 3,600 − 4 = 3,596.

P9. Market sizing: bottom-up mini-estimate. Estimate the number of smartphone screen-repair jobs per year in a city of 12 million people, assuming 70% own a smartphone and each owner needs a screen repair, on average, once every 4 years. State the estimate and your logic.

Solution. Smartphone owners = 12,000,000 × 0.70 = 8,400,000. Repairs per year = 8,400,000 ÷ 4 = 2,100,000 repair jobs per year. It's an estimate. If more data turns up, stress-test the "once every 4 years" assumption first, since the answer scales linearly with it.

P10. Hypothesis, hunch, opinion, or question? Classify each: (a) "Something feels off about our customer support costs." (b) "We should just outsource support." (c) "Why did support costs rise?" (d) "Support costs rose because average handle time per ticket increased, not because ticket volume did."

Solution. (a) is a hunch, too vague to test. (b) is an opinion/recommendation. It has skipped straight to "what to do" without passing through structure or analysis. (c) is the question a hypothesis exists to answer. (d) is the only real hypothesis. It's specific and falsifiable, because a test of handle-time vs. volume data would confirm or kill it, and it names in advance what evidence would decide it.

P11. Design the minimum killing analysis. For hypothesis (d) in P10, what is the single fastest, cheapest analysis that would confirm or kill it?

Solution. Pull two numbers already likely to exist in the support system: (i) average handle time per ticket, this period versus the prior period, and (ii) total ticket volume, this period versus the prior period. If handle time rose materially while volume was flat or fell, the hypothesis is confirmed. If handle time was flat and volume rose instead, it's killed, and a different hypothesis takes its place: a volume-driven cost problem. You don't need a cohort study, a customer survey or a headcount audit to take this first, decisive step.

P12. Rewrite bottom-up as answer-first. Rewrite this chronological sentence as an answer-first governing thought: "We looked at the data, found that late deliveries were concentrated in one warehouse, investigated further, and discovered that warehouse was short-staffed on weekends, so we think adding weekend staff there would probably help."

Solution. "The company should add weekend staff to the [named] warehouse, because it is understaffed specifically on weekends and that gap accounts for the concentration of late deliveries there." The governing thought comes first, carrying the recommendation and its reason. The discovery process shrinks into the reason clause, so nobody has to sit through it step by step.

P13. Framework recognition. A national grocery chain wants to decide whether to enter the meal-kit-delivery category as a new line of business. Which framework most directly evaluates whether this specific industry is worth entering, independent of the grocery chain's own capabilities? (a) 7S (b) Porter's Five Forces (c) Weighted-average cost of capital (d) DuPont decomposition

Solution. (b) Porter's Five Forces. It evaluates the meal-kit-delivery industry's structural attractiveness (rivalry, buyer/supplier power, entry threat, substitutes) on its own terms, independent of who is considering entering it. (a) evaluates internal organizational fit, which is a separate and later question about capability. (c) and (d) are finance-galaxy tools for cost of capital and return decomposition, and neither measures industry attractiveness.

P14. Spot the framework-as-theater mistake. A junior consultant, asked "should we raise prices?", opens with a full BCG Growth-Share Matrix classification of the company's nine product lines before addressing pricing at all. What is wrong with that?

Solution. BCG's matrix answers a portfolio capital-allocation question: which businesses to fund, hold, or divest. It isn't built to answer a pricing question at all. Running it first is framework-as-theater, reaching for a familiar, impressive-looking tool on reflex before diagnosing what "should we raise prices?" requires. That question needs something closer to a cost/value/competition pricing framework, which you'll meet in full later.

Applied mini-project

The six-stage dry run.

Pick one real, currently unresolved, moderately ambiguous decision. It can come from your own work, a business you know well, or a public situation you can describe without non-public information. A plausible synthetic scenario is just as acceptable. Either way, keep it neutral, and don't name yourself or any private individual. Run it through all six stages of the cycle and produce a single one-page memo (350–600 words) with six labeled parts, no more and no fewer:

1. Define. A situation–complication–question statement for your chosen problem, written so that a stranger who knows nothing about the situation could tell what decision is being asked for.
2. Structure. A MECE issue tree with at least one level and at least three branches, naming which cut (algebraic, process, segment, or stakeholder) you used, and a one-line note showing you tested it for overlap and gaps.
3. Prioritize. A ranked short-list of which branch(es) you would analyze first, with your reasoning. An impact-versus-cost-to-test judgment is enough. If the real situation doesn't have exact numbers, you don't need them, and a defensible qualitative ranking is fine.
4. Analyze. State your day-one hypothesis for the highest-priority branch. Describe the single minimum analysis, real or realistically described, that would confirm or kill it, and the result you'd expect under each outcome.
5. Synthesize. One governing "so what" sentence that answers the question from part 1. Give the insight. A list of findings won't do.
6. Recommend. A short answer-first paragraph: the recommendation first, then two or three supports, a named risk, and a concrete next step.

Self-scoring rubric. Score each of the six parts 0/1/2.

• 0 means missing, or done in the wrong order, such as a recommendation with no synthesis behind it.
• 1 means present but flawed: an issue tree with a real overlap, a hypothesis that's actually an opinion, or a recommendation that buries its conclusion after the reasoning.
• 2 means it meets the bar the worked examples above set.

Pass at 9 of 12, with no zero on any single part. That mirrors the whole idea of the cycle. A strong performance on five stages doesn't excuse skipping the sixth.

The gate for this node. There's no mapped Case-Engine case here. The placement diagnostic above is the "challenge" component, alongside the mastery quiz below. Your first full Case-Engine case in the Consulting galaxy arrives with the branch's own belt node and the modules right after it. The memo is deliberately a stand-alone, self-graded dry run, and it doesn't stand in for that later, richer case work.

Reading & resources

Learn from these, and check any specific claim you cite in real work against a primary source. None of them can build the skill for you.

The core canon:

• Barbara Minto, The Pyramid Principle. [Paid] [Advanced]: the original source for MECE and answer-first pyramid logic. You've met only its opening idea so far, and CN1.03 applies it in full.
• Charles Conn & Robert McLean, Bulletproof Problem Solving. [Paid] [Beginner–Advanced]: the clearest modern statement of hypothesis-driven, issue-tree-based problem solving, and the source of the "exhaustive versus focused" distinction used above.
• Ethan Rasiel, The McKinsey Way and The McKinsey Mind. [Paid] [Beginner]: short, readable introductions to MECE, hypothesis, and the "so what," good for building intuition before the heavier texts.
• Richard Rumelt, Good Strategy / Bad Strategy. [Paid] [Advanced]: the diagnose-before-prescribe principle behind the framework-as-theater warning above. Full application comes in CN3.03.

Casing practice (for the diagnostic's framework and structure slices):

• Marc Cosentino, Case in Point. [Paid] [Beginner]: the standard case-format reference. Full depth comes in CN2.03–CN2.04.
• Victor Cheng, Case Interview Secrets, and the free materials at caseinterview.com. [Mixed] [Beginner–Intermediate]
• Free MBA consulting-club casebooks: Wharton, Kellogg, Fuqua/Duke, Darden, INSEAD, LBS, Ross, Booth, Columbia. [Free] [Intermediate]: downloadable PDFs with real practice cases and model answers.
• Official free practice cases published by McKinsey, BCG, and Bain's own career sites. [Free] [Beginner–Advanced]

General:

• Harvard Business Review (HBR), across strategy and problem-solving topics. [Mixed, free/paid] [Beginner–Advanced]: the general-purpose anchor for the whole Consulting galaxy.
• M10.03 (The Communication Toolkit), if you arrived via the Finance galaxy. It's the identical Minto logic, already in your hands, and CN1.03 applies it to a consulting recommendation without re-deriving it.

Do this as well as reading. Take one real headline about a company you follow, say a profit warning, an entry into a new market or an acquisition. Before you read any analyst commentary on it, write your own one-paragraph situation-complication-question and a single-level MECE tree for "what could explain this." Ten minutes of doing this cold will teach you more about the structure-first reflex than an hour of reading about it.

Teach it back & journal

Feynman prompt. Explain to a friend who has never heard of "case interviews" or "strategy consulting" why a smart, careful person who insists on "seeing all the data first" would actually do worse at this job than someone willing to guess an answer on day one, using the fixed-clock, fixed-budget, must-act argument from the opening. Then sketch a single-level MECE profit tree on a napkin, for a business they know, and ask them to break it. Can they find a fact that fits two branches, or none? If the tree survives their attack, and you can explain in a sentence why "gather everything first" is backwards here, the idea has stuck.

Journal prompt. For the next real ambiguous problem you personally face, write down, before you do anything else, your day-one hypothesis and a one-level MECE tree for it. Then act, and afterwards write down whether your hypothesis was confirmed, killed or needed to change. Which habit would most change how colleagues experience your judgment: defining the question that matters in place of the presenting one, prioritizing before analyzing, or stating your conclusion first? Note what you'll do this week to make it automatic. Keep the entry. The next node asks you to revisit it once you've run a fuller engagement end to end.

This module's flashcards and mastery quiz are wired into the app: see the node's Quiz and Reviews.