CaseClimbThe most common piece of bad advice in case interview prep is "learn the frameworks." Profitability trees and 4Ps are fine as a starting vocabulary, but interviewers at the top firms can spot a memorized template within thirty seconds, and they read it as a red flag. Interviewers reward a repeatable process for building a structure from scratch, tailored to the business in front of you.
Step 1: Restate the objective in your own words
Before you structure anything, say back what you think the client wants, in one sentence. "They want to grow revenue" is too vague to structure against in most cases. "They want to decide whether to enter the Brazilian market within the next 18 months" is precise enough. If you get this wrong, everything you build on top of it is wrong too, and checking costs you ten seconds.
The restatement also fixes the unit of success. If the client wants to "improve performance", ask whether that means profit, growth or share, and over what period. A good clarifying question narrows the problem: "Is the goal to restore the margin to where it was two years ago, or to reach a specific target?" Two sharp questions read as discipline; five vague ones read as stalling.
Step 2: Build 2-4 buckets that are MECE for THIS problem
MECE (mutually exclusive, collectively exhaustive) is the standard most candidates quote and few apply. The practical test: could you file any fact about this business under exactly one of your buckets, with nothing left over? For a profitability case, that's often revenue and cost. For a market-entry case, it's more often market attractiveness, right to win, and entry approach. The buckets should come from the specific prompt rather than from a template you memorized before walking in, which is why market-entry and profitability structures shouldn't look alike, even though both are "cases."
A reliable way to generate buckets without a template is to ask what would have to be true for the client to get what they want, then group the answers. Once the buckets exist, test them. If a single fact such as "pricing" could sit under both revenue and competition, merge or redraw. Then name three facts about the business at random and check that each has a home. Candidates fail that second test most often by forgetting the external side (competitors, regulation, customers) when the prompt sounds internal.
Step 3: Attach a hypothesis to each bucket, not only a question
This is the difference between a hypothesis-led structure and a checklist. Instead of "I'll look at costs," say "I suspect costs are the bigger issue here because revenue has held steady while margin has fallen, so I'd want to check fixed versus variable cost trends first." A hypothesis gives the interviewer something to react to and gives you a reason to ask for one exhibit before another. It also signals that you are thinking like someone solving the problem rather than someone reciting a framework.
A hypothesis does not need to be right. It needs to be specific enough to be tested by the next piece of data you ask for. "I think store labor cost has risen faster than sales, because the chain has added service counters" can be confirmed or killed by one exhibit.
Step 4: Say it out loud in under 90 seconds
A structure that takes four minutes to explain has, in most cases, grown too many branches to be useful, even if you can defend each branch on its own. Practice compressing: objective, two to four MECE buckets, one hypothesis per bucket, and which one you want to start with and why. If you can't say it in about a minute and a half, prune the structure rather than explaining it for longer.
Delivery has an order that interviewers recognize. Lead with the objective in one sentence. Name the buckets as a list before you go into any of them, so the listener has the map. Take each bucket in turn with its hypothesis and the data you would want, then close by saying where you would start and why. If you are using paper, turn it so the interviewer can read it and point as you talk.
A worked example: a grocery chain's falling margins
Suppose the interviewer opens with: "Our client is a regional grocery chain with 400 stores. Its operating margin has fallen for two consecutive years, and the CEO wants to know why and what to do about it." Two clarifying questions tell you that revenue has been flat at about $2 billion a year and that operating margin has slipped from 4% to 2.5%. Before you say a word about buckets, do the arithmetic. Four percent of $2 billion is $80 million of operating profit, 2.5% is $50 million, so the client has lost $30 million of annual profit with no change in sales. Flat revenue plus a $30 million profit drop means costs rose by about $30 million, or 1.5% of revenue.
That one calculation shapes the structure. Revenue stays in as a bucket, since flat sales can hide a mix shift toward lower-margin products, and cost becomes the lead bucket with a hypothesis you can size: cost of goods in grocery runs near 70% of sales, so about $1.4 billion here, and a 2% rise in that line alone would be $28 million, most of the $30 million gap. Quick arithmetic done before you speak is what lets a structure carry a point of view. Said out loud, with the interviewer able to see your page, it sounds like this.
"So the client runs 400 grocery stores, revenue is flat at around $2 billion, and operating margin has dropped from 4% to 2.5% over two years. That is a fall from $80 million to $50 million in operating profit, so we are looking for about $30 million of lost profit a year, and the goal is to find where it went and recover it. I would split this into three areas. First, cost, which I think is where most of the answer sits, because sales have not moved while profit has fallen. Second, revenue quality, because flat revenue can still hide a shift toward lower-margin categories or heavier promotions. Third, what the fix looks like once we know the cause, including what is achievable in the next 12 to 24 months."
"On cost, I would start with cost of goods, since it is around 70% of sales in this business, so about $1.4 billion, and a 2% rise there would explain almost the whole gap. I'd want cost of goods as a percentage of sales by year, then store labor and occupancy, since those are the next largest lines. On revenue quality, I'd look at category mix and promotional intensity over the same two years. I'd leave the fix until we know the driver. If that works for you, I'd like to begin with the cost of goods trend, because that is where the arithmetic points."
Notice what this leaves out: no fixed-versus-variable split, no 4Ps, no competitor bucket added for completeness. It sizes the problem, names three buckets that fit a flat-revenue margin squeeze, puts a testable hypothesis on the lead bucket and asks permission to start there. Spoken at a normal pace it runs about two minutes.
Interviewer-led versus candidate-led cases
The example above assumes a candidate-led case: you build the structure, pick where to start, ask for data and steer the analysis. In an interviewer-led case, the interviewer holds the agenda. You still open with a structure, but after that the interviewer asks a series of set questions, and your structure serves as a reference point rather than a route map. Candidate-led cases test whether your structure can carry the whole case, so the hypothesis and the starting point matter a great deal. Interviewer-led cases test breadth, since you will be asked about areas you ranked lower.
Adjust the structure to the format. In a candidate-led case, three buckets with a strong lead hypothesis beat five with none, because you will have to drive the analysis yourself and depth wins. In an interviewer-led case, go a little broader and shallower, four buckets rather than three, so that whichever question comes next you can say "that sits under my second area, customer economics". In both formats, if an exhibit arrives from a bucket you had not planned to start with, say so, update your hypothesis out loud and move on.
Common mistakes and how to fix them
- Reciting a template. The buckets match a case book rather than the prompt. Fix: build buckets from what would have to be true for the client to succeed, then check them against MECE.
- Too many branches. Six buckets with sub-buckets take four minutes and lose the interviewer. Fix: cap the structure at four top-level buckets and hold the sub-points until you enter a bucket.
- Buckets without hypotheses. "I'd look at revenue and costs" says nothing about your thinking. Fix: one sentence per bucket that starts with "I suspect" and contains a "because".
- Skipping the arithmetic. Candidates structure a margin problem without sizing the gap. Fix: if the prompt gives you two numbers, do something with them before you speak.
- No starting point. The structure ends in silence and the interviewer has to prompt. Fix: close with "I'd like to start with X because Y" and ask whether that works.
How interviewers score the structure
Structuring is the first of the five skills on the scoresheet, and it sets up the other four. Structuring itself is scored on whether the buckets fit the prompt, whether they are MECE in practice rather than in name, and whether you prioritized. Quantitative analysis shows up here if you size the problem before structuring. Business judgment is read from your hypotheses: an interviewer can tell within a sentence whether you know that cost of goods dominates grocery economics or are guessing. Synthesis depends on the structure too, because a recommendation that answers the objective you restated at the start is far easier to deliver than one assembled from scattered findings. Communication is scored on the delivery: the order, the length and whether the interviewer could follow.
The habit that builds this skill
None of this becomes fast under pressure because you read about it once. It becomes fast the way any timed skill does: repeated reps against prompts you haven't seen before, scored where possible on whether you tailored the structure, the first of the five skills interviewers grade, or reached for something familiar-sounding. That is a different exercise from reading case books, and it is the one that carries over into a real interview room.
Common questions
How long should the structure take? Aim for ninety seconds to two minutes spoken, plus a minute or so of silent preparation before you begin. Ask for that preparation time; interviewers expect it. Past three minutes you are spending time you will need later.
Is it fine to use a framework like the profitability tree? Yes, as a starting vocabulary. The problem is reciting one without adapting it. If revenue and cost are the right split for the prompt, use them and name the hypotheses that make them specific to this client.
What if my hypothesis turns out to be wrong? That is normal and costs you nothing if you handle it well. Say that the data does not support your initial view, state what it does suggest and move to the next bucket. Interviewers score the update, and a candidate who changes course on evidence looks stronger than one who guessed right.
Should I ask clarifying questions before structuring? Yes, one to three, and make them count. Ask about the objective, the timeframe and any constraint that would change the structure, such as whether the client is willing to close stores.
