# The Case Interview, Run From Prompt to Recommendation on the Clock

*You will run a full case on the clock, hitting the specific graded output at each stage from clarifying question to closing recommendation.*

- Canonical URL: https://www.refolk.ai/candidates/guides/case-interview-prompt-to-recommendation
- Pillar: Interviewing
- Format: Playbook
- Published: 2026-10-10
- Last reviewed: 2026-10-10
- Reading time: 17 min

This is the playbook for one job: getting through a live consulting case interview and handing over a clear, defensible recommendation before the clock runs out. It is written for early-career candidates targeting MBB and strategy roles, who have read the "seven frameworks" lists and still cannot run a whole case on time. What follows runs a single case minute by minute - clarify, structure, drive the analysis, carry the math, synthesize - with the specific output each stage is graded on, so you can rehearse against deliverables instead of memorizing templates.

The audience for this is large and front-loaded. In Refolk's index of professional profiles, there are 133,147 US profiles carrying the "Management Consulting" skill at entry level against 43,507 at senior level, a 3.1x drop from the bottom of the funnel to the top. Most people reading this are competing to get in, against firms that state they accept under 1% of applicants. The case is the gate, and the gate is a timed exercise with a defined deliverable at every stage.

## Why the whole case must be on a clock

A case interview is a timed exercise where front-loaded minutes are stolen directly from the scored analysis phase, so pacing is not a style preference, it is structural. A standard MBB-style case runs 30 to 45 minutes, and the published allocation puts roughly 60% to 70% of that time on analysis. The clarifying, structuring, and recommendation phases share what is left.

The mistake almost every self-prepared candidate makes is treating the opening as free. It is not. On a 30-minute case the analysis block is 18 to 22 minutes and it is non-negotiable, because that is where the interviewer watches you think. A candidate who burns ten minutes clarifying and structuring has stolen that time from the only phase that scores highly, and there is no recovery inside a fixed-length case.

**3.1x - US entry-level to senior drop in consultants with the "Management Consulting" skill, in Refolk's index**

133,147 entry-level profiles against 43,507 senior - the funnel this guide serves is overwhelmingly early-career.

The practical rule the coaches enforce is that clarify plus structure plus present should fit inside about five minutes, and you should be into analysis six to seven minutes in. The generic guides that allot three to five minutes just to clarify and four to six minutes to structure are describing a case you will lose on time. Treat the opening as something you execute crisply and leave, not something you dwell in to look thorough.

> **Rule:** Protect the analysis block
>
> Clarify, structure, and present should take about five minutes combined, so you are driving analysis six to seven minutes in. Time spent in the opening is subtracted directly from the 65% of the case that scores.

## The time budget, phase by phase

Here is the published per-phase split for a 30-minute case. Treat these as proportions, not fixed minutes: a 25-minute case compresses every row and a 40-minute case stretches analysis most, but the shape holds.

| Phase | Time | % of case |
|---|---|---|
| Clarifying questions | 1-3 min | ~10% |
| Structure/framework | 2-4 min | ~10% |
| Analysis | 18-22 min | ~65% |
| Recommendation | 3-5 min | ~15% |

The two numbers to internalize are the 65% and the 15%. The 65% is where structured thinking, quantitative analysis, and business judgment are actually observed. The 15% recommendation is short on purpose - a strong synthesis runs 60 to 90 seconds and is a decision, not a replay. If your recommendation is running past two minutes, you are recapping, and recapping does not score.

#### Where the clock goes on a 30-minute case

| Stage | Figure | Note |
| --- | --- | --- |
| Analysis | 18-22 min | the scored core |
| Recommendation | 3-5 min | 60-90 sec of talking |
| Structure | 2-4 min | build silent, present fast |
| Clarifying | 1-3 min | one or two questions |

*Analysis dominates at roughly 65% of the case, so the opening phases must be executed fast and left behind.*

## The seven starting structures and how to pick one

The commonly cited set covers seven case types: profitability, market entry, M&A, pricing, growth strategy, operations, and market sizing. These are starting points for building a tailored tree, not templates to recite. In one analysis of case types asked across McKinsey, BCG, and Bain, profit improvement topped the list at 20% of reported questions, so if you drill one pattern hardest, drill profitability.

Each type suggests a different first split. Learn the first branches as a launch pad, then tailor from there.

| Case type | First branches |
|---|---|
| Profitability | Revenue vs costs, then price, volume, fixed, variable |
| Market entry | Attractiveness, competition, right to win, entry strategy |
| M&A | Strategic fit, value, synergies, risks, integration, valuation |
| Pricing | Cost-based, value-based, competitive, implementation |
| Operations | Map process, bottlenecks, quantify impact, sequence |

Market entry has a documented logical order worth memorizing because it doubles as your path to the answer: first confirm the market is attractive, then check competition is weak enough to win meaningful share, then confirm the company can actually enter, and finally verify it will be profitable. That ordering is the structure and the sequence of analysis at once.

Picking the pattern is a quick read of the prompt's verb. "Should we enter" is a decision-options or market-entry case. "Why are profits falling" is a profit-driver tree. "Should we acquire" is deal value. The pattern tells you the shape; the client specifics fill the branches. A tree with no client-specific driver in it is a memorized tree, and interviewers are trained to catch exactly that.

## Clarify without over-clarifying

Clarifying means asking the one or two questions that would change your structure, then stopping. Recap the prompt in your own words, ask your sharp questions, and move on, two minutes maximum. More questions signal uncertainty, not thoroughness.

The test for whether a question earns its place is simple: would the answer change a branch of your tree? "What is the client's goal - growth or margin?" changes the tree. "What industry is this?" was usually in the prompt and asking it signals you did not listen. Cap yourself at two to three scope questions, and make each one load-bearing.

> **Watch out:** Over-clarifying reads as nervous, not rigorous
>
> Three or four clarifying questions tells the interviewer you are hunting for certainty before you can think. Cap at one or two, and only ask what would actually reshape your structure.

Done looks like this: the objective, the success metric, and the scope are stated back and the interviewer has agreed. Now you go quiet and build.

## The full procedure, start to finish

Run the case in this order. The timings assume a 30-minute case; scale the proportions for shorter or longer. The whole point is that each step has a defined "done," so you can rehearse the step in isolation and know when you have hit it.

#### One case, prompt to recommendation

1. **Open and take notes** - Capture client, industry, metric, and the exact question. Turn paper landscape, draw a vertical line, notes on the right third, framework on the left two-thirds. Done when you can restate the question verbatim.
2. **Clarify the objective** - Recap the prompt and ask one or two sharp questions, two minutes max. Each must be a question that would change your tree. Done when objective, metric, and scope are confirmed and agreed.
3. **Build the structure silently** - Take the silence. Name the decision, split into three to five mutually exclusive branches with sub-drivers, form a lead hypothesis. Done when you have a tree you can execute in the time left.
4. **Present the structure** - Lead with the top line - how many buckets and the overall logic in one sentence - then walk each branch in 60 to 90 seconds. Flag the branch you will attack first. Done when the interviewer can see your path.
5. **Drive the analysis** - Explore the priority branch, read exhibits, brainstorm. Give an interim synthesis after each branch and propose the next area in candidate-led cases. Done when each branch is tested and the driver isolated.
6. **Carry the math** - Restate the numeric goal and assumptions, sketch labeled intermediates, compute with units, sanity-check magnitude. Strip zeros, calculate, add them back. Done when you have a correct number plus its business implication.
7. **Synthesize the recommendation** - Answer in one firm sentence, two or three reasons from case data, one risk, one next step, in 60 to 90 seconds. Lead with the branch you prioritized first. Done when you hand over a decision the interviewer could act on.

### Build the structure silently, present it loud

Structuring fast means naming the decision, splitting it into three to five mutually exclusive branches, prioritizing the highest-value branch, and setting up your next question, all built inside about 90 seconds. Do not narrate while you build. Interviewers assess how you think when you present, not while you scribble, so take the silence, build it properly, then present cleanly. Announce the pause - "I'd like about 90 seconds to structure this" - so the quiet does not read as being stuck.

When you present, lead with the top line. Say how many buckets you have and the overall logic in one sentence before you walk each branch. The branch you prioritize first should be the one your lead hypothesis points at, because the link between structure and synthesis is direct: the branch you prioritized is usually the reason you lead with at the end. A hypothesis-led tree pre-writes your recommendation.

#### The structuring-to-synthesis loop

1. **Name the decision** - Restate the objective as the question your tree answers
2. **Split into 3-5 branches** - Mutually exclusive, with a client-specific driver in at least one
3. **Prioritize one branch** - Pick the highest-value branch and form a lead hypothesis
4. **Test and isolate** - Drive analysis on the priority branch first, confirm or kill the hypothesis
5. **Lead the close with it** - Open the recommendation with the branch you prioritized

*The branch you prioritize while structuring becomes the sentence you lead with when you close.*

## Carry the math so it cannot sink you

Case math is a gate, not a dimension. Industry data cited by coaches puts single mental math errors at roughly a 95% automatic-rejection rate, which means a narrated, sanity-checked answer beats a fast wrong one every time. You will be asked to estimate market sizes, calculate profitability, or solve for breakeven, all without a calculator.

The correct approach is the same every time. Restate the numeric goal and your assumptions aloud. Sketch the structure, labeling intermediates like population segments or cost per unit. Compute carefully, rounding to keep speed. Report the answer with units and a short business implication. Then propose a quick check or a sensitivity test. The structure-first habit is what protects you, because it lets the interviewer follow your logic even if a digit slips, and it lets you catch the slip yourself.

Two documented failure points account for most rejections. First, the breakeven denominator is contribution margin per unit - price minus variable cost per unit - not total cost. Breakeven volume equals fixed costs divided by contribution margin, and getting that denominator wrong throws off the entire analysis. Second, losing track of zeros is the single most common mental math mistake. The fix is mechanical: strip the zeros, do the core calculation, then add them back.

**Case math run-through, spoken out loud**

```
Goal: I need breakeven volume - units to cover fixed costs.
Assumptions: Fixed costs 2,000,000. Price 50. Variable cost 30.
Structure: Breakeven = Fixed costs / contribution margin per unit.
Intermediate: Contribution margin = 50 - 30 = 20 per unit.
Strip zeros: 2,000,000 / 20 = 200,000 / 2 = 100,000 units.
Sanity check: at 50 each that is 5,000,000 revenue, plausible for the market.
So what: they must sell 100,000 units to break even, which is X% of the market we sized - reachable or not, and that is the decision.
```

*Say each line as you work. Swap the breakeven example for your case numbers.*

The published speed target for a clean breakeven calculation is under 15 seconds, but speed is the last thing to build. Build the sequence first, then compress it. And never stop at the number. After calculating the market is worth some figure, the weak candidate says "so the market is that big" and waits. Every number ends with a business implication that moves the case toward the recommendation.

> **Tip:** The "so what" is part of the answer
>
> A number without an implication is an unfinished thought. Attach the consequence to every result: what it means for the client's decision, and what you would check next.

## Candidate-led versus interviewer-led: know which case you are in

The difference between a candidate-led and an interviewer-led case is who steers between questions, and getting it wrong wastes time or looks passive. In a candidate-led case you are expected to summarize findings, state your hypothesis, and propose the next area to investigate. In an interviewer-led case the interviewer does that steering for you, question by question.

The split tracks the firm, and it tracks how each firm builds cases. McKinsey cases come from a central team with a set script, so they tend to be interviewer-led. BCG cases are built by interviewing consultants from their own work, so they tend to be candidate-led. Bain, once known for candidate-led cases, is shifting toward interviewer-led.

| Firm | Predominant format | What you do between questions |
|---|---|---|
| McKinsey | Interviewer-led | Answer the posed question; interviewer steers |
| BCG | Candidate-led | Summarize, hypothesize, propose next area |
| Bain | Shifting to interviewer-led | Expect steering, but prep to drive too |

If you only know how to navigate interviewer-led cases, you may struggle to drive a candidate-led one. The reverse is easier. Prep the harder candidate-led mode, because the skill of summarizing and proposing the next step transfers down to the interviewer-led format, where it still reads as initiative. Confirm the format from the interviewer's opening prompt wording - whether they invite you to lead or signal they will direct you.

Note that the live case is often not the first gate. BCG online screens include the Consulting Career Assessment, the BCG Online Case known as the Casey chatbot, Pymetrics, and at some offices a Cognitive Test. Those are separate exercises; this playbook is for the live case the online rounds qualify you for.

## Close with a recommendation, not a recap

A strong synthesis is a 60 to 90-second recommendation that states a decision first, not a replay of your analysis. Use the four-part top-down formula based on Barbara Minto's Pyramid Principle: Answer, Reasons, Risks, Next steps. One taught variant extends this to a 5R close - recap, recommendation, reasons, risks, retain - but the core is the same: lead with the answer.

State the recommendation in one firm sentence. Give two or three reasons that reference specific case data, not generalities. Name one or two risks. Propose one next step. Lead with the branch you prioritized when you structured, because that is where your hypothesis lived and where your strongest evidence should be.

**The answer-first recommendation**

```
Answer: My recommendation is that [client] should [do X].
Reason 1: [case data point] shows [implication].
Reason 2: [case data point] shows [implication].
Reason 3 (optional): [case data point] shows [implication].
Risk: The main risk is [specific risk], which I'd monitor via [signal].
Next step: I'd [concrete next action] before committing further.
```

*Fill in from your case. Keep the whole thing to 60-90 seconds spoken.*

> The branch you prioritize while structuring is the sentence you should lead with when you close.
> </pull>
>
> ## How this goes wrong
>
> Most case rejections come from a handful of repeatable failure modes, and each has a tell you can check for in rehearsal. The dangerous ones are the false positives - habits that feel thorough while quietly sinking the case.
>
> - **Over-clarifying.** Feels thorough; reads as uncertain. More questions signal doubt, not rigor. Check: cap at two or three scope questions, and each must change your tree.
> - **Memorized framework.** Looks clean; is generic. A market-entry structure on a profitability case tells the interviewer you did not read the question. Check: does any branch name a client-specific driver?
> - **Overbuilding.** Five or six branches with sub-branches sound thorough but cannot be executed under pressure. Two or three branches explored properly beat five explored shallowly. Check: can you execute every branch in the time left?
> - **Silence misread as stuck.** Going quiet while structuring is correct, but unannounced it looks like freezing. Check: say "I'd like 90 seconds" before you go silent, and narrate reasoning when you present.
> - **Zero and unit errors.** A confident wrong number is worse than a slow right one, given the ~95% auto-reject rate on single math errors. Check: strip zeros, carry units, sanity-check the magnitude.
> - **Math with no "so what."** Calculating a figure and stopping leaves the analysis unfinished. Check: every number ends with a business implication.
> - **Rambling close.** A long recap feels thorough and scores badly. Check: answer first, 60 to 90 seconds, one risk, one next step.
> - **Treating all MBB the same.** Prepping only interviewer-led leaves you unable to drive a candidate-led case. Check: confirm the format from the opening prompt.

figure
kind: matrix
title: Reading your own performance under pressure
caption: The dangerous cases are top-left - habits that feel thorough while quietly failing the case.
x: Feels weak :: Feels thorough
y: Actually hurts :: Actually helps
quadrant: Visible struggle :: Honest, recoverable - narrate and regroup
quadrant: Over-clarify and overbuild :: The trap - looks rigorous, burns the clock and loses analysis time
quadrant: Awkward but tested branch :: Fine - a messy but real driver beats a clean generic one
quadrant: Answer-first close with "so what" :: The target - thorough and correct, keep doing it
```

## Rehearse against stage outputs, and get a real coach

The fastest way to improve is to rehearse one stage at a time against its defined "done," then run full cases on the clock. Memorizing more frameworks does not move the needle; hitting each stage's deliverable under time does.

Build a drill loop. Run the structuring step alone against a timer until you can present three to five tailored branches in 90 seconds. Run the math step alone until breakeven and market-sizing calculations come out with units and a "so what" every time. Then stitch the whole thing together and run full cases end to end, because the skill being tested is doing all of it on the clock, not any one piece in isolation.

#### Before you call a practice case done

- [ ] You restated the exact question verbatim before clarifying
- [ ] You asked no more than two or three clarifying questions, each one that reshaped your tree
- [ ] Your structure had three to five mutually exclusive branches with at least one client-specific driver
- [ ] You announced the pause, built silently, and presented top-line-first in 60 to 90 seconds
- [ ] You drove the priority branch first and gave an interim synthesis after each branch
- [ ] Every number came out with units and a business implication
- [ ] You checked magnitude and the breakeven denominator was contribution margin, not total cost
- [ ] Your close was answer-first, 60 to 90 seconds, two or three data-backed reasons, one risk, one next step
- [ ] You confirmed whether the case was candidate-led or interviewer-led and behaved accordingly

The highest-leverage practice partner is someone who has sat on the other side of the table. Mock cases with a former MBB consultant surface the tells you cannot see in yourself - the unannounced silence, the generic branch, the number with no "so what." Platforms like PrepLounge give you a practice-case marketplace and coach answers, and finding an ex-consultant directly is often faster. Refolk writes and tailors your application materials from your own history, and once you are in the loop, you can use it to find people worth practicing with.

Ask me this: `Find former McKinsey, BCG, or Bain consultants in New York who now coach case interviews.` - [run the search](https://www.refolk.ai/start?q=Find%20former%20McKinsey%2C%20BCG%2C%20or%20Bain%20consultants%20in%20New%20York%20who%20now%20coach%20case%20interviews.).

*Returns profiles of ex-MBB consultants offering case coaching, so you can line up mocks with someone who has graded cases.*

Keep the playbook current by re-checking two things that drift. First, firm formats: Bain is mid-shift toward interviewer-led, so confirm the current mode from your interviewer's opening rather than assuming. Second, online screens change - the Casey chatbot and assessment battery at firms like BCG get revised - so verify the current gate set with a recent candidate before you walk in. Everything else here is mechanics, and mechanics you own by rehearsing against the stage outputs until the whole case runs on time without improvising.

## Frequently asked questions

### How long should a case interview actually take, and how do I split the time?

A standard MBB-style case runs 30 to 45 minutes, and analysis should take about 60% to 70% of it. On a 30-minute case the published split is clarifying 1 to 3 minutes, structure 2 to 4 minutes, analysis 18 to 22 minutes, and recommendation 3 to 5 minutes. Coaches push the opening tighter so you are driving analysis roughly six to seven minutes in.

### How do I build a case interview structure without using a memorized framework?

Restate the objective in your own words, ask the one or two clarifying questions that would change your tree, then pick the pattern that fits the actual problem: profit drivers, market attractiveness, decision options, operational flow, or deal value. Build three to five mutually exclusive branches from scratch and name at least one client-specific driver. Interviewers are trained to spot a generic tree, so a market-entry structure pasted onto a profitability case reads as memorization.

### What is the difference between a candidate-led and interviewer-led case?

In a candidate-led case you summarize findings, state your hypothesis, and propose the next area to investigate yourself. In an interviewer-led case the interviewer does that steering for you. McKinsey cases are predominantly interviewer-led because they come from a central team with a set script, BCG cases are predominantly candidate-led, and Bain is shifting toward interviewer-led. Prep the candidate-led mode because the skills transfer down to the interviewer-led format.

### How much does one math mistake matter in a case interview?

A lot. Industry data cited by case-prep coaches suggests roughly 95% of the time a single mental math error results in automatic rejection, so arithmetic is a gate rather than one dimension among several. That is why narrated, sanity-checked math beats fast math. Strip the zeros, do the core calculation, add the zeros back, carry units throughout, and check the magnitude is plausible before you report.

### What should the closing recommendation in a case interview contain?

Follow the top-down structure based on the Pyramid Principle: Answer, Reasons, Risks, Next steps. State the recommendation in one firm sentence, support it with two or three reasons that reference specific case data, name one or two risks, and propose a next step. Keep it to 60 to 90 seconds. It is a decision, not a recap of everything you did.

### How do I calculate a breakeven point in a case interview?

Breakeven volume equals fixed costs divided by contribution margin per unit, where contribution margin is price minus variable cost per unit. The most common error is dividing by total cost instead of contribution margin, which throws off the whole analysis. With practice the calculation should take under 15 seconds. Always report the number with units and the business implication, not just the figure.

---

*From the Refolk guide library. I revise these guides rather than replacing them, so the current version is always at https://www.refolk.ai/candidates/guides/case-interview-prompt-to-recommendation*