Method
How to do a case interview, step by step
Clarify · Structure · Analyze · Synthesize · Recommend.
Step 1
Clarify
Repeat the decision. Ask only what changes the tree.
Step 2
Structure
A MECE tree you could steal. Two levels is enough.
Step 3
Analyze
Equation before numbers. Then the so-what.
Step 4
Synthesize
What we know, what it implies, what we do next.
Step 5
Recommend
A close a partner could take into the client dinner.
Step 1 · 30–45 seconds
Clarify
The interviewer already knows the prompt. You are proving you heard the decision, not buying time. Repeat it in one sentence, then ask only what would change the tree.
How to structure it
- Decision. What yes/no or A-vs-B are we answering?
- Metric. Profit, share, NPV, go / no-go — pick one.
- Horizon. 90 days and three years are different trees.
- Constraints. Brand, regulation, capacity, a union, a retailer you cannot anger. One or two. Then start.
Do not ask the company history, the market size, or “is there any other context?” If they say “what do you want to know?”, you already waited too long.
“We’re deciding whether BrightTrail should replace blanket promotions with retailer-specific hurdle rates. Success is higher operating profit without losing priority shelf space. Any retailer, brand, or capacity constraints I shouldn’t break?”
Usual miss. Clarifying six things. Two that change the structure. The rest you learn from exhibits.
Step 2 · 60–90 seconds
Structure
Write a tree the interviewer could steal. Two levels is enough. Each leaf must be sizeable: a number could sit on it. Reciting “the 3Cs” is not a structure. Naming the decision, then hanging customized branches off it, is.
How to structure it
- Write the identity first. Profitability is profit = revenue − cost. Growth is where the extra dollar comes from. Entry is attractiveness × can-we-win × how × economics.
- Split each side once. Revenue: price, volume, mix. Cost: variable, fixed, one-time.
- Rename the leaves in the client’s language. Trade spend is price. Store count × velocity is volume.
- Announce the first branch you will test, and why. That is the hypothesis: “I think X because Y, so I want exhibit Z.”
On the page
BrightTrail profit tree
- Revenue
- Price / trade spend / discounting
- Volume — distribution and velocity
- Mix — core vs adjacent vs tail
- Cost
- Variable — COGS, slotting
- Fixed — opex
First branch: realized price. The founder already named trade spend. Volume is the alternative, not the default.
“Profit is revenue minus cost. On revenue I’ll split price, volume, and mix — trade spend sits in price. Cost is variable versus opex. Given they grew the top line and the founder is worried they bought it, I’d start on realized price, then volume.”
Usual miss. A framework name with no leaves. If a number cannot hang on the branch, it is not a branch yet.
Step 3 · most of the case
Analyze
Exhibits and math are the same job: turn a table into a decision. Do not read every cell. Name the biggest swing in dollars, then what it implies.
How to structure it
- Equation first. Write profit = revenue − COGS − opex before you touch a number.
- Units. $m versus $. Percent versus points. Say them.
- Compute. One or two significant figures. Interviewers grade the path, not the eighth digit.
- So-what. Which side of the decision just flipped? What exhibit do you want next?
| Move | BrightTrail math |
|---|---|
| Base operating profit | 107 − 58 − 23 = $26m |
| Push volume: −4% price, +8% volume | Revenue 107 × 0.96 × 1.08 ≈ $110.9m. COGS 58 × 1.08 ≈ $62.6m. Opex stays $23m. |
| New operating profit | 110.9 − 62.6 − 23 = $25.3m |
| So-what | The promo destroys about $0.7m. Target trade spend instead of buying broad volume. |
“Equation is operating profit equals revenue minus COGS minus opex. Today that’s $26m. If we cut price 4% and volume rises 8%, revenue goes to about $110.9m, COGS to $62.6m, and profit to $25.3m. The broad promo destroys about $0.7m. I want retailer-level lift next so trade spend goes only where it clears a hurdle.”
Usual miss. A correct number with no implication. That scores like a calculator. Also: mixing $m and $ halfway through.
Step 4 · 20–30 seconds, more than once
Synthesize
A synthesis is not a recap. It is an interim close so the interviewer knows you are driving. You should be able to do this every two minutes: what we know, what it implies, what we do next.
How to structure it
- What we know. One fact. Prefer a number.
- What it implies. Which branch of the tree is now leading.
- What we do next. The next exhibit, the next cut, or the close. Then stop talking.
“We know a broad promotion takes operating profit from $26m to about $25.3m. That implies trade-spend quality, not lack of volume, is the issue. Next I want retailer-level incrementality so we can set hurdle rates.”
On the page
Core accounts are 44% of the book at $24 unit contribution. Growth is $19. Tail is $12. Once price is the driver, stop funding low-return promotions in the tail.
Usual miss. Restating the last three exhibits. If you could delete the number and the sentence still works, it is not a synthesis.
Step 5 · 45 seconds
Recommend
The close is a decision a partner could take into the client dinner. Answer first. Then two supports, one of them a number. Then a risk with a mitigation, then what happens Monday. Then you stop.
How to structure it
- Answer. Protect price / push volume / enter / do not enter. No throat-clearing.
- Two supports. One is a number from the case. The other is the mechanism (mix, cost, capacity).
- Risk + mitigation. Name who gets hurt and the test that bounds it.
- Monday next step. A person, a file, a store test. Not “do more analysis.”
“Replace blanket promotions with retailer hurdle rates. Operating profit is $26m on $107m; a 4% cut with 8% more volume takes profit down to about $25.3m. Freeze low-return trade events and shift support toward the $24-contribution core rather than the $12 tail. Risk is retailer pushback, so run a two-banner test. Monday: pull the trade calendar and score the next six events on incremental contribution.”
Usual miss. Ending on “it depends” or a tour of the tree. The interviewer asked for a recommendation. Give them one.
Then sit it
A page is not practice
BrightTrail in learning mode walks these five checkpoints before you run it live. Juniper Cup is the same loop on a growth prompt. The free method lesson is eight minutes if you want the 60-second version first.