PartnerRound- 01AnswerOne sentence, a clear call
- 02ReasonsThree, each with a number
- 03RisksOne or two, with a mitigation
- 04Next stepsConcrete actions, with timing
By the end you can
- Deliver a recommendation in 60 seconds that leads with the answer
- Build it as a pyramid: one answer, three reasons, numbers under each
- Give a crisp interim synthesis whenever the interviewer asks where you are
- Make a clear call when the data is mixed, without retreating to 'it depends'
On this page
01The moment that decides the case
Near the end, the interviewer will say something like 'The CEO has just walked in. What do you tell her?' You have about a minute. It is the last thing they hear, and often the clearest evidence of how you think.
A recommendation is not a recap. The CEO does not want the order in which you did the analysis. She wants the answer, why you believe it and what to do next.
Most candidates who stumble here did the analysis well. They lose marks because they tell the story chronologically, hedge the answer or forget the numbers they worked so hard for.
02The pyramid principle
The pyramid principle, set out by Barbara Minto at McKinsey, says: start with the answer, then give the arguments that support it, then the facts behind each argument. The listener can stop at any level and still have the point.
Three rules keep the pyramid honest. Each reason supports the answer directly. The reasons do not overlap. Every reason carries at least one number from the case.
03The 60-second script
Sixty seconds is roughly 130 to 160 spoken words. When you sense the end coming, write the skeleton on a fresh sheet: the answer at the top, three short lines for reasons, one for the risk, one for next steps. Then speak from that page, not from your analysis.
If the interviewer offers time to prepare, take it. If not, ask: 'Could I take 30 seconds to pull this together?' It is almost always granted.
- Name the client and the decision in your first sentence.
- Round the numbers: '$54m', not '$53.8m'.
- Say 'three reasons', then give exactly three.
- Stop after the next steps. No 'so yeah, that's about it'.
04Worked example: should FreshCo launch delivery?
Worked example
FreshCo: from findings to a 60-second answer
FreshCo runs 120 grocery stores in one region, each selling about $15m a year, so $1.8bn in total. It asks whether to launch home delivery. Here is what the case has established.
The data is mixed. On its own, delivery misses the payback hurdle. As a defence it looks essential, but the defensive number rests on the assumption that the 3% loss spreads. A strong answer makes a call and tests the weakest assumption.
Weak
So, we started by looking at the market, which is about $1bn and growing quickly. Then we looked at the economics: $60m of revenue and $6m of contribution, but cannibalisation takes $3m, so the payback is five years, above the hurdle. Then we saw store sales fall where rivals deliver. So overall it's a mixed picture, and it really depends on FreshCo's appetite for risk.
Strong
FreshCo should launch home delivery, starting with a six-month pilot in the 30 stores where rivals already deliver. The market is moving, standing still is already costing it sales, and a pilot proves the economics before it commits $15m.
Why: The weak version makes the CEO wait for an answer that never comes: 'it depends' is not a recommendation. The strong version gives the call and all three reasons in two sentences. Everything after that is supporting detail she can interrupt.
05Interim synthesis mid-case
Interviewers often stop you mid-case: 'Can you summarise where we are?' or 'What do we know so far?' It checks that you have not lost the thread. Answer in 15 to 30 seconds.
- 01What we knowTwo or three findings, with numbers
- 02What it meansYour current hypothesis
- 03What's nextThe next analysis, and why
For FreshCo, halfway through: 'So far we know the market grows from $1bn to $1.5bn, and delivery on its own pays back in five years against a four-year hurdle. So the standalone case is marginal. Next I'd like to check whether not launching costs FreshCo store sales, because that could change the answer.'
Use it unprompted too. A one-line synthesis after each major step is the easiest way to show drive in a candidate-led case, and it makes the final recommendation almost write itself.
06Recommending when the data is mixed
Real cases rarely point one way, and interviewers often design them not to. 'It depends' is not an answer. Make a call, say what it depends on, and propose how to find out.
- On balance: 'On balance, I'd recommend X, because the strongest evidence, Y, outweighs Z.'
- Conditional: 'I'd recommend X, provided Y holds. If it doesn't, I'd switch to Z.'
- Staged: 'I'd recommend a pilot of X, which tests our weakest assumption, Y, for a fraction of the cost.'
Name the single assumption that would flip your answer. It shows you understand the case well enough to know where it is fragile, which is exactly what a partner wants to hear.
07Check yourself
One test before you speak: could the CEO repeat your recommendation to her board in one sentence, with one number? If not, cut until she could.
Check yourself
Which is the best first sentence of a final recommendation?
Check yourself
Mid-case, the interviewer asks 'Where are we?'. What is the best shape for your answer?
Check yourself
The evidence is mixed, and the investment is large and hard to reverse. Which recommendation fits best?
Cheat sheet
The lines worth rereading the night before.
- 1Answer first: the CEO hears your call in the first sentence.
- 2Answer, three reasons with numbers, the main risk, next steps: about 60 seconds, about 150 words.
- 3Each reason supports the answer, none overlap, and every one carries a number.
- 4Write the skeleton on a fresh page before you speak, then speak from it.
- 5Interim synthesis: what we know, what it means, what's next, in under 30 seconds.
- 6Mixed data still gets a call: on balance, conditional or staged. Never just 'it depends'.
- 7Name the one assumption that would flip your answer, and the test for it.
Now drill it
Reading builds the model; drills make it automatic. Start with the first topic, easiest drills first.
Questions candidates ask
How long should a case interview recommendation be?
About a minute, or roughly 130 to 160 spoken words. That is long enough for an answer, three reasons with numbers, a risk and next steps, and short enough that the interviewer never waits for the point.
Can I ask for time before giving my recommendation?
Yes. 'Could I take 30 seconds to pull this together?' is normal and almost always granted. Use the time to write the skeleton: answer, three reasons, risk, next steps.
What if I haven't finished the analysis when asked to recommend?
Recommend anyway, based on what you know, and say what you would check next: 'Based on what we've seen, I'd recommend X. The open question is Y, which I'd test first.' Interviewers sometimes ask early on purpose, to see whether you can commit.
What is the pyramid principle?
A way of structuring communication with the answer at the top, supporting arguments beneath it and facts beneath each argument. Barbara Minto developed it at McKinsey, and it underpins how consultants write and present.
Should I mention risks in a case recommendation?
Yes, briefly: one or two risks, each with how you would manage it. Risks show judgement, but they belong after the answer and the reasons, not before them.