Recovering Mid-Case When You Get Stuck or Go Wrong

One firm gives written permission to say you are stuck in a case

The instruction above sits on Kearney's case page, read in a Wayback snapshot at kearney.com because the live domain refuses our requests. The page continues: "Remember the simulation is designed to reflect the reality of consulting, including the fact that it can be complex and ambiguous."

Three things are worth pulling out of those two sentences.

First, the firm names the failure mode by name. Toughing it out is the thing it tells you to avoid, which means it has seen candidates do it and lose on it.

Second, it gives three exits rather than one: acknowledge, seek information, or change path. Those are different moves and they suit different situations. Acknowledging works when you have lost the thread entirely. Seeking information works when you know what you need. Changing path works when the branch you chose has stopped producing anything.

Third, it supplies the justification. The exercise is built to be ambiguous, so being stuck is inside the design rather than outside it. That reframe matters because the reason candidates tough it out is a belief that admitting difficulty is itself the failure. One firm has published the opposite, in writing, on its own careers site.

Accenture supports the same move from the other side. Its careers blog at accenture.com states: "The interviewer is your biggest asset in the room. They have the information you need to "solve the case" successfully." Oliver Wyman puts it as a stance: "Think of your case interviewer as your client. They want you to solve the problem, and can help you to do so. Work together."

The two case interview pitfalls BCG names, pointing opposite ways

BCG's case preparation page at careers.bcg.com publishes six things to avoid: Panic Under Pressure, Rush Your Response, Overcomplicate Solutions, Ignore Feedback, Stay Silent, Interrupt.

Look at what that list does. Rush Your Response and Stay Silent cannot both be fixed by slowing down or by speaking more. Interrupt and Stay Silent are the same pair from the other side. The firm is describing a rhythm rather than a speed, and the fault it is warning about is being out of step with the conversation.

Ignore Feedback belongs in the same family. The same page states that "there isn't always a single 'right' answer", so a nudge from the interviewer is information rather than a verdict, and a candidate who keeps going after receiving one has chosen to discard data.

Our reading of the six, and it is our reading rather than a ranking the firm supplies: four of them are about the handoff between you and the interviewer, and two are about the content of what you say. That puts most of the published risk on the interaction, which is exactly the part candidates practice least when they rehearse alone.

Recomputing out loud after a unit error in a case interview

The most recoverable mistake in a case is arithmetic, and the most expensive one is a unit. A firm's own blog supplies the example.

On L.E.K.'s careers blog at lek.com, a first-year associate in the London office describes a market sizing where he treated each person as marrying separately rather than two people per wedding. The interviewer credited the method and pointed out the error. He adds: "The key was not necessarily getting the right answer but demonstrating structured problem-solving skills." That is a single account from one person about his own experience, published by his employer, and it is not evidence about how the firm grades.

The factor is two, and here is what two costs. These figures are our own illustration with stated assumptions. The firm publishes no numbers for this anecdote.

Verified by hand on 30 September 2026: 800,000 divided by 2 is 400,000, and 400,000 times 25,000 is 10,000,000,000, so 10 billion dollars. The wrong reading gives 800,000 times 25,000, which is 20,000,000,000.

The recovery sequence has three moves, and the order matters.

1. Name the error before you fix it. I have been counting people rather than weddings, so my number is double what it should be. Six seconds, and it tells the interviewer you found it yourself.
2. Redo only the affected step. The spend per wedding is unaffected, so the whole correction is one division by two.
3. Restate the conclusion, because it may have changed. A 10 billion dollar market supports a different recommendation than a 20 billion dollar one, and stopping at the corrected number leaves the work half done.

Three sentences that buy thirty seconds when you get stuck

Being stuck in silence and being stuck out loud are scored differently, and both L.E.K. and Bain publish something about the pause.

Bain's case interview page lists nine general tips, and two of them cover this ground: think aloud even when you are unsure, and take a pause to gather your thoughts. Those two run against each other in real time, which is why a sentence that covers the pause is useful.

Three that work, and they are ours rather than any firm's wording.

1. Let me take twenty seconds to lay this out before I answer. Announced silence is not the same event as unannounced silence. One is a plan, the other looks like a stall.
2. I want to check one thing before I commit to that number. This buys a recomputation without conceding the answer, and it is honest.
3. I am not getting traction on this branch. Can I tell you what I would look at instead? This is Kearney's third exit, said in a way that keeps you leading.

None of these works twice in the same case. Used once, a pause is deliberate. Used three times, it is the rhythm fault BCG named.

Changing your mind without losing the room

L.E.K. publishes both sides of this and resolves neither, which is the honest state of the problem.

Step three of its seven is Stick to your approach, glossed as "Outline how you are thinking and move methodically to address each issue, point-by-point". Step six is Come up with an answer: "Commit yourself to a course of action, but show flexibility and a willingness to rethink in the face of new data."

Kearney publishes the same tension in one breath: "Be coachable, soliciting feedback and integrating it quickly. At the same time, be firm about the things you believe in."

Neither firm gives a rule for choosing. Our reading, and we label it as ours: the trigger for changing course is new data, and the trigger for holding course is discomfort. If a number arrived that your structure cannot absorb, change the structure and say why. If nothing new arrived and the branch is merely hard, work it.

Said out loud, a legitimate change sounds like this. The margin data you just gave me says the rate is flat, so cost mix is not where this is coming from. I want to move to the fixed cost base instead. That sentence names the new fact, names the conclusion it forces, and names the next move. A change without those three parts reads as abandonment.

What a recovery costs in our rubric

The weights below are ours, read from the grading code in src/lib/grading. No consulting firm publishes a rubric.

Practicing the recovery on purpose

Nobody rehearses recovery, because rehearsal usually means running cases until they go well. Invert that.

1. Run a case and deliberately choose a weak branch. Work it for four minutes, then execute Kearney's third exit out loud and change path.
2. Have someone hand you a contradicting number halfway. Then practice the three-part change sentence: new fact, conclusion forced, next move.
3. Introduce a unit error into your own sizing and catch it. Say the correction out loud, including the restated conclusion.
4. Time your pauses. Twenty seconds announced is fine. Forty seconds unannounced is the pitfall BCG named.
5. Do one case where you say I am stuck on purpose. The first time you say it, you will find out it is survivable, which is the whole point of the Kearney line.

Our structure drill will accept a revised structure and score it fresh, which is the closest a text tool gets to practicing a change of course. Our math drill reports Working separately from Answer, so you can see what a corrected chain is worth. Both are free and neither needs an account. Neither can simulate an interviewer pushing back in real time, which is the part that needs a person.

For the composure side of the same problem, see case interview pressure. For what to do when the interviewer challenges a conclusion you still believe, see handling pushback. For the firm's own page in detail, see the Kearney case interview guide.