The arithmetic error is not what goes on the sheet. The next twenty seconds are. In a typical first-round panel of eight candidates, Ned's estimate is that five or six make at least one recoverable mistake; the two or three who stall, apologize, or defend the wrong path are the ones who drop a notch, not the ones who said "nine million" when the product was nine billion.
The thesis most prep advice never states: interviewers score the correction, not the error. A clean self-catch on volume units can raise your quantitative mark. Digging in for ninety seconds after the interviewer hands you the exhibit insight can erase a structure that was otherwise fine. Recovery is coachability under the clock, and the firms publish the expectation even if they never use the word "recovery."
By the end you will have four exact lines an interviewer wants to hear, a scoring heuristic for what each recovery buys you, and one worked transcript you can rehearse aloud.
What the sheet rewards after you slip
I cannot quote any firm's confidential scorecard. I can tell you what I listen for when a candidate hits a wall, and how that maps to language the firms put on their own careers pages.
Bain's case interview guidance is blunt on the point: "Listen and adapt: Pay attention to interviewer cues and adjust accordingly," and "Think aloud: Share your reasoning and calculations, even if you're unsure." BCG's case interview preparation page says there is not always a single right answer and that what matters is "how you approach the problem and the quality of your reasoning." McKinsey's interviewing page frames the case as a problem-solving conversation about a client scenario, not a quiz with a locked key. A public McKinsey careers story from Yan Yan describes incorporating interviewer feedback on mental math between rounds and improving immediately; that is the same muscle mid-case, compressed into one turn.
| What happened | What I score up | What I score down |
|---|---|---|
| Math error | Name the break, fix units, state the new implication in one breath | Apologize twice, restart the whole case, or hide the miss |
| Wrong structure | Restate the decision, replace one branch, ask the next useful question | Defend the framework, or rebuild twelve MECE boxes from scratch |
| Blank | Anchor on the objective and one known fact, then propose one next step | Silence longer than ten seconds, or "I don't know" with no path |
| Missed exhibit | Say what you missed, re-read the cell that matters, update the hypothesis | Argue with the chart, or summarize every row without a so-what |
Ned's rule. After a mistake, your next sentence must contain a diagnosis and a next move. If it contains only regret, I stop adding points until you give me something I can score.
The recovery line for a math error
Say the break, not the emotion. The interviewer already saw the wrong number. What they need is proof you can find the crack and still drive the decision.
Line to use: "That result is off by an order of magnitude; I divided monthly revenue by twelve when the prompt was already monthly. Corrected monthly contribution is $1.2 million, which puts break-even share at about 8% of the local market, so I would test whether we can capture that before I recommend a build."
Three parts, always: name the break (units, period, or formula), state the corrected figure, and say what changes for the recommendation. If the correction does not change the decision, say that in eight words and move on: "The rounding error does not flip the go or no-go."
What I do not want: "Sorry, I'm bad at math under pressure." That is autobiography. It is not analysis.
The recovery line for a wrong structure
A wrong structure is usually a generic split that never touches the client's actual decision. Profitability into revenue and cost is often true and still useless.
Line to use: "My structure is accounting-led and misses the mechanism behind the decision. The question is whether HarborFresh should keep the same-day delivery service. I would replace the cost branch with three tests: whether delivery orders are incremental or shifted from stores, contribution per delivery order net of last-mile cost, and whether capacity caps volume before the service turns profitable. Can I start with incrementality?"
Preserve anything that still maps to the objective. Replace only the broken branch. Ask permission to start on the highest-leverage test. BCG's older interview preparation brochure still lists structuring, prioritization, accurate analysis, and synthesis as what the case tests; a mid-case rebuild that re-aims at the original question is synthesis applied early, not a restart.
The recovery line for a blank
Blanks feel like personality failure. They are usually a lost objective. Re-anchor before you invent a framework.
Line to use: "I want to restate the decision before I choose the next analysis. HarborFresh needs a yes or no on keeping same-day delivery, and we already know orders are up 18% but contribution is unknown. The missing piece that would reverse a keep decision is negative contribution after last-mile cost. I would ask for unit economics on a delivery order next."
If you truly have nothing, buy ten seconds with a mechanical move: write the prompt verb, write one number you already have, and speak the next question. Bain's published tip to pause and gather thoughts is permission, not a weakness.
The recovery line for a missed exhibit insight
This is the one that looks coachable or not. The interviewer often designed the exhibit so one cell flips the story. Missing it once is common. Missing the hint after they point at the cell is expensive.
Line to use: "I underweighted the mix shift in column C. Same-day volume is up, but grocery AOV on those orders is $42 versus $68 in-store, so revenue growth is volume-led and margin-dilutive. That revises my hypothesis from 'demand validates the service' to 'demand is real but unit economics may not clear.' I would check last-mile cost per order before recommending keep."
Name the cell or column. State the revised hypothesis. Name the next analysis. Do not re-narrate the entire table.
Worked transcript: HarborFresh, minute 14
Prompt (minute 1): HarborFresh, a regional grocer with $420 million in annual revenue, launched same-day delivery in three cities eighteen months ago. Should leadership keep, expand, or cut the service?
Candidate has just said contribution is healthy because orders grew 18%. The interviewer slides an exhibit: delivery AOV $42, in-store AOV $68, last-mile cost $9.50 per order, basket gross margin 28%.
Weak recovery (do not sound like this):
Candidate: "Oh. Sorry, I didn't notice that. Orders are still growing though, so I think we should expand."
Interviewer: "Walk me through the unit economics."
Candidate: "Right, so AOV is lower… and then costs… I'm not sure. Maybe we should keep it for now?"
What I write: missed insight, soft on numbers, no correction that changes the recommendation.
Strong recovery:
Candidate: "I underweighted the AOV gap. Delivery AOV is $42 against $68 in-store. At 28% gross margin, delivery gross profit is about $11.80 before last-mile. Last-mile is $9.50, so contribution is roughly $2.30 per order, thin but positive on the current basket. That revises my hypothesis from 'demand proves keep' to 'keep only if we can raise AOV or cut last-mile without killing volume.' I would test whether a $10 delivery fee or a $55 minimum order clears another $3 to $4 of contribution before I recommend expand."
Interviewer: "What if last-mile is $14 in the smallest city?"
Candidate: "Then contribution goes negative by about $2.20 per order in that city. I would cut or redesign there and keep the two denser cities where last-mile is closer to $9.50, pending a fee test."
Reproduce the math: $42 × 0.28 = $11.76; $11.76 − $9.50 ≈ $2.26; at $14 last-mile, $11.76 − $14 ≈ −$2.24. The recovery is scored on the revision and the conditional recommendation, not on the fact that volume growth alone was a weak first pass.
How I score the four recoveries in practice
Think in notches, not perfection. One clean recovery rarely creates an offer by itself. It often keeps you in the conversation when the alternative is a polite no.
| Recovery type | Time budget | Passes if | Fails if |
|---|---|---|---|
| Math | 20–40 seconds | Corrected number plus decision implication | Corrected number with no so-what |
| Structure | 30–45 seconds | Decision restated, one branch replaced, next question asked | Full framework reboot or defense of the original tree |
| Blank | 15–25 seconds | Objective + one fact + one next step | Silence, or a new framework with no link to the prompt |
| Exhibit | 25–40 seconds | Named miss, revised hypothesis, next analysis | Chart tour without updating the call |
If you recover well and later find the recommendation still soft, stop recovering and close. A second apology at minute 28 does not buy points. Synthesis does. For how to land the last minute, see the synthesis guide.
Drill the four lines today
Open a case you have already run, pause at a deliberate mistake, and say the matching line out loud once without notes. Then run one live case and force a mid-case self-correction even if you did not need one: invent a wrong intermediate number, catch it, and finish the decision.
If you want the correction scored by someone who interrupts the way an interviewer does, use the voice case with Ned (CopperCup, about eighteen minutes, seven-score debrief). Everything on CoachNed is open for seven days with no card; then $120 for a recruiting season or $49 a month. The math drill and structure drill are the shortest places to rehearse the first two lines above.
Frequently asked questions
Should I apologize after a case interview mistake?
Once, briefly, if it costs nothing: "That was the wrong period; correcting now." A second apology is noise. The interviewer is waiting for the corrected implication, not for remorse.
Does one math error end a McKinsey or BCG case?
No firm publishes a one-error reject rule. BCG's own materials stress quality of reasoning over a single right answer, and Bain explicitly asks you to think aloud even when unsure. A hidden or defended error is worse than a named one that you fix and use.
What if the interviewer stays silent after I recover?
Finish the thought, state the implication, and propose the next analysis. Silence is often a test of whether you can drive without constant prompts, not disapproval.
How do I recover if I misunderstood the objective from the start?
Say that the objective was wrong, restate the prompt verb, and rebuild only the branches that serve the real decision. Do not salvage a tree aimed at the wrong question.
Is recovery the same as coachability?
In the room, yes: listening to a cue, updating your view, and showing the change in the next sentence. Firms describe that as listen-and-adapt or as incorporating feedback; they rarely label a mid-case script "recovery," but that is what the sheet rewards.
Sources
- Preparing for the Case Interview | Bain & Company — listen-and-adapt and think-aloud guidance for Bain cases. Checked 2026-09-24.
- Case Interview Preparation | BCG Careers — BCG's public statement that reasoning quality matters more than a single right answer. Checked 2026-09-24.
- Interviewing at McKinsey | McKinsey & Company — official description of the problem-solving (case) interview as a client-scenario conversation. Checked 2026-09-24.
- BCG Recruiting Interview Preparation Brochure (Nov 2018) — structuring, analysis, creativity, and synthesis as case skills. Checked 2026-09-24.
- Taking interviews step by step | McKinsey Careers Blog (Yan Yan) — public account of incorporating interviewer feedback on math between rounds. Checked 2026-09-24.