Your case is not scored while you are in the room. It is scored later, often the same afternoon, in a room you are not in, by people comparing notes on candidates most of them never met. The only things that survive that meeting are the things someone wrote down: a sentence you said, a number you produced, the moment you changed course. Candidates prepare as if the sheet were an average of impressions. It is an evidence log, and the boxes with no evidence in them lose.
I have filled in several hundred of these sheets on two continents. What follows is what goes on them, why the sheet exists, what the research says about the four-minute folklore, and one case moment scored twice so you can score your own recordings the same way.
What the firms actually publish
Less than the prep industry implies. None of the three largest strategy firms publishes a scoring sheet, a weighting, or a pass mark. What they publish is a list of things the case is meant to reveal.
| Firm | What its own pages say the case assesses | What it does not say |
|---|---|---|
| McKinsey | A business case "to evaluate your analytical thinking and approach to solving complex problems," after a personal experience interview (McKinsey careers) | Weights, scales, how PEI and case combine |
| BCG | "Problem-solving and analytical skills" (BCG interview process); its 2018 brochure adds "structure a problem, prioritize and perform accurate analyses, synthesize your thoughts, and be creative" (BCG brochure, 2018) | Any numeric scale |
| Bain | Interviews "reward precision, but also creativity," with "the same questions for each role to reduce bias" (Bain interviewing); the case shows "how you think, structure problems, and build on ideas" (Bain case prep) | The form, the scale, the cut-off |
The lists overlap almost completely: structure, analysis, synthesis, creativity or judgment, and communication appear under some name at all three. Bain's line about identical questions is the only place any of them says out loud what all of them do: run a structured interview, same prompt, same exhibits, same dimensions, scored on paper. Everything beyond that, including the numbered 1-to-5 sheets that circulate on forums, is candidate folklore: some of it roughly right, none of it published, and forms change by office and by year.
Why there is a sheet at all
Dimension-by-dimension scoring is the difference between an interview that predicts anything and one that does not. A 2022 review of selection methods in the Journal of Applied Psychology ranked structured interviews first among every predictor it examined, with a mean validity of .42 against .19 for unstructured interviews (Sackett, Zhang, Berry, and Lievens, 2022; the authors restate the figures in Industrial and Organizational Psychology, noting that validity varies widely by interview quality). A badly run structured interview is a conversation with a form stapled to it.
The difference is whether the boxes get filled with evidence or with adjectives. Daniel Kahneman tells the founding story in Thinking, Fast and Slow: as a young officer he redesigned the Israeli army's recruit interview so that interviewers scored six traits one at a time, on factual questions, and were only allowed an overall judgment after the sixth box. It predicted later performance better than the old free-form chat. That is the design of every case sheet I have used: dimensions first, evidence in each, one holistic line at the bottom. The holistic line decides your fate. The boxes are what your interviewer will use to defend it.
The four-minute myth
The most repeated folklore is that the decision is made in the first four minutes and the rest is confirmation. It traces to a 1954 doctoral thesis, and a field study of 166 interviewers and 691 real interviews found something different (Frieder, Van Iddekinge, and Raymark, 2016). Interviewers reported deciding within the first minute 4.9% of the time and within five minutes about 30% of the time; 17.7% decided after the fifteenth minute, and 22.5% had still not decided when the interview ended (BPS Research Digest). Interviewers who asked consistent questions took longer to decide, not shorter.
Those were campus recruiters, not partners running cases, so treat the numbers as directional. In a 45-minute first-round case I usually have a lean by the first exhibit, around minute fifteen, and I change it roughly one time in five. What changes it is never a vibe; it is something specific I can write down. So a shaky opening is recoverable and a strong opening is not banked: "restructured cleanly after the first hint, unprompted" is a sentence a colleague can read out later, and "great start" is not.
What my sheet looks like when I fill it in
The form varies by firm and year; the boxes do not vary much. The 1-to-5 scale is what most forms I have seen use; the anchors are mine.
| Box | A note that helps you | A note that sinks you |
|---|---|---|
| Problem definition | "Restated the goal as margin, not revenue; asked the time horizon" | "Built a tree before asking what success meant" |
| Structure | "Three branches specific to the client; said which to open first and why" | "Profitability tree from a book; no prioritization" |
| Analysis | "Set up the calculation aloud, stated units, sanity-checked" | "Correct answer, silent for ninety seconds, no interpretation" |
| Exhibits | "Led with the one line that mattered; asked for the missing denominator" | "Read every bar left to right" |
| Judgment and creativity | "Two hypotheses, said which was more likely and what would falsify it" | "Listed six ideas; could not rank them" |
| Synthesis | "Answer first, two reasons, one risk, one next step, under sixty seconds" | "Summarized chronologically; recommendation in the last clause" |
| Communication | "Signposted every transition; I could write while listening" | "I had to ask for the number twice" |
| Bottom line | "Would staff on a Monday. Yes." | "Not distinctive. Would not fight for." |
Three things are not obvious from outside.
An empty box is not a neutral box. Nothing under Judgment defaults to a 3, and in the debrief a 3 with no note reads as "nothing happened." In a first round where roughly one candidate in three advances (my estimate; it varies by office and season), "nothing happened" is a no.
The notes are quotes, not adjectives. "Strong structure" is worthless an hour later when a colleague says the same candidate's structure was generic in their case. "Split cost into aluminum, labor, and yard overhead, and went to aluminum first because it is 60% of cost" wins that argument by itself.
The bottom line is not the average. Seven 4s with "would not want on my team" is a no. A 2 in Analysis with "made one slip, caught it, and gave the best synthesis I heard today" often goes through. Averages are what candidates compute. Interviewers compute whether they would sit next to you at a client for three months.
Ned's rule. If I cannot quote you, I cannot score you. Every box on my sheet needs one sentence you actually said; a box with an adjective and no quote gets a 3 in the room and loses the argument in the debrief.
The debrief, reconstructed
Firms do not publish how their decision meetings run, so this is the shape I have seen, not any one firm's process. Each interviewer arrives with a sheet per candidate. Someone reads the bottom lines around the table. Where they agree, the candidate takes about ninety seconds. Where they disagree, it becomes an evidence contest: what did the candidate say when you pushed on the number, did they build the structure themselves or after a hint, what was the recommendation in one sentence. The interviewer with quotes wins.
The arithmetic should change how you prepare. Eight candidates with two 45-minute interviews each, the first-round format BCG's brochure describes, is sixteen sheets and, in my experience, a debrief of 30 to 45 minutes: four to six minutes per candidate to represent 90 minutes of your interviewing. So the question to ask about every moment in a case is not "did that go well" but "what could the interviewer write down just now." If the honest answer is "nothing," that minute produced no evidence.
One case moment, scored twice
An invented case with invented numbers, so you can reproduce every step. Halvard Marine builds aluminum workboats in Norway. Two years ago revenue was EUR 240 million at a 12% operating margin. This year revenue is EUR 270 million at 7%. The interviewer hands over that exhibit and asks what you make of it.
Candidate A says: "So revenue is up but margin is down, which means costs have grown faster than revenue. This looks like a cost problem, so I would look at the cost side, probably materials and labor."
Candidate B says: "Revenue grew 12.5%, from 240 to 270. Operating profit went from 28.8 million to 18.9 million, so it fell by about a third while sales grew. If margin had held at 12% we would have 32.4 million this year, so the gap is 13.5 million, or five points of margin. I want to split that gap two ways: is it unit cost, most likely aluminum, or is it mix, meaning we sold more of a lower-margin boat? Can I get cost of goods by boat type?"
Check B's arithmetic: 240 × 0.12 = 28.8; 270 × 0.07 = 18.9; 270 × 0.12 = 32.4; 32.4 − 18.9 = 13.5; 13.5 ÷ 270 = 5.0 points. Nothing clever: one-line multiplications, said aloud with units, ending in a question that moves the case.
| Box | Candidate A: note and score | Candidate B: note and score |
|---|---|---|
| Analysis | "Correct direction. No numbers. Did not size the gap." 3 | "Sized the gap at 13.5m, five margin points; sanity-checked." 4 |
| Judgment and creativity | "Went to 'materials and labor' by default. No hypothesis on why now." 2 | "Two competing hypotheses, asked for the data that separates them." 4 |
| Exhibits | "Read the trend. Computed nothing from it." 3 | "Turned two margins into a profit gap in one step." 4 |
| Communication | "Fine. Nothing to quote." 3 | "Led with the growth rate, ended with a specific ask." 4 |
Neither candidate said anything wrong. A produced no evidence and will be represented in the debrief by the word "fine," the most dangerous word on a scoring sheet. B produced four quotable lines in about forty seconds, with no framework name and no rehearsed transition: a number, a comparison, a pair of hypotheses, and a question. Those four moves are available in every exhibit you will ever be shown.
Practice this today: score your own recording
You cannot see your interviewer's sheet, so build one. Run a full case, and before you read any feedback, draw the eight boxes from the table above and write only quotes in them: your words, verbatim, with the timestamp. Any box with no quote gets a 2, not a 3, because you will be generous with yourself. Write the bottom line as a staffing decision, "would I put this person on a client team on Monday," with one sentence of reason. Then fix the emptiest box, not the lowest score; an empty Judgment box is a habit, while a low Analysis score is usually one slip. A first-round pass, in my experience, produces at least one quotable line per box, roughly ten over a 25-minute case. Fewer than six and the interviewer is defending you with impressions.
CoachNed's live voice case with Ned at /interview takes about 18 minutes and ends with a seven-score debrief; do your quote-only sheet first, then compare which boxes the debrief filled that you left empty. With five minutes rather than twenty, the typed first rep at /start gives you three scored turns without an account. Then take the emptiest box to the matching drill: structure, synthesis, charts, or math, with the full library of 64 cases at /cases. Everything is open for seven days with no card; after that it is $120 for a recruiting season or $49 a month, and the quick-math drill stays free.
CoachNed is an independent product and is not affiliated with McKinsey, BCG, Bain, or any other firm named here.
Frequently asked questions
Do case interviewers use a numerical scoring sheet?
At every firm I have scored for, yes. The firms confirm the structured part without publishing the form: Bain says it asks every candidate the same questions to reduce bias, and BCG lists the abilities its case tests. Weights and pass marks are not public.
Can a strong recommendation make up for a math mistake?
Often, if you caught the mistake yourself and the recommendation was specific. "Made one slip, corrected it unprompted, best synthesis of the day" gets a candidate through. "Math wrong, never noticed, generic recommendation" does not.
How is a final-round case scored differently from a first round?
Same boxes, different bottom line. In a first round the closing question is roughly "could this person do the analysis." In a final round it becomes "would I put this person in front of my client next month," so judgment, synthesis, and how you handle being pushed carry more of the decision.
Sources
- Interviewing at McKinsey — the interview sequence and what the case evaluates. Checked 2026-09-24.
- BCG consulting interview process — the case assesses problem-solving and analytical skills. Checked 2026-09-24.
- BCG interview preparation brochure, November 2018 (PDF) — abilities the case tests; two 45-minute first-round interviews. Checked 2026-09-24.
- Interviewing at Bain — same questions for every candidate to reduce bias. Checked 2026-09-24.
- Bain case interview preparation — what the case reveals. Checked 2026-09-24.
- Sackett, Zhang, Berry, and Lievens, Journal of Applied Psychology, 2022 — structured interviews ranked first. Checked 2026-09-24.
- Sackett et al., Industrial and Organizational Psychology — the .42 versus .19 figures and their variability. Checked 2026-09-24.
- Frieder, Van Iddekinge, and Raymark, Journal of Occupational and Organizational Psychology, 2016 — the field study of decision timing. Checked 2026-09-24.
- BPS Research Digest on the four-minute myth — the decision-timing percentages and the 1954 origin. Checked 2026-09-24.
- Thinking, Fast and Slow by Daniel Kahneman — traits scored separately, holistic judgment last. Checked 2026-09-24.