Most mock hours die as theater. Two friends trade cases, say "that was solid," log another tick, and wonder why first-round scores barely move.
A mock is only worth the hour if it is scored against a stable rubric and produces one named leak you can drill before the next session. Volume without that loop is rehearsal of your current habits, not improvement. After this piece you should be able to run a mock as candidate or interviewer, fill a printable score sheet in under five minutes, and decide how many scored mocks belong in your calendar versus how many isolated drills.
McKinsey's own prep page frames the problem-solving interview as a fast discussion of structure, facts, conclusions, and recommendations on a client-style scenario (Interviewing at McKinsey). BCG's public case guide stresses structured thinking, clear communication, and showing your reasoning while you work (BCG Case Interview Preparation). Your mock has to train those behaviors under a clock, or it is just chatting about business.
What the sheet actually rewards
I stop listening when a candidate narrates activity without committing to a decision. Interviewers are not grading effort. They are grading whether you can be trusted in a client room for the next three months of staffing.
Use the same six dimensions every time. Score each 1–5. Write the leak in one line. That is the whole governance model.
| Dimension | 1 (leaking) | 3 (passable) | 5 (hire signal) |
|---|---|---|---|
| Structure | Laundry list of MECE-ish buckets | Covers the ask, some overlap | Hypothesis-led map keyed to the decision |
| Prioritization | Boils the ocean | Picks a start, drifts | Names the 20% that moves the answer |
| Math | Silent errors, no units | Correct but slow | Clean path, check, business meaning |
| Exhibits | Reads cells aloud | Pulls one insight | Ties chart to next question |
| Communication | Rambling, buried lead | Clear enough | Top-down, interruptable, collaborative |
| Recommendation | Soft hedge, no risk | Answer plus reasons | Answer, 2–3 supports, risk, next test |
Print that table. Keep one copy for the interviewer. The candidate never sees scores until the case is over. Mid-case coaching kills the signal you came to measure.
How to run the hour as the candidate
Treat the partner like a stranger who has a score sheet, not like a study buddy.
- Agree the rules before the clock: 25–35 minutes of case, five minutes of fit optional, ten minutes of scored debrief. No mid-case tips.
- State your structure out loud. If you cannot say it in 45–60 seconds, you do not have one yet.
- Drive the analysis. Ask for data. Narrate math. When interrupted, answer the question, then return to your map.
- Force a recommendation two minutes early. Answer first. Reasons. Biggest risk. What you would check next week.
- Sit through the scores without defending. Your job in the debrief is to name the leak, not to litigate the grade.
If you leave with "I need to be more structured," you failed the debrief. Translate it: "Opening map had overlapping cost and ops buckets; tonight I write five structures with a forced MECE check."
How to run the hour as the interviewer
Your job during the case is to be a fair examiner. Coaching belongs after the clock.
Hand out one prompt. Stay in role. Push once on a fuzzy assumption. Interrupt unclear speech once. Withhold the exhibit until they ask a question that earns it. Ask for the recommendation before they feel ready. Then switch seats and fill the sheet.
Do not invent a new rubric each week. Consistency is the point. If last Tuesday's "structure" meant something different from this Tuesday's, you cannot tell whether the candidate improved or you just felt nicer.
Keep feedforward brutal and short. One primary leak. One drill. One date for the next scored mock. Everything else is noise dressed as thoroughness.
Ned's rule. If the debrief names more than one leak, you did not prioritize. Fix the leak that would have cost the hire vote first; the rest can wait until that score moves.
Worked example: scoring one bad hour
Prompt (invented client): HarborLine Logistics runs 420 regional delivery routes. Operating margin fell from 8.2% to 5.1% in twelve months. The CEO wants to know whether to cut low-density routes or renegotiate last-mile contracts.
Candidate A spends nine minutes listing Porter-ish buckets, then asks for "any data on costs." Margin math: revenue $180m, prior profit $14.8m, current profit about $9.2m, so a $5.6m hole. They never separate volume from rate. Recommendation: "Maybe both, depending."
Candidate B opens with a driver tree: margin = price × volume − fixed − variable. Hypothesis: variable cost per stop rose faster than density fell. Asks for stops, cost per stop, and contract rates. Shows the $5.6m hole is roughly 60% rate inflation and 40% density. Recommends renegotiate the top three carriers first; keep thin routes only if contribution stays positive after the new rate.
| Dimension | Candidate A | Candidate B |
|---|---|---|
| Structure | 2 | 5 |
| Prioritization | 2 | 4 |
| Math | 3 | 5 |
| Exhibits | 2 (none used well) | 4 |
| Communication | 3 | 4 |
| Recommendation | 1 | 5 |
| Total /30 | 13 | 27 |
A's named leak is recommendation under time pressure. Next drill: three timed closes, 90 seconds each, answer-risk-next-test only. B's leak is exhibit speed. Next drill: ten chart reads with a forced "so what" sentence. Same case count for both; different payoff.
Reproduce the margin math yourself: 0.082 × 180 = 14.76; 0.051 × 180 = 9.18; gap ≈ 5.6. The skill is not the arithmetic. It is forcing the gap into drivers before you prescribe a fix.
How many mocks actually move scores
Firms do not publish a "do N mocks" prescription. What exists is the deliberate-practice literature, and it is stricter than recruiting folklore.
Ericsson, Krampe, and Tesch-Römer (1993) defined deliberate practice as effortful, feedback-rich activity designed to improve a specific aspect of performance—not mere repetition or playful rehearsal (Psychological Review paper via DOI). Their Berlin violin studies linked higher attainment to more accumulated deliberate practice, with expert preparation typically spanning many years, not a cram month. They also argued that high-concentration practice hits an effort constraint: effective daily sessions are limited, often on the order of roughly an hour early on and a few focused hours at most for elites.
Translate that to case prep without pretending consulting has a published dose-response curve:
| Practice type | Counts as deliberate practice? | Typical use |
|---|---|---|
| Scored mock + one named leak + drill | Yes | 2–4 per week in peak season (Ned's range) |
| Unscored "friendly" case | Mostly no | Warm-up only |
| Isolated drills (structure, math, charts) | Yes, if scored against a criterion | Daily 20–40 minutes |
| Watching someone else's mock | Weak unless you score it live | Optional, capped |
Ned's estimate from scoring hundreds of practice and live cases: candidates who run eight to twelve scored mocks with a stable sheet and intervening drills usually move one to two points on the weak dimensions. Candidates who log thirty unscored cases often plateau. That is not a peer-reviewed consulting study. It is what the sheet shows when the rubric does not change.
If you only have two weeks, prefer six scored mocks and twenty short drills over twenty friendly cases. Ericsson's core claim is about quality of practice design and feedback, not about collecting reps for a tracker.
Where a live voice AI case fits next to humans
Human mocks give you interruption style, ego threat, and a second brain that can invent follow-ups. They are uneven: partners drift, rubrics mutate, friends go soft.
CoachNed's live voice case at /interview (CopperCup, about 18 minutes, seven-score debrief) is the opposite trade. Same case, same scoring dimensions, every time. Useful when you need a stable baseline between human sessions, or when you cannot book a partner tonight. It does not replace a tough human interviewer. It replaces the week of theater where nobody wrote scores.
A sane mix in a recruiting season: two human mocks per week with the printable sheet, plus one voice case when you want a clean measurement without scheduling. Between them, drill the named leak on structure, math, or charts—not another full case for the sake of the count.
Everything on CoachNed is open for seven days, no card; then $120 for a recruiting season or $49 a month. CoachNed is not affiliated with McKinsey, BCG, Bain, Deloitte, or any firm named above.
Practice this week
- Print the six-row score sheet above.
- Run one 30-minute mock with a partner. Fill the sheet. Name one leak in a single sentence.
- Drill that leak for two days (structures, timed math, or chart "so whats").
- Re-run a scored case—human or /interview—and check whether that dimension moved.
If you need a cold start without a partner, take the five-minute typed first rep at /start, then come back to a full mock when you have someone who will stay in role.
Frequently asked questions
How many mock case interviews should I do before MBB?
Enough scored mocks to move your weakest dimension at least once, then maintain. For most candidates that is closer to eight to twelve scored sessions with drills between them than to thirty friendly cases. Firms do not publish a magic number; treat volume claims on forums as folklore.
Should the mock interviewer coach during the case?
No. Mid-case tips destroy the measurement. Stay in role until the recommendation, then switch to feedback with the same rubric every time.
What if I do not have a strong practice partner?
Use a stable AI voice case for baseline scores, and save scarce human hours for interruption practice. Pair that with short drills so the partner hour is not spent teaching basics.
How long should a mock consulting case interview last?
Twenty-five to thirty-five minutes of case plus a ten-minute scored debrief is enough to stress structure, math, and the close. Longer sessions usually add chat, not signal.
Can I practice mocks alone and still improve?
Yes, if you record yourself, score against a fixed sheet, and drill the leak. Solo practice without a rubric is just talking to a wall. Pair alone work with occasional human or voice cases so interruption and pacing get tested.
Sources
- Interviewing at McKinsey — official description of the problem-solving interview and what it evaluates. Checked 2026-09-24.
- BCG Case Interview Preparation — BCG's public guidance on structured thinking, communication, and collaborative case style. Checked 2026-09-24.
- Ericsson, Krampe, and Tesch-Römer (1993), Psychological Review — deliberate practice definition, feedback, effort constraints, and expert-performance evidence. Checked 2026-09-24.
- BCG Recruiting Interview Preparation Brochure (Nov 2018) — firm-published notes on case skills (structure, analysis, synthesis) and interview length norms of that era. Checked 2026-09-24.