What Is a Good McKinsey Solve Score? The Honest Answer
Written by Tom Prescott, founder of SolvePrep. Last updated: August 31, 2026.
The honest answer has two halves, and the first one surprises people: you will never see a McKinsey Solve score. Not during the test, not after, not in the rejection email, not in the interview invitation. McKinsey scores the assessment internally and tells candidates nothing. So when someone asks "what's a good Solve score," they're really asking one of two better questions — what is McKinsey measuring? and how do I know if I'm ready? Those we can answer.
What McKinsey actually measures
The Solve isn't graded like a math test with one right total. Across the three games — Sea Wolf, the Red Rock Study, and the Sustainable Future Lab — the assessment reads both your product (did the choices work?) and your process (how you got there: what you examined, in what order, how consistently). McKinsey has said publicly that the way you work matters, not just the outcome, and the game design backs that up: several mechanics exist only to observe decision patterns.
That second layer is where candidates lose without noticing. In our data, the single worst-performed component anywhere in the Solve is consistency — Sustainable Future Lab candidates lose 60.7% of the available consistency points (we published the full breakdown, confidence intervals included in our study of candidate practice data). You can feel like you did fine on the surface while the process layer quietly graded you down.
So what counts as "ready"?
Since McKinsey won't show you a number, the only score you'll ever see is a practice score — and that's genuinely useful, because it's comparative. When you finish a run on SolvePrep you get a percentile against other practicing candidates: not "you got 71%," but "you did better than X% of people preparing for the same test." That's the readiness signal that matters, for one simple reason: the Solve is a screen, and screens are relative. You're not trying to beat the test; you're trying to beat enough of the field.
Two patterns from our data worth calibrating against. First: your first score is not your score. Users score 20% higher on their best run than their first — and even after we strip out the statistical noise (regression to the mean; we ran a placebo test and publish both numbers), roughly half of that gain is real improvement. Second: candidates who pass report an average of 8 full practice runs. One run tells you where you started. It doesn't tell you where you'd finish.
What actually moves the number
Not tips. Reps — but aimed reps. The pattern we see in score progressions: the first run is dominated by format shock (time pressure, unfamiliar mechanics), runs two through four fix the mechanical losses, and the runs after that are where the process layer improves — the consistency and prioritization signals that the assessment weighs and that you can't fake in a single sitting. That's the reasoning behind the full breakdown on Elite: it shows which specific decisions cost points, so run five isn't a rerun of run four's mistakes.
Common questions
Can I see my real McKinsey Solve score?
No. McKinsey doesn't disclose scores, percentiles, or per-game feedback to candidates at any stage. The only outcome you'll ever see is the interview decision how results actually reach you.
Is there an official passing score?
None is published, and the bar likely moves with the applicant pool and office. McKinsey doesn't publish pass rates either — most estimates put the failure rate around 70%.
What's a good practice percentile before test day?
There's no magic threshold, but directionally: if your best runs sit comfortably in the top bands against other practicing candidates — people motivated enough to practice — you're ahead of a field where 70% of candidates, by our data, don't even start until the final 48 hours.