McKinsey & Company · Online assessment

McKinsey Solve Practice Test 2026

Updated 2026-08-06 · All practice scenarios on this page are original CasingLab creations.

McKinsey Solve is the gamified online assessment McKinsey uses to screen candidates before first-round interviews. Instead of a multiple choice aptitude test, you play ecology themed mini games for roughly 65 minutes while the platform records both your answers and how you reach them. McKinsey retired its old written test, the PST, and Solve is now the standard first hurdle for most offices worldwide.

This page gives you the current 2026 format: which mini games you are likely to see, how the product score and process score work, what advancing actually requires when no official cutoff is published, and a free three question mini simulation you can play right now, no account needed.

One important note on sourcing: every practice scenario on this page is an original CasingLab creation. It mirrors the format of Solve, the question types, the timing pressure and the answer structure, never the real test content.

Format and timing

A standard invitation runs about 65 minutes and pairs Redrock with Sea Wolf. Longer invitations, around 85 minutes, also include the Sustainable Future Lab. The older Ecosystem Building simulation has been largely phased out and most candidates will not see it. McKinsey publishes none of these figures, so treat them as commonly reported rather than official.

SectionDurationQuestionsFormat
Redrock StudyAbout 35 minRoughly 10 to 15 tasksA guided research study: read a brief, drag data into a research journal, run calculations, then answer standalone quantitative mini cases.
Sea WolfAbout 30 minRotating micro tasksCategorize organisms, compare candidates against criteria, and make sequential selection decisions as the constraints shift between rounds.
Sustainable Future Lab (longer invitations)About 20 minScenario decisionsA judgment module with no calculations: rank priorities inside a research team, then make evolving decisions that are profiled across work style traits.

Practice the real McKinsey Solve format

Full simulations, real timing, detailed scoring.

Practice the full simulationTry the free diagnostic

Question types on the McKinsey Solve

Sequential decisions under shifting criteria (Sea Wolf style)

Sea Wolf runs as a multi step scenario: you review a situation, sort what you find into groups, then choose between candidates against criteria that move between rounds. Each decision narrows the options for the next one, so what is measured is adapting a strategy rather than executing a fixed plan. There is no arithmetic shortcut here. The work is reading every criterion precisely and staying consistent once the thresholds change.

Scored skills: Sequential decision making · Criteria based classification · Adapting to changing constraints · Consistency across rounds

Example question

Invented example in the Sea Wolf style. Early in the module you sort organisms into groups before you are allowed to use them.

An organism joins the fast responder group only if it reproduces in under 5 days AND survives below 10 degrees. Organism W reproduces in 3 days and survives to 6 degrees. Organism X reproduces in 4 days and survives to 14 degrees. Organism Y reproduces in 7 days and survives to 4 degrees. Organism Z reproduces in 2 days and survives to 8 degrees. Which organisms qualify?

  • W and X
  • W and Z
  • X and Z
  • Y and Z

Answer: W and Z

Both conditions have to hold at once. X reproduces fast enough but only survives to 14 degrees, above the 10 degree line. Y survives the cold but takes 7 days. W and Z clear both tests. The module chains these classifications, so a single misread rule quietly propagates into every decision that follows it.

Data analysis tasks (Redrock style)

You are handed a research brief about an animal population with tables, charts and scattered facts, and asked to extract the right numbers, compute growth rates, ratios or expected values, and drag conclusions into a structured report. Accuracy is binary: a value that is off by one rounding step is simply wrong. The challenge is less the math itself than staying precise while a visible timer runs.

Scored skills: Data extraction from mixed sources · Percentage and growth calculations · Precision under time pressure

Example question

Invented example in the Redrock style. A coastal reserve tracks its sea otter population to plan feeding capacity.

The survey counts 320 otters at the end of Year 1. The model assumes 15 percent growth in Year 2 and 10 percent growth in Year 3. How many otters does the model predict at the end of Year 3, rounded to the nearest whole number?

Answer: 405

Year 2: 320 times 1.15 equals 368. Year 3: 368 times 1.10 equals 404.8, which rounds to 405. Chained percentages are the single most common Redrock computation, and the classic error is adding the percentages (25 percent on 320 gives 400, which is wrong).

Standalone quantitative mini cases

After the guided study, Solve serves short self contained questions: a small dataset, one precise question, usually multiple choice or a single numeric entry. Topics rotate through rates, proportions, simple probability and reading comprehension of charts. Each one is quick, but there are many, so pacing decides whether you finish.

Scored skills: Ratio comparison · Basic probability · Chart reading · Pacing across many small tasks

Example question

Invented example. A marine station compares bleaching across two reef zones it monitors.

Zone A has 84 coral colonies, of which 21 are bleached. Zone B has 140 colonies, of which 42 are bleached. Which statement do the data support?

  • Zone A has the higher bleaching rate
  • Zone B has the higher bleaching rate
  • Both zones have the same bleaching rate
  • The rates cannot be compared from these figures

Answer: Zone B has the higher bleaching rate

Convert to rates before comparing: 21 of 84 is 25 percent, 42 of 140 is 30 percent. Absolute counts mislead here, which is exactly the trap this question type sets. Divide first, compare second.

Try three practice questions, free

Three original questions in the McKinsey Solve format. No account, no email, instant corrections with explanations.

Question 1 of 3

Sequential selection, original scenario in the Sea Wolf style. You must pick one microbial strain to treat a contaminated inlet. It has to tolerate the inlet salinity of 18 ppt, act within the 9 days remaining, and stay inside the 40 unit oxygen budget. Strain A tolerates 15 ppt, acts in 6 days, uses 30 oxygen. Strain B tolerates 22 ppt, acts in 11 days, uses 25 oxygen. Strain C tolerates 20 ppt, acts in 7 days, uses 38 oxygen. Strain D tolerates 25 ppt, acts in 8 days, uses 44 oxygen.

Practice the real McKinsey Solve format

Full simulations, real timing, detailed scoring.

Practice the full simulationTry the free diagnostic

What is actually scored

McKinsey has publicly described Solve as producing two signals. Neither your raw score nor the exact rubric is disclosed to candidates, so the precise weighting below is based on McKinsey's own general descriptions plus consistent candidate reporting, not on a published scoring key.

Product score

What you achieved: were your calculations correct, did your selections satisfy the criteria, how many tasks did you complete accurately. Candidate reporting in recent cycles suggests this outcome score carries the dominant weight.

Process score

How you worked: the platform records clicks, sequencing, hesitation and corrections while you play. McKinsey has described measuring how candidates approach problems, not just whether they solve them, which is why guessing wildly is risky even when it happens to land on a right answer.

Consistency across games

Both assigned games feed the same decision. A strong first game does not excuse a collapsed second one, so pacing your energy across the full sitting is part of the test.

What score do you need on McKinsey Solve?

The direct answer: McKinsey does not publish an official cutoff, a pass rate or a score range for Solve, and it never shows candidates a numeric result. Any precise threshold you see quoted online is an estimate, not an official figure.

What the publicly available reporting does suggest: a strong product score, meaning a high accuracy rate across the study tasks, appears to matter more than any single process metric, and candidates who leave tasks unfinished tend not to advance. You will also find specific cutoffs and percentile bands quoted around the internet. Treat them as unverified: they come from prep vendors aggregating self reported outcomes, not from McKinsey, and we will not repeat a number we cannot source.

Two things are worth planning around. First, Solve is one input among several: the same performance can pass at one office and fail at another depending on the strength of the local applicant pool and the rest of your file. Second, because results are relative, the practical goal is not a magic number but maximizing completed, correct tasks. In practice that means keeping calculation accuracy near perfect, satisfying every criterion in the selection rounds, and not stalling on any single task.

The PST is retired: what replaced it

McKinsey phased out the Problem Solving Test, the paper based multiple choice exam that screened candidates for over a decade, and rolled Solve out globally in its place. If you find PST practice materials, treat them as math warm up only: the format, the interface and a large part of what is measured have changed.

The practical consequence: you can no longer grind one fixed question bank. Solve rewards transferable habits, fast two and three digit arithmetic, chained percentages, reading data tables without misreading a row, and calm sequencing under a visible timer. Those habits are exactly what a good simulation trains.

How to prepare for Solve in 2026

Start with format familiarity: knowing how each module is structured before test day removes the single biggest source of wasted minutes. Then drill the arithmetic that Redrock style tasks repeat endlessly: growth chains, ratios, weighted averages, expected values. Finally, rehearse under real timing at least twice, because most candidates report the clock, not the content, as what beats them.

A sensible two week plan: three untimed sessions to learn the mechanics, then alternating timed simulations and targeted drills on whichever computation type you miss most, finishing with two full dress rehearsals in exam conditions.

Frequently asked questions

A standard invitation is commonly reported at about 65 minutes across two modules, plus setup and tutorials. Longer invitations run around 85 minutes because they add a third module. Either way, plan a full uninterrupted window with a stable connection, since the assessment runs in one sitting.

Keep reading

Practice the real McKinsey Solve format

Full simulations, real timing, detailed scoring.

Practice the full McKinsey Solve simulation