Updated September 2026 · 12 min read · Global management consulting firm
Search for McKinsey's assessment and you will be told, confidently and repeatedly, that McKinsey acquired Imbellus, the assessment startup that built the original Problem Solving Game. It did not.
On 1 December 2020, Roblox acquired Imbellus's intellectual property assets, and Imbellus founder Rebecca Kantar moved to Roblox to lead a new assessment team there. Roblox said at the time it was "creating a new team within Roblox, led by Imbellus' visionary founder Rebecca Kantar to build a system on Roblox for fair and objective assessments." McKinsey was Imbellus's flagship client, not its buyer.
This matters for more than pedantry. It is a useful early signal about the quality of information circulating on this assessment. A great deal of what you will read about Solve — scenario names, timings, scoring models, pass rates — comes from prep vendors and candidate forums rather than from McKinsey, and some of it is years out of date. McKinsey itself publishes remarkably little. Knowing which is which is the first practical advantage you can give yourself.
This guide separates the two throughout. Where McKinsey says something, it is quoted. Where the only source is a coach, a forum or a prep site, that is stated plainly.
McKinsey names four components on its own interviewing page, and the naming is worth learning precisely because the firm's internal vocabulary shows up in recruiter emails.
Solve. McKinsey describes it as "one of the exciting steps in our process for most consulting roles" and as "a gamified assessment that tests your natural problem-solving abilities." It typically sits immediately after CV screening.
Personal Experience Interview (PEI). McKinsey instructs candidates to "come prepared with two personal examples that demonstrate how you've made an impact in each of the four areas." The PEI is a structured behavioural interview, and it is scored.
Problem-Solving Interview. This is McKinsey's own label for what everyone else calls the case interview — the firm's page says "also called case interview." McKinsey describes it as presenting "a business case to evaluate your analytical thinking and approach to solving complex problems."
Expertise Interview. For technical and specialist roles, this may include "coding challenges, problem-solving tasks, or other role-specific exercises."
A typical generalist funnel therefore runs: application and CV screen, Solve, first-round interviews combining PEI and case, second and final round with Partners, then offer. Third-party accounts put the end-to-end process at roughly 1 to 3 months, with campus and MBA cycles opening between August and October and experienced hires recruited year-round. McKinsey does not publish that timeline itself.
One official detail is worth flagging before you assume you will take Solve at all. McKinsey groups its non-consulting jobs under the heading internal roles, spanning ten functions, and that hub describes no assessment and never mentions Solve. The firm's own wording is "most consulting roles" — not all roles, and not all applicants.
McKinsey's public description of Solve is short and slightly disarming: "a gamified assessment created to showcase your problem-solving abilities," accompanied by the advice that "there's no need to prepare—approach it with curiosity and creativity."
The firm's FAQ document adds three things that are more revealing than the marketing copy.
First, every candidate gets a different test. McKinsey states that Solve "uses advanced AI methods to ensure that there is ample variation in each scenario, and it is designed so that no two individuals have the same parameters and combinations of data." There is no answer key circulating, because there is no single set of answers to circulate.
Second, McKinsey reserves the right to check your reasoning in person. The FAQ says you "may also be asked at random to take an additional task in person and/or to explain your logic used." This is the closest official confirmation that how you reached your answers matters, not only what you submitted — and it is a meaningful deterrent against anyone who did not do the work themselves.
Third, Solve is not a standalone gate in McKinsey's own account of it. The firm says twice, in two different documents, that results are "taken into consideration together with the rest of your application and any other assessment results."
That last point is genuinely contested. On Wall Street Oasis, a self-identified McKinsey employee described the bar as "top decile usually," while in the same thread a consulting intern reported that "my recruiter directly told me they don't use the score at all," and a third participant wrote simply, "Guys the game does not matter." Consultants who coach candidates disagree just as sharply, with some describing Solve as a hard percentile filter applied before a recruiter opens your CV and others calling it a secondary signal. McKinsey's own wording supports the holistic reading. Nobody outside the firm can settle it, and you should be suspicious of anyone who claims to have.
The widely circulated model of a "product score" plus a "process score," assessed against five constructs — critical thinking, decision-making, metacognition, situational awareness and systems thinking — traces back to Imbellus-era material rather than to anything currently published by McKinsey. Treat it as a plausible framework, not a specification.
The scenario set has changed materially, and this is where out-of-date guides do the most damage.
Candidate reports and consulting coaches converge on the same account: Ecosystem Building was phased out around the middle of 2025, and the older Plant Defense, Disaster Management, Disease Management and Migration Management scenarios are legacy content no longer in general use. On PrepLounge in autumn 2025, a candidate reported hearing that "since July 2025, they've eliminated Ecosystem and the only two remaining are RedRock and Sea Wolf," and a coach responded confirming that most candidates now see only those two.
What appears to be in use now:
Two formats are described by third-party sources: a shorter version pairing Redrock and Sea Wolf, and a longer version adding Sustainable Futures Lab. Candidates do not choose; the format is assigned by office and role.
Every one of those scenario names is candidate- and coach-derived. McKinsey names no scenario anywhere on its own site, and it does not publish durations either — the FAQ says only that "the length of your Solve assessment is outlined in your invitation email." Any guide quoting exact minute counts as fact is quoting a forum.
Two things McKinsey does confirm about the mechanics: each task "will begin with an untimed tutorial," and Solve "does not require previous business knowledge or gaming experience." It runs on PC or Mac, and sound is not required. If you have a disability, health condition or specific learning difficulty, McKinsey says to speak to your recruiter, and extra time is available.
The PEI and the Problem-Solving Interview are conducted by consultants, not by a platform. There is no third-party test vendor involved in the interview stages, and McKinsey names none for Solve either.
In January 2026 the Financial Times reported that McKinsey had begun testing candidates with its internal AI assistant, Lilli — candidates analysing a case with the tool's help and being assessed on how well they formulated their questions. Follow-up coverage described the exercise as targeted at recent graduates, positioned around curiosity and judgement rather than pass-or-fail, and piloted in selected US offices.
McKinsey declined to comment on the report. Its own interviewing page mentions Lilli only as a tool used by staff, not as an interview stage. Treat this as a real but unconfirmed development, and do not plan your preparation around it.
Generalist consulting (Business Analyst, Associate). The full funnel: Solve, then PEI plus case interviews across two rounds. This is the population that most Solve content is written about.
Technical and specialist roles. The Expertise Interview replaces or supplements the standard case, and McKinsey explicitly mentions coding challenges and role-specific exercises. Solve may still apply.
Experienced hires. Recruited year-round rather than on a campus cycle. Whether experienced hires universally take Solve is not stated officially; some third-party sources say nearly all applicants do.
Internal and business-services roles. McKinsey's own internal-roles hub describes no assessment at all. If you are applying to a function rather than to consulting, do not assume Solve is coming.
Students in the UK applying through McKinsey's Solve It case competition are told directly that they "will be required to complete our digital assessment, Solve" as part of entry.
McKinsey publishes no sample questions and no scenario detail, and reproducing real Solve content would be both a copyright problem and useless to you, since no two candidates receive the same parameters. What follows are original practice items built to exercise the same underlying reasoning. They are not McKinsey questions and do not imitate McKinsey's interface.
Data-interpretation practice, in the spirit of a case-analysis module.
A regional distributor operates 4 depots. Weekly dispatch volumes are 1,240, 890, 2,110 and 1,760 units. Handling cost per unit is £0.85 at the two largest depots and £1.10 at the two smallest. A proposal would close the smallest depot and redistribute its volume equally between the two largest.
Which of the following would you need to verify before concluding the proposal reduces total handling cost? (a) Whether per-unit handling cost at the two largest depots stays flat as volume rises (b) The lease expiry date at the smallest depot (c) The average delivery distance from each depot (d) All of the above
The reasoning being practised here is not arithmetic. It is noticing that the cost saving depends entirely on an assumption the question quietly makes — that unit costs do not change with volume — and that a capacity constraint would invalidate the whole calculation. That habit of interrogating the assumption behind a number is what case-style assessment rewards.
Systems-reasoning practice, in the spirit of an ecology module.
You are asked to build a stable food chain from a set of organisms, each with a stated calorie requirement and a stated set of prey. One candidate chain satisfies every calorie requirement but places two species in the same niche competing for a single food source.
What is the weakness in the chain, and what would you check first?
The habit being practised is looking past whether individual constraints are satisfied to whether the system as a whole is stable — the difference between a valid answer and a robust one.
Structuring practice, in the spirit of a case interview.
A national coffee chain has seen same-store revenue fall 6% over 12 months while transaction count rose 3%.
Lay out the two or three branches you would investigate, in order, and say what data you would ask for first.
There is no single right structure. What is being assessed is whether your branches are mutually exclusive, whether they collectively cover the problem, and whether you can say out loud why you chose to look at one before another.
The gap between McKinsey's "there's no need to prepare" and the industry that has grown up around Solve is the most consistent theme in candidate discussion.
On Wall Street Oasis, a candidate discussing preparation options in 2024 concluded it was "hard to practice for I'm afraid," while another who had paid for a third-party simulator reported finding the real assessment easier than the paid imitation. A 2026 reply in the same thread was blunter: "you do not need to pay for any Solve preparation materials at all."
On PrepLounge in January 2026, a candidate gave an unusually precise self-assessment — completing the first two case sections accurately, running out of time on the next two, never starting the last two, and scoring around 80% on each of three sections in the ecology scenario. They attributed the shortfall partly to internet dropouts and household interruptions, which they flagged in the assessment's own comment box. The practical lesson is contained in that detail: sit the assessment somewhere you will not be interrupted, on a connection you trust, and use the comment facility if something goes wrong.
Also on PrepLounge, a candidate described building a spreadsheet to work through calculations quickly, and a coach warned against relying on external tools during a proctored sitting, advising mental arithmetic and scratch paper instead. McKinsey's own integrity rules are explicit here: candidates may not "use applications, websites, AI tools, or pre-prepared notes," may not "record or capture screenshots of any part of the assessment," and must work alone.
After Solve, results feed into the same pool as your CV and any other assessments. McKinsey does not publish a score, a band or a pass mark, and you should not expect to be told one.
After the first interview round, strong candidates progress to a final round conducted largely by Partners. Third-party accounts describe first rounds of one to three interviews and final rounds of two to three, with the PEI and case weighted comparably. McKinsey does not confirm weightings.
After the final round, outcomes are typically communicated by your recruiter. McKinsey publishes no decision SLA.
At every stage, McKinsey's stated sanctions for integrity breaches are severe and specific: "exclusion from consideration, rescinding of an offer, restrictions on future applications, or legal action."
McKinsey says you do not need to prepare for Solve. That is true in the narrow sense that no amount of studying will let you memorise a scenario you have never seen and that nobody else will see either. It is not true in the broader sense, and the firm's own advice to approach it "with curiosity and creativity" is easier to follow when you are not also fighting unfamiliar mechanics.
Get comfortable with the format, not the content. The tutorial before each task is untimed. Use it properly. Read every instruction rather than skimming towards the first interaction.
Build calculation fluency without a calculator. Case-style modules reward quick, confident mental arithmetic with percentages, ratios and unit conversions. This is the single most trainable component.
Practise reading data critically. Given a table, ask what it does not tell you. Given a total, ask what it is composed of. Given a trend, ask what would have to be true for it to continue.
Prepare the PEI properly, because it is entirely preparable. McKinsey asks for two examples in each of four areas. That is eight stories, each with a concrete situation, your specific actions, and a measurable result. Write them out. Time yourself telling them in around 3 minutes. The PEI is scored, and unlike Solve it rewards preparation directly.
Do enough cases to stop panicking about structure. The Problem-Solving Interview is a conversation, not a monologue. Practise saying what you are doing and why before you do it.
Protect the sitting itself. Stable connection, quiet room, charged laptop, no interruptions, nothing on your screen that you would not want a proctor to see.
Do not use AI during the assessment. McKinsey's rules prohibit it explicitly, the firm reserves the right to ask you to explain your logic in person, and the stated consequences run to rescinded offers and legal action.
No. Roblox acquired Imbellus's intellectual property assets on 1 December 2020, and Imbellus founder Rebecca Kantar joined Roblox to lead assessment work there. McKinsey was Imbellus's client.
Candidate and coach reports point to Redrock Study and Sea Wolf as standard, with Sustainable Futures Lab appearing on longer sittings since 2025. Ecosystem Building was phased out around mid-2025. McKinsey names no scenario publicly, so treat all scenario naming as unofficial.
McKinsey says only that "the length of your Solve assessment is outlined in your invitation email." Two format lengths are described by third-party sources, but McKinsey publishes no duration and neither should anyone else.
McKinsey publishes no pass mark, benchmark or percentile. Claims that a specific proportion of candidates fail are unverified and mutually inconsistent across sources.
McKinsey does not publish a scoring model. The strongest official signal that reasoning matters is the FAQ line stating you may be asked at random "to take an additional task in person and/or to explain your logic used."
No. McKinsey describes it as part of the process "for most consulting roles." Its internal-roles hub, covering ten non-consulting functions, describes no assessment and does not mention Solve.
No. McKinsey's integrity rules prohibit applications, websites, AI tools and pre-prepared notes, and prohibit screenshots or recordings. Sanctions include rescinded offers and legal action.
The Personal Experience Interview is a structured behavioural interview requiring two examples in each of four areas. The Problem-Solving Interview — McKinsey's own term for the case — presents a business problem to assess analytical thinking. Both are scored.
The Financial Times reported a pilot in January 2026 in which candidates worked through a case with McKinsey's internal assistant Lilli and were assessed on the quality of their questioning. McKinsey declined to comment on the report and does not list it as a stage.
Use TestSolve to crack the McKinsey & Company assessment
AI-powered answers and reasoning sent to your phone. Free first solve, no signup.
Try it free →See pricing