The McKinsey Solve Assessment: Complete 2026 Guide

Everything a 2026 candidate needs: which version of the test you’ll sit, what each of the three modules actually asks of you, how Solve is scored, and how long to spend preparing for each part.

Last updated: August 2026 · Written from candidate reports and real past PSG data

1. What the McKinsey Solve actually is

Solve is the online assessment McKinsey uses to screen candidates before interviews. You’ll see it called several things — the Problem Solving Game, the PSG, the Digital Assessment, or Imbellus, after the company that originally built it. They all refer to the same test. McKinsey’s own name for it is simply Solve.

It is not a case interview in software form, and it is not a knowledge test. There is nothing to revise. Instead, Solve places you inside a set of ecological scenarios and watches how you handle them: how you read information, how you handle constraints, how you decide when the data is incomplete, and how consistent your reasoning stays under time pressure.

You sit it at home, unsupervised, usually within a week or two of being invited. It runs in the browser, it is timed, and you cannot pause it once it starts.

2. Which version you’ll sit: 65 vs 85 minutes

There are currently two versions in circulation, and your invitation email tells you which one you have by stating the duration.

By mid-2026 most offices had moved to the 85-minute version, so assume you’ll face all three modules unless your invitation explicitly says 65 minutes. The rollout has been uneven across offices, roles and recruiting cycles, and McKinsey has varied the format by region before — so check your own invitation rather than relying on what a friend sat last cycle.

One point that trips people up: Ecosystem Building is no longer part of the standard assessment. It was the original scenario, and it was replaced by Sea Wolf in 2024. A lot of preparation material still circulating online was written for Ecosystem. If what you’re practising involves building food chains from a species library, you are preparing for a test McKinsey has retired.

3. The three modules at a glance

The three are genuinely different skills, which is the point — Solve is designed so that being strong at one won’t carry you through the others.

4. Redrock: the data study task

Redrock comes first, and for most candidates it is the hardest part of the assessment. You are given a research scenario about an animal population and a body of supporting material — data tables, charts, written findings — and asked to work through it.

It splits into two halves. The first presents case-based questions alongside report tasks where you assemble a written and visual answer from the material provided. The second is a set of standalone questions that test the same underlying numeracy without the narrative wrapper.

An on-screen calculator is provided, which tells you something important: the difficulty isn’t the arithmetic itself, it’s doing it accurately while reading carefully and watching the clock. The most common failure mode is not getting sums wrong — it’s misreading which figure the question actually asks for.

Because Redrock runs first and is the most demanding, a bad start here tends to affect everything that follows. If your preparation time is limited, this is where it should go. Redrock practice tests →

5. Sea Wolf: the constraint puzzle

Sea Wolf, introduced in 2024, is an ocean-cleanup simulation. You act as a researcher designing a microbe-based treatment for a set of polluted ocean sites. Each site comes with its own target characteristics, traits that are desirable, and traits that must be avoided, and you assemble a combination of microbes that satisfies those requirements.

Structurally it is a constraint satisfaction problem. The scenario is dressed as marine biology, but what you are doing is filtering a large option set against multiple simultaneous conditions and finding a combination that works. There is a correct answer, and there is a systematic way to find it.

That makes Sea Wolf the most learnable module in the assessment. Candidates who go in cold tend to search haphazardly, run out of time, and submit something that satisfies most but not all of the constraints. Candidates who have practised have a method: filter first on the hard requirements, eliminate anything carrying a forbidden trait, then optimise within what survives. Sea Wolf practice tests →

6. Sustainable Futures Lab: the judgement scenario

SFL is the newest addition, introduced in March 2026, and it is unlike the other two. There is no maths, no calculator and no single correct answer. You join an environmental research team and work through thirteen sequential decisions about priorities, interpretation and competing demands.

The critical thing — and the thing most candidates get wrong — is that your answers are assessed as a pattern rather than marked individually. SFL is looking at whether your thirteen decisions together describe someone with a coherent way of thinking. A consistent set of choices generally reads better than an inconsistent set containing several individually ideal answers.

It also comes last, after 65 minutes of dense quantitative work, when your judgement is at its weakest. Read the full Sustainable Futures Lab guide →

7. How Solve is scored

McKinsey publishes very little about scoring, so be sceptical of anyone quoting precise thresholds. What can be said with reasonable confidence:

You are assessed on more than your final answers. Solve captures how you work — the order you do things in, what you revisit, how you use the time available — not just what you submit. This is the whole reason the assessment exists in a game format rather than as a multiple-choice test.

Scores are relative. Candidates are ranked against a comparison group rather than measured against a fixed pass mark, which is why “what score do I need?” has no fixed answer. It also means the bar shifts by office and by cycle.

You almost certainly won’t be told your score. Most candidates receive only a progression decision. Be wary of any product whose guarantee depends on you producing a score McKinsey doesn’t give you.

8. How long to prepare, and on what

Solve rewards familiarity far more than it rewards study hours. The goal is not to learn content — there isn’t any — but to reach test day with no surprises, so your attention goes on the problem rather than the interface.

For most candidates that means somewhere between five and fifteen hours. Where those hours go matters more than how many there are:

Two practical notes. Practise full timed runs, not isolated questions — the pressure is a large part of what’s being tested, and untimed practice rehearses the wrong thing. And practise in sequence, Redrock first, so you experience SFL in the fatigued state you’ll actually meet it in.

If you have days rather than weeks, do one timed Redrock, one timed Sea Wolf, and one SFL run-through, and spend whatever remains on Redrock.

9. The mistakes that cost most candidates

10. Frequently asked questions

PSG Solver is an independent preparation resource and is not affiliated with, endorsed by or connected to McKinsey & Company. Assessment formats vary by office, role and recruiting cycle — always check the details in your own invitation.