McKinsey Solve: What Your Invitation Length Tells You About the Test

McKinsey publishes almost nothing about Solve. No module names, no durations, no scoring model, no score report. Its FAQ contains one sentence that is more useful than everything the preparation market has written about the assessment put together: “the length of your Solve assessment is outlined in your invitation email”. That number tells you which modules you are about to sit.

What the two common lengths mean

As of September 2026, dated candidate and coach reports converge on two configurations, and they have been stable since the spring.

Each module runs its own clock, and the clock pauses between them, so minutes saved on Redrock cannot be spent on Sea Wolf. Inside Redrock the opposite is true: its four phases share one 35 minute budget which you allocate yourself, and that is where most of the time pressure in Solve actually lives.

Redrock: four different tasks wearing one name

You are a researcher studying an ecosystem on a fictional island. The screen has three panels: the material in the centre, a Research Journal on the right, the phase list on the left. What most guides miss is that the four phases test four unrelated things.

Investigation has no answer to submit at all. A research objective is stated, a page of text and tables appears, and specific figures on it can be collected into your journal. Most of what is on screen is deliberate noise. The task is relevance: gather exactly what the objective needs and leave the rest. Collecting everything feels safe and is the most common way candidates run the module out of clock, because the next phase then has to search their own journal.

Analysis introduces the calculator, and the calculator is the point. It does four operations and percentages, no exponents, and it logs every result it produces at full precision so you can reuse it in the next calculation instead of retyping a rounded version. Two to four questions, and they chain: one wrong intermediate poisons every answer after it, including the report.

Report asks for three things. Fill the blanks in a pre-written summary, choose the chart type that fits the data, and populate that chart. It is graded on accuracy, not prose, and the chart choice is marked. Choosing a line chart where the data compares categories is named by more than one source as a common and entirely avoidable loss.

Cases is six short independent mini-cases with no shared narrative, under whatever is left of the 35 minutes. There is no per-case timer.

Sea Wolf: it is the average that matters

Three contaminated ocean sites, each treated with exactly three microbes chosen from a pool you assemble yourself. Microbes have three numeric attributes on a 1 to 10 scale plus qualitative traits, and their names are deliberate nonsense so that no real knowledge helps.

The scored rule is public, in the sense that three independent sources state it identically, and it is the single most useful thing to know before you start:

• The meanof each attribute across your three chosen microbes must fall inside the site's range for that attribute. Not each microbe. The mean.
• At least one of the three must carry the site's desired trait.
• None of the three may carry its undesired trait.

Efficiency starts at 100% and loses 20 points for each attribute whose mean falls outside range, 20 if no microbe carries the desired trait, and 20 for each microbe carrying the undesired one. Candidates report scores like 60%, 80% and 100% across their three sites in one sitting, and coaches describe that spread as unremarkable.

Two consequences follow, and both are counterintuitive. First, a microbe whose value is wildly outside the range is not disqualified: with a target of 2 to 4, a microbe at 8 alongside two at 1 and 2 gives a mean of 3.67 and is fine. Filtering microbe by microbe throws away the combinations that work. Second, 100% is not always available. Sometimes the pool simply cannot produce a perfect treatment, and recognising that 80% is the ceiling and moving to the next site is the right call rather than a failure. Time management across the three sites is described as a significant issue, and the clock does not stop while you hunt for a treatment that does not exist.

Sustainable Futures Lab: reading, not arithmetic

The newest module and the least documented. Thirteen questions in about twenty minutes on a text-based environmental project, with the theme rotating per candidate between wetland restoration, air quality and coastal habitat. The first item is a drag-and-drop prioritisation of about four possible actions. The remaining twelve are multiple choice with long, paragraph-length options that begin with phrases like “You suggest...” or “You recommend...”. One continuous narrative, so later questions build on what has already happened. No calculator, and no arithmetic anywhere in it.

What is actually known about the scoring

Less than you will be told. The one peer-reviewed source, presented at EDM 2018 by the Imbellus team with a McKinsey co-author, describes the 2017 and 2018 assessment as deriving its scores from telemetry, capturing “both their cognitive process and product”, and rolling roughly 100 constructs up into five: critical thinking, decision-making, metacognition, situational awareness and systems thinking. It reports correlations of 0.43 to 0.71 with passing McKinsey's old paper test. That is solid evidence about the original design, and it describes an assessment none of whose modules are still in rotation.

For the current modules, the sources contradict each other. One major preparation site says process scores “appear to have minimal influence”. Another says process and product are weighted roughly equally. Neither offers evidence, and the first concedes elsewhere on its own site that McKinsey has never stated how process is measured and that it is working from unverified hypotheses. Pass-rate claims are worse: the range in circulation runs from 10% to 33%, and every figure sits next to an offer to improve it.

The practical consequence is that nobody can tell you how to optimise for the process score, and anybody who says otherwise is selling something. What is left is the advice that is honest because it also produces a better answer: read the objective before you collect anything, work in the tools the game gives you, update a decision when the game shows you something that makes it wrong, and do not let one task eat a clock that other tasks are sharing.

The logistics that catch people out

If you want to sit the tasks rather than read about them, our McKinsey Solve practice simulations cover all six: the four Redrock phases as separate drills, Sea Wolf with a generated deal and a known optimum, and Sustainable Futures Lab. And if somebody has recommended you build a spreadsheet solver for Sea Wolf, read what the rules actually say first.