McKinsey Sea Wolf Practice Simulation

The complete McKinsey Sea Wolf guide

McKinsey's public description of Sea Wolf is short: the Solve game where you select and evaluate microbes to clean up contaminated sites. It does not publish the steps or the scoring, so everything below about structure comes from what candidates and prep providers report. On the core flow they largely agree.

The useful mental model is simple: Sea Wolf is a constraint-matching game, not a biology test. As reported, you treat 3 ocean sites in about 30 minutes, run the same four steps at each, and finish each site with 3 microbes judged on their averages, not one by one.

The labels vary between write-ups (Energy, Adhesion and Speed in one; Rigidity, Mobility and Size in another), which suggests the names are drawn per test. The structure does not change. Write-ups of the 2023 to 2024 beta, sometimes called Ocean Cleanup, describe a shorter one- or two-site version, and Ecosystem Building is a separate Solve game whose advice does not transfer. If your tutorial uses different wording than a guide you read online, trust the tutorial in front of you.

Who gets Sea Wolf, and when?

Sea Wolf shows up inside McKinsey Solve, which most applicants complete early in the process, typically after the application and before first-round interviews. The most commonly reported 2026 format pairs Sea Wolf with Redrock Study in a roughly 65-minute sitting, with Sea Wolf usually second. Longer invitations add the Sustainable Futures Lab: candidates report a 20-minute version in 85-minute invitations and, since late August 2026, a newer 30-minute version in 95-minute invitations. McKinsey does not publish the line-up, so your invitation email states the length and the in-test instructions confirm the games. The McKinsey Solve guide covers the full assessment flow, the variants, and how Solve feeds into the interview rounds.

What are the steps in the Sea Wolf game flow?

Candidates and prep providers describe the same flow: an untimed tutorial, then 3 ocean sites that share one 30-minute clock. Time carries over, so minutes you save on site 1 are yours for site 3. Each site runs the same four steps, and before sites 2 and 3 there is a short review of the microbes you saved for that site.

1. Read the tutorial once, carefully. It is untimed, and it is the only place the exact rules for your version are guaranteed to be correct, so do not skim it to save thirty seconds you will lose later.
2. Step 1: build the microbe profile. With the site's requirements on screen (a range for each of 3 attributes, 1 desired and 1 undesired trait), you pick exactly 2 of 7 characteristics: the 3 attributes, each with a min to max range, and 4 traits. Prep providers agree it is not part of the treatment score. The tutorial implies your profile shapes the microbes you see later; test-takers report no visible effect, and nobody outside McKinsey knows for sure.
3. Step 2: categorize 10 microbes. One at a time, each with a made-up name, 3 attribute values from 1 to 10 and 1 trait, you send each to this site, the next site, or reject it. Choices are final. Most sources report that you see the full current site but only one characteristic of the next site.
4. Step 3: build the prospect pool. Six microbes are already in the pool. Four times, you pick 1 of 3 new microbes, so the pool ends at 10. These picks are final too.
5. Step 4: create the treatment. Choose 3 of the 10 and submit; you can swap picks until you do. This step sets the site's efficiency score.
6. Review saved microbes, then start the next site. Before sites 2 and 3, you re-check the microbes you saved for that site against its full requirements and keep or reject each. Fewer sources describe this screen, and they number it differently.
7. Watch the shared clock. Prep providers suggest about 10 minutes per site. A slow site 1 is paid for on site 3.

What microbe and attribute mechanics do you track?

You track three things at once: each microbe's attribute values, its one trait, and how a group of 3 averages out against the site. The names change between versions; the building blocks candidates and prep providers report do not.

The mechanic candidates most often underweight is the set-level view. A microbe with one value outside a range is not a failure: a partner on the other side of the range can pull the average back in. The only single-microbe problem the treatment cannot absorb is the undesired trait. So ask what the full set needs to add up to, then assemble microbes that hit that target together.

How does Sea Wolf scoring work per site?

McKinsey does not publish a Sea Wolf score formula. Candidates and prep providers do report a consistent efficiency rule for the treatment, and a second, behavioral score that nobody outside McKinsey can see.

• Efficiency starts at 100% and loses 20% per miss. Prep providers report five checks on the treatment: each of the 3 attribute averages in range, at least one desired trait, and no undesired trait. Sources disagree on whether the undesired trait costs 20% once or 20% for every microbe that carries it.
• Only the treatment earns efficiency. The profile, categorize and pool steps are not part of it, though the pool you build in step 3 decides which trios are possible.
• Some sites cannot reach 100%. Candidates and prep providers report that some sites have no 100% treatment at all; one provider puts it at about 30% of sites. When 100% is out of reach, take the best treatment available, often 80%, and move on instead of burning the clock.
• A process score is reported too, and how it works is unknown. A 2018 paper by the Imbellus and McKinsey team behind Solve describes scoring both product (your result) and process (how you got there, read from clicks, actions and timing). It covers an earlier pilot, not Sea Wolf. How Sea Wolf scores process, and how much it weighs against efficiency, is not public, so work in a calm, consistent order.

Also unknown: whether the profile quietly changes which microbes you get, and where the pass bar sits. McKinsey publishes no pass rate or percentile, so treat any figure you see online as a prep provider's estimate. The takeaway: clear the five checks at every site, accept a capped site without chasing it, and keep your process orderly. The simulation on this page scores your treatment on the same five checks, and its debrief is practice evidence, not an official McKinsey score.

Worked example

Suppose a site needs:

• Energy average 3 to 5
• Adhesion average 6 to 8
• Speed average 2 to 4
• at least one Heat-Resistant microbe (desired trait)
• no Light-Sensitive microbe (undesired trait)

The trio needs totals of Energy 9 to 15, Adhesion 18 to 24 and Speed 6 to 12. Three microbes from your pool (the names are invented):

• Vorin Alga (Heat-Resistant): Energy 4, Adhesion 7, Speed 3
• Pelto Virus (Aerobic): Energy 2, Adhesion 6, Speed 4
• Quask Amoeba (Aerobic): Energy 6, Adhesion 7, Speed 2

Now check the sums against 3 times each end of the range:

• Energy = 4 + 2 + 6 = 12, inside 9 to 15, so the average is 4
• Adhesion = 7 + 6 + 7 = 20, inside 18 to 24, so the average is 6.67
• Speed = 3 + 4 + 2 = 9, inside 6 to 12, so the average is 3

All five checks pass: three averages in range, one desired trait, no undesired trait, so under the reported rule this treatment scores 100%. Notice that Pelto Virus and Quask Amoeba are each outside the Energy range on their own. The group fits, and the group is what counts. Swap Vorin Alga for a look-alike with the same numbers and a neutral trait, and the treatment drops to 80%. That is the practical point of Sea Wolf prep.

The four-step method that survives any version

Use one move per step: profile the hardest constraint, serve the current site first, pick for the trio, validate the set. Prep providers largely agree on these moves, even where they number the steps differently.

1. Profile the hardest constraint. In step 1, pick the narrowest or most extreme attribute range, plus the desired trait. The profile is not in the treatment score, so spend under a minute on it.
2. Serve the current site first. In step 2, judge each microbe against this site first, and save for the next site only what fits the part of it you can see. Reject only what would cost points, such as the undesired trait; one out-of-range value is not a reason, because averages can absorb it.
3. Pick for the trio, not the best single microbe. In step 3, look at which attribute your pool is short on and fill that gap. Skip the undesired trait.
4. Validate the treatment as a set. In step 4, drop undesired-trait microbes, keep at least one desired trait, then check each attribute sum against 3 times the low and high ends. If no trio reaches 100%, take the best one and move on.

This method is intentionally less glamorous than a solver. That is the point: it remains usable when the labels, values, or interface change.

How to prepare for Sea Wolf

The best prep is boring in a good way:

• Run one untimed attempt on the simulation above to learn the four steps and the vocabulary.
• Do quick average and range-to-sum drills. Timed case interview math practice, data interpretation reps, and scored mental math drills make the arithmetic reflexive.
• Practice screening trait rules without hesitation. Structure practice builds the same explicit constraint handling, the discipline covered in structured thinking for case interviews.
• Repeat under a clock, about 10 minutes per site, and keep the same checklist.
• Review by error type, not only by score: which step went wrong, then whether it was a missed rule, a wrong comparison, or a weak combination.

A timed self-simulation also works on paper: invent a few sites with three attribute ranges, a desired and an undesired trait, and a pool of ten microbes, then pick the best trio against a clock and score it with the 20% rule. It drills the same muscles: range-to-sum, trait screening and set-level checking.

If your invite does not list module names, do not panic about memorizing labels. Focus on clean arithmetic, calm reasoning, and reading the tutorial carefully once the task opens.

How to review a Sea Wolf practice attempt

Do not review only the final treatment. Find the first step where your reasoning left the rules. A weak treatment can begin with a profile that ignored the hardest range, a microbe rejected for one out-of-range value, a pool pick that doubled up on what you already had, or a trio that looked fine one microbe at a time.

Then replay the logic without changing the rule order. Improvement comes from making the checking routine dependable, not from discovering a new trick after every attempt.

How Sea Wolf fits into the rest of McKinsey prep

Sea Wolf concentrates on multi-constraint selection. Redrock concentrates on evidence retention, exhibit interpretation, and calculation: the Redrock study guide goes phase by phase and the Red Rock practice study runs the loop free. SFL concentrates on prioritization and stakeholder judgment, and the SFL practice simulation covers it. All three reward structured problem solving, but practicing only one does not reproduce the switching cost of a longer Solve invitation. Bain runs a comparable digital screen you can rehearse in the Bain SOVA simulator, and the MBB assessment simulator roundup maps every free option in one place.

Solve is only one gate. After that, you still need to handle live case interviews and the Personal Experience Interview.

Sources

• McKinsey Solve page (checked June 17, 2026): https://www.mckinsey.com/careers/mckinsey-digital-assessment
• McKinsey Problem Solving Game FAQ PDF (checked June 17, 2026): https://www.mckinsey.com/~/media/McKinsey/Careers%20REDESIGN/Interviewing/Main/McKinsey-Problem-Solving-Game-FAQ-v2.pdf
• McKinsey careers blog on Solve development (checked June 17, 2026): https://www.mckinsey.com/careers/meet-our-people/careers-blog/mck-problem-solving-game-team
• IGotAnOffer Solve guide (checked June 17, 2026): https://igotanoffer.com/blogs/mckinsey-case-interview-blog/mckinsey-problem-solving-game
• Kantar, McNulty, Snow et al., "Constructing Cognitive Profiles for Simulation-Based Hiring Assessments", EDM 2018 (checked September 25, 2026): https://educationaldatamining.org/files/conferences/EDM2018/papers/EDM2018_paper_19.pdf
• MConsultingPrep Seawolf deep dive, step structure and efficiency rules (checked September 25, 2026): https://mconsultingprep.com/mckinsey-solve-seawolf-deep-dive
• Hacking the Case Interview Sea Wolf guide, step structure and efficiency rules (checked September 25, 2026): https://www.hackingthecaseinterview.com/pages/mckinsey-sea-wolf-game

McKinsey and Solve are trademarks of McKinsey & Company. Road to Offer is independent and is not endorsed by McKinsey.