The candidates who struggle on Sea Wolf are rarely the ones who cannot read a range. They are the ones who build an elegant three-microbe team, then discover one average sits two points outside the target band — or that a forbidden trait snuck in on the third pick. Illegal-but-pretty loses to ugly-but-legal every time.
Sea Wolf is not a marine-biology quiz and it is not "pick the strongest microbes." It is a timed constraint-satisfaction task inside McKinsey Solve. After this guide you should know what McKinsey publishes, what is only candidate-reported, how to score a three-site decision, and which four errors burn the clock.
McKinsey's Game-Based Innovation Lab describes Sea Wolf as selecting and evaluating microbes to clean up contaminated sites — analytical thinking in context, not oceanography (McKinsey careers blog). Treat that as the floor. Everything more specific below is labeled candidate-reported.
What McKinsey publishes versus what candidates report
Start with the firm's pages. They matter because prep sites invent certainty McKinsey never offered.
| Claim | Status | Where it comes from |
|---|---|---|
| Solve is a gamified assessment of problem-solving | Firm-stated | Solve page |
| No preparation required or expected | Firm-stated | Same page; interviewing overview |
| Results considered with the rest of the application | Firm-stated | Solve FAQ (Aug 2025) |
| Length is in your invitation email; each task starts with an untimed tutorial | Firm-stated | Same FAQ |
| PC or Mac only; complete without interruptions | Firm-stated | Same FAQ |
| Scenarios vary so no two candidates share the same parameters | Firm-stated | Same FAQ |
| No AI tools, notes, screenshots, or outside help | Firm-stated | Solve page integrity rules |
| Sea Wolf = microbes cleaning contaminated sites | Firm-stated | GBIL careers blog |
| ~30 minutes for Sea Wolf alone; three sites; three microbes per site | Candidate-reported | Invite folklore and post-game accounts; confirm against your tutorial |
| Numeric attributes must average into site ranges; one required trait in; one forbidden trait out | Candidate-reported | Same — follow the on-screen rules if they differ |
| Paired with Redrock (~65 min total) or plus Sustainable Futures Lab (~85 min) | Candidate-reported | Invitation length is the only firm-controlled signal |
If your invitation says 65 or 85 minutes, believe the email, not a blog table. You will not see a Sea Wolf score or cutoff. The FAQ warns you may be asked at random to explain your logic later — write decisions as if someone might ask why you benched a strong microbe.
How the game works when you treat it as constraints
Candidate reports converge on a loop that looks like this. Your tutorial is the real rulebook; this is the mental model that matches what most people describe as of 2026.
- You get several contaminated ocean sites. Each site posts target ranges on numeric attributes (things like salinity, acidity, or nutrient load — labels vary) plus trait rules: at least one microbe must carry a required trait, and none may carry a forbidden trait.
- You browse a database of microbes. Each has numeric scores and categorical traits. Sorting and filtering are part of the work; McKinsey's public demonstration of Solve shows database search as a core action.
- For each site, you assemble a small team — commonly reported as three microbes. The team's average on each numeric attribute should land inside that site's range. Required trait present. Forbidden trait absent.
- You submit and move on. Time is shared with the rest of Solve, so dawdling on site one is a tax on site three.
What this is not: Ecosystem Building, the older food-chain placement game some 2021-era writeups still describe. If your invite or tutorial is about food webs and species cards, you are in a different module — see the ecosystem building guide. Sea Wolf, on McKinsey's own description, is microbes and cleanup.
The scoring intuition that survives scrutiny is simple. Process signals (did you open the database, sort, check averages before locking) and product quality (did the final teams satisfy the stated rules) both matter in Solve generally. McKinsey has never published Sea Wolf weights. Candidate folklore that "one illegal trait zeros the site" is plausible for a constraint game and should shape your play, but treat it as folklore until the tutorial says otherwise.
Worked decision: three sites, nine microbes, one trap
Invented numbers. Reproduce the arithmetic yourself. Assume each site needs a three-microbe team; averages must fall in range; required trait must appear at least once; forbidden trait must appear zero times.
Site A — Coastal lagoon
| Attribute / rule | Target |
|---|---|
| Acidity (avg) | 4.0–5.5 |
| Salinity (avg) | 20–28 |
| Required trait | Filter-feeder |
| Forbidden trait | Pathogen |
Microbe pool (subset)
| Microbe | Acidity | Salinity | Traits |
|---|---|---|---|
| M1 | 3.0 | 18 | Filter-feeder |
| M2 | 6.0 | 30 | Pathogen |
| M3 | 5.0 | 25 | Filter-feeder |
| M4 | 4.5 | 22 | Motile |
| M5 | 5.5 | 26 | Filter-feeder, Motile |
| M6 | 2.0 | 15 | Motile |
| M7 | 7.0 | 35 | Pathogen, Motile |
| M8 | 4.0 | 24 | None listed |
| M9 | 5.0 | 21 | Filter-feeder |
Bad team candidates often lock first: M3 + M5 + M2.
Averages: acidity (5.0 + 5.5 + 6.0) / 3 = 5.5 (in range). Salinity (25 + 26 + 30) / 3 = 27 (in range). Required trait: Filter-feeder present on M3 and M5. Forbidden trait: Pathogen on M2. The averages look clean. The team is illegal. In a case interview, this is the recommendation that ignores the client's hard constraint.
Legal team: M3 + M5 + M9.
Averages: acidity (5.0 + 5.5 + 5.0) / 3 = 5.17. Salinity (25 + 26 + 21) / 3 = 24. Filter-feeder present three times. No Pathogen. Slightly "weaker" looking microbes than M2, fully legal.
Near-miss that wastes time: M1 + M4 + M8.
Averages: acidity (3.0 + 4.5 + 4.0) / 3 = 3.83 — below 4.0. Salinity (18 + 22 + 24) / 3 = 21.3, inside range. Filter-feeder present on M1. No Pathogen. Two of three constraints pass; the acidity miss still fails the site. Single-attribute blindness.
Sites B and C use the same checklist. Recompute averages after every swap. If you cannot hit all four rules, drop the forbidden-trait carrier first, then the microbe pulling an average out of band. Incomplete is ugly; illegal is worse when the tutorial treats rule breaks as product failures — and most candidate accounts say it does.
Ned's rule. Before you chase a prettier average, ask whether the team is legal. If it is not legal, the prettier number is a distraction. In interviews I stop listening when someone optimizes a branch I already told them was out of bounds; Sea Wolf rewards the same discipline.
Errors that look like strategy
Optimizing before filtering. Candidates sort by the highest salinity score, then try to "fix" the team. Start by deleting anything with the forbidden trait for that site. The pool shrinks. Decisions get easier.
Checking one attribute. The lagoon example fails acidity while salinity looks fine. Treat each numeric range as a separate veto.
Ignoring the required trait until the end. Perfect averages with no Filter-feeder still fail. Keep one slot for a required-trait carrier until averages settle; swap it only if another carrier can hold the ranges.
Confusing Sea Wolf with Ecosystem Building. Food chains and reef placement are a different module. Those heuristics invent rules the Sea Wolf tutorial never stated.
Integrity shortcuts. McKinsey forbids apps, websites, AI, notes, screenshots, and outside help except written disability accommodations. Do the work yourself.
Inventing a cutoff. There is no public pass mark. Controllable inputs: read the tutorial, check every rule, finish all sites without panic-locking.
What "scored" likely means when the sheet is secret
McKinsey publishes that Solve measures analytical thinking, that parameters vary by candidate, and that recruiters weigh the result with the rest of the file. It does not publish a Sea Wolf rubric. From the interviewer chair, these are the behaviors I would bet show up in the data:
| Behavior | Why it maps to the job | How it shows up in Sea Wolf play |
|---|---|---|
| Constraint-first | Clients give hard rules | Forbidden traits out before maxing averages |
| Multi-factor control | Cases have several moving parts | Tracking two averages plus two trait rules at once |
| Evidence use | Open the data before guessing | Using sort/filter instead of random picks |
| Time allocation | Interviews are paced | Leaving buffer for the last site |
| Recovery | Bad branches get pruned | Swapping one microbe instead of restarting from zero |
If you practice only "find any legal team," you train the product side. If you practice only speed-clicking, you train neither. Aim for legal teams with a short audit trail you could explain aloud: "M2 was out because Pathogen; M9 replaced it; averages recomputed to 5.17 and 24."
One sitting that teaches the loop
Do this once on a quiet evening. Do not turn it into a week-long game camp. McKinsey says no prep is required; format familiarity is the only honest reason to practice.
- Read your invitation email for total Solve length. Note whether you likely have Redrock only with Sea Wolf, or a third module.
- Open CoachNed's Sea Wolf–style practice at /mckinsey-sea-wolf. Everything on the site is open for seven days, no card; then $120 for a recruiting season or $49 a month.
- Run one timed round. After each site, write one sentence: which rule almost broke, and which microbe you nearly kept illegally.
- Run a second round with a hard rule: no submission until you have written the four checks (avg1, avg2, required, forbidden) on paper.
- Stop. Sleep. The skill transfers to cases — kill infeasible branches before you polish synergy — but Sea Wolf reps do not replace interviewer-led case practice or PEI.
If Redrock is also on your invite, keep the tempos separate: Sea Wolf is fast constraint math; Redrock is slower investigation and report writing. Mixing the mindsets mid-assessment is a common candidate-reported failure.
Frequently asked questions
How long is the McKinsey Sea Wolf game?
McKinsey does not publish a fixed Sea Wolf duration. The Solve FAQ says your invitation email states the full assessment length, and each task begins with an untimed tutorial. Candidate reports often describe Sea Wolf itself as roughly 30 minutes inside a longer Solve sitting — treat that as folklore until your invite and tutorial say otherwise.
Is Sea Wolf the same as Ecosystem Building?
No. McKinsey's own description of Sea Wolf is microbe selection for contaminated sites. Ecosystem Building is an older food-web style module still discussed in older candidate accounts. Confirm which modules your invitation and on-screen tutorials actually launch.
Does McKinsey publish Sea Wolf scoring or a cutoff?
No. Official guidance is that Solve results are considered with the rest of your application and any other assessments, not as a standalone pass/fail number you can see. Anyone selling a "guaranteed cutoff" is inventing a figure the firm has not released.
Can I use notes, another screen, or AI during Solve?
McKinsey's Solve page prohibits applications, websites, AI tools, pre-prepared notes, screenshots or recordings, and help from others, unless you have a written disability accommodation from McKinsey. Violations can end consideration or worse. Play alone on one machine.
Should I practice Sea Wolf if McKinsey says preparation is unnecessary?
The firm says no preparation is required or expected. Format familiarity — knowing you will filter constraints before you optimize averages — is optional insurance, not a substitute for reading the tutorial carefully on test day. A short independent practice sitting is enough; multi-week grinding is not.
Sources
- Solve, McKinsey's assessment game — official description of Solve, integrity rules, FAQ link. Checked 2026-09-24.
- Solve FAQ (August 2025 PDF) — invitation length, untimed tutorials, no business/gaming prerequisite, multi-factor decisioning, scenario variation. Checked 2026-09-24.
- From apprenticeship in space to selecting microbes (McKinsey careers blog) — firm description of Sea Wolf as microbe selection for contaminated sites; Solve volume among early-career candidates. Checked 2026-09-24.
- Interviewing at McKinsey — Solve as gamified assessment for most consulting roles; no preparation required. Checked 2026-09-24.
CoachNed is independent practice material and is not affiliated with McKinsey & Company.