The Sea Wolf (Ocean Cleanup):
rules and worked examples.

The Sea Wolf is the second of the three McKinsey Solve games. For the overview of all three, including durations, versions and official rules, start with the complete McKinsey Solve guide. This page goes deeper on Sea Wolf only: how a site works, what you are trying to achieve, and a full site solved by hand.

What Sea Wolf asks you to do

You play a researcher treating three contaminated ocean sites. At each site you choose a treatment of three microbes, and the treatment is judged on how well the three together match the site's requirements. McKinsey does not publish the length of each game, but candidate reports consistently describe about 30 minutes for the three sites, with the same version in both the 85-minute and the 95-minute invitations.

Each microbe has three numeric attributes, scored from 1 to 10, and one trait. Each site sets a required range for every attribute, one desired trait and one undesired trait. Attribute names vary between reports; the example on this page uses Permeability, Mobility and Energy, the labels our simulator uses, and four traits: Aerobic, Heat Resistant, Light Sensitive and Phosphorus Removal.

The key word is together. For the attributes, what matters is the average of the three microbes, not each microbe on its own. For the traits, it is enough for one of the three to carry the desired trait, and it is enough for one of them to carry the undesired trait to lose that condition.

The four stages of each site

Candidate reports describe the same sequence at every site, and our simulator follows it. Some guides count five steps because they treat the final confirmation as a separate step; the work is the same.

1. Characteristics. You choose two characteristics to prioritise: two attributes, two traits or one of each. In our simulator, as in the reports, this choice shapes the starting prospect pool but not the microbes you sort next.
2. Categorise microbes. Ten microbes appear one at a time, and you send each to Current Site, Next Site or Return. The game's own guidance is to look for microbes with at least two attributes inside the site's ranges and without the undesired trait. A short insight about the next site tells you what is worth sending forward.
3. Prospect selection. Your pool starts with six microbes. Four times, you pick one microbe out of three offered, so the pool ends with ten. Two microbes offered in the same round can never both be in your pool.
4. Treatment. You choose three microbes from the pool of ten and confirm. Once confirmed, the site is closed and you move to the next one.

At Site 2 and Site 3, the microbes you sent forward come back first for a second sort, then the four stages repeat. This repetition is the main reason Site 1 feels slow and Sites 2 and 3 feel faster: you are no longer learning the interface, only applying the same routine to new numbers.

Rules of the task versus conditions to optimise

Before deciding what to exclude, it helps to separate two different kinds of condition. Some are rules of the task: the game enforces them, and you cannot break them even if you want to. Others are objectives: the game lets you miss them, and each one you miss costs part of the score.

Use these five conditions to check each treatment. A combination that meets all five may not be available, so compare the conditions each option meets. McKinsey does not publish its scoring weights; the method below focuses on meeting the stated requirements rather than assuming a particular points formula.

None of the five is a hard filter, which changes how you should read a pool:

• Set aside first: microbes with the undesired trait, because any treatment containing one misses that condition. Reconsider them only if the remaining pool cannot produce a better overall combination.
• Anchor on: microbes with the desired trait. You need one, not three, so pick the anchor first and build around it.
• Only compare: microbes outside one or more ranges. Because the score depends on the average, a high value can offset a low one. Excluding them on sight is the most common way to throw away the best treatment, as the example below shows.

Lukas, on writing the conditions down before touching anything.

Turn averages into totals

Checking whether an average falls in a range means adding three numbers and dividing by three, for each attribute and for every treatment you try. The division is the slow, error-prone part, and you can remove it once per site.

If the average of three numbers must be between A and B, then their total must be between 3A and 3B. So you multiply the ends of each range by three once, write the totals next to the ranges, and from then on you only add. At least one detailed candidate report describes no calculator in Sea Wolf, and the numbers are small whole values, so this is easy to do in your head.

The same idea gives you a second shortcut. Once you fix one microbe, you can subtract its values from the totals and see what the other two must add up to. That remaining budget tells you, before any trial, which pairs can and cannot work.

A worked example: one site solved by hand

The site, microbes and values below are fictional and written for this guide. They follow the structure described above, but they are not taken from the real assessment or from our paid scenarios.

The final check uses three sums and a check of the two traits. Calculating the partner ranges first ruled out three pairs before you needed to test them.

Two treatments compared

Now compare the treatment above with the one many candidates build first. The instinct is to keep the microbes that look right individually and add an Aerobic microbe to meet the trait condition. With A and J as the two "safe" picks, the treatment is A + J + B.

Both treatments contain B and J. The only difference is the third microbe: A, which is inside every range, against H, whose Energy of 3 is below the range. Yet H is the right choice, because the treatment needs a low Energy value to offset B's 9, and A's 6 is too high to do that. A + J + B misses the Energy range by one point.

The general lesson is that a microbe is never good or bad on its own; it is good or bad for a given treatment. In this fictional pool, every treatment that meets all five conditions contains C or H, the two microbes a "keep only what fits every range" filter would have thrown away. A treatment of A + F + J, all three individually perfect, still misses a condition, because none of them is Aerobic.

Use the earlier stages to set up the treatment

The treatment is where the score is reported, but the earlier stages decide what you have in your pool when you get there. A few habits help.

Characteristics. Choose the attribute range that looks hardest to meet and the desired trait. The first narrows the pool towards the values you need; the second makes it more likely that you start with an anchor.

Categorise. Return microbes with the undesired trait unless the next-site insight says they are useful there. Keep microbes with two attributes in range, and send forward those that fit the insight about the next site better than the current one.

Prospect selection. Each round offers three microbes and you keep one. Think in terms of the treatment, not the single microbe: if your pool has no anchor yet, take the microbe with the desired trait; if it already has one, take the microbe that offsets your anchor's extreme values, as H offsets B above.

Pacing: about ten minutes per site

With about 30 minutes for three sites, ten minutes per site is a useful reference, and it is the split most preparation sources suggest. Site 1 usually takes longer because you are still learning the screens, so aim to recover a little time on Sites 2 and 3 rather than rushing the first.

A full score is not always available. The game itself says a fully effective combination may not exist, and a site where the best treatment meets four conditions out of five is normal. When you have a treatment that meets four conditions and a few minutes of searching have not found a fifth, confirm it and move on. A good result on all three sites is worth more than a perfect first site followed by two rushed ones.

Treat this split as a starting point and adjust it after your first timed run.

What our practice report adds

After a run, it is hard to tell from memory why a site scored less than you expected. The detailed report in full access shows, for each site, the three microbes you sent, each of the five conditions met or missed with your totals against the targets, and the time each site took. It also shows which conditions you missed most often across the three sites, so you know what to focus on in your next run.

Aisha, on using a checklist of conditions and seeing in the report where she lost time on the third site.

A practice routine for Sea Wolf

A few reviewed runs teach more than many runs played without a review.

Play one timed run to see where you lose time or conditions, practise that point, then play again and compare. Sea Wolf comes after Redrock Study, whose rhythm is quite different, and before the Sustainable Futures Lab, which is about judgement rather than numbers.

Where to start

The quickest way to learn the four stages is to play them once under time. Start with the free Sea Wolf run, then use the totals and budget method from the example on your next attempt.

About Franco

Franco Borgonovo

Former BCG Principal · 10-year tenure · Official BCG coach

Franco spent ten years at BCG, from intern to Principal, interviewing and selecting candidates as a global lead recruiter. He is still an official BCG coach for new hires, and today works one to one with candidates on applications, McKinsey Solve and interviews.

More about Franco →