How this was measured
- Practice runs on SolveForge's browser recreations of two McKinsey Solve games. The games are rebuilt from public descriptions, and the scoring is ours, not McKinsey's.
- McKinsey doesn't publish its scoring or a pass mark, and nothing here predicts a McKinsey result.
- Free players replay one fixed scenario, so their later runs would overstate their skill. For anyone without a McKinsey pass we count only their first run of each game. Every run made with a pass counts, because each one is a new scenario. That leaves 152 of 211 Red Rock runs and 113 of 152 Sea Wolf runs; 30 of the 265 were made with a pass.
- Our own test accounts are left out. Only totals are shown, never a player.
- We publish no figure based on fewer than 30 runs or 20 players.
Red Rock Study
The score is the share of questions answered correctly across the study and the six cases. 152 runs from 138 players.
Share of the 152 counted Red Rock runs in each score band.
Time used
51 runs from 47 players recorded the clock, all on the 30-minute version our simulator used until 10 October 2026 (it now runs 35 minutes, as candidates report for the real game). The median run took 14 minutes 47 seconds, and the middle half took 11:14 to 18:39. Only 4% ran out of time. Most people finish early, which is worth knowing: the time pressure is real but rarely binding, and the points go on accuracy.
The question types missed most often
From the same 51 runs, the share of attempted questions of each type that were answered wrong. Each type was attempted in at least 47 runs by at least 44 players.
| Question type | Missed |
|---|---|
| Weighted average (case) | 48% |
| Probability (case) | 46% |
| Percent change | 28% |
| Percent versus percentage points | 27% |
| Compound growth | 24% |
| Mean versus median (case) | 23% |
| Totals from the right rows | 21% |
| Copying values into the report table | 19% |
| Choosing the chart | 19% |
| Rates versus raw counts (case) | 13% |
Most of these runs are each player's first go at the same free study, so the rates partly reflect that study's particular numbers. Still, the top two are the late cases, where people are tired and short of time, and both need a formula rather than a lookup. The Red Rock guide works through each one.
Sea Wolf
The score is the average treatment efficiency across the three sites. 113 runs from 105 players.
Share of the 113 counted Sea Wolf runs in each score band.
Sea Wolf scores bunch at the top: a site is either right or loses at least 20 points, so runs land at 100 or in steps below it. The cluster in the 60s is mostly runs that got one site badly wrong.
Time used
32 runs from 29 players recorded the clock, on a 30-minute clock. The median run took 16 minutes 58 seconds, the middle half 14:24 to 23:19, and 3% ran out of time.
What goes wrong on a site
The same 32 runs cover 96 sites. 57% of sites were treated perfectly. On 38%, at least one attribute's average across the three microbes fell outside the site's range, which is by far the most common way to lose points. Missing the desired trait (7% of sites) and picking a microbe with the undesired trait (7%) were much rarer.
Adhesion was the attribute most often out of range (28% of sites), ahead of Speed (18%) and Energy (17%). Most of these runs use the free set of sites, so treat the attribute ranking as specific to those sites. The general lesson holds: check the average of your three against every range before you commit. The Sea Wolf guide shows how.
What we can't show yet
- Whether practice raises scores. We wanted to compare each player's first and third runs, but only 3 players on Red Rock and 2 on Sea Wolf have three counted runs. We'll add it when 20 players qualify.
- The Sustainable Futures Lab and the Bain sections. None has 30 runs from 20 players yet.
- The 35-minute Red Rock clock. Runs on it start from 10 October 2026, so the time figures above are all from the 30-minute version.
We refresh these figures monthly. The date at the top says when the data was taken.
Using these numbers
Compare your practice score with the median for the same game, not with a number someone quotes as McKinsey's cut-off: McKinsey doesn't publish one. A first-run Red Rock score in the 60s is normal here. If yours is well below that, the question-type table shows where to look first. Our page on how the McKinsey Solve is scored covers what's known about the real scoring.
To get comparable numbers of your own, play the Red Rock and Sea Wolf simulators. One fixed scenario of each is free; a McKinsey pass gives you a new one every run.
FAQ
What is a good McKinsey Solve practice score?
On SolveForge, the median practice run scores 75% on Red Rock and 87% on Sea Wolf, from 265 counted runs up to 11 October 2026. These are practice scores on our recreations. McKinsey doesn't publish its scoring or a pass mark, so no practice score tells you whether you'll pass.
Are these McKinsey's real scores?
No. They come from practice runs on SolveForge's browser recreations of Red Rock Study and Sea Wolf, scored by our own rules. SolveForge is not affiliated with McKinsey and has no access to McKinsey's results.
How long do people take on Red Rock practice?
On our 30-minute clock, the median run took about 15 minutes and only 4% of runs ran out of time, from 51 runs that recorded the clock. Our simulator now runs 35 minutes, matching candidate reports for the real game.
Which Red Rock questions do people miss most?
In our practice runs, the weighted-average case (48% missed) and the probability case (46%) were missed most, followed by percent change and percent versus percentage points (about 27% each). Most runs were first attempts at our free study, so the exact rates partly reflect that study.
SolveForge is an independent practice tool and is not affiliated with or endorsed by McKinsey & Company. Red Rock Study, Sea Wolf and the Sustainable Futures Lab are names used by McKinsey for its assessment; our games are recreations from public descriptions.