The arithmetic on Red Rock Study is rarely what sinks people. The base is. Wrong denominator, an unweighted average of unequal areas, a report sentence that claims causality the study pack never supports: those are the quiet ways a data module bleeds points while you feel busy and correct.
Red Rock Study (also called Redrock or Red Rock Island in candidate reports) is the data-interpretation task inside McKinsey Solve, the firm's gamified recruiting assessment. McKinsey's public pages describe Solve as a whole; they do not publish a full Red Rock scoring sheet or a fixed minute count for every invite. This guide separates what the firm states from what candidates commonly report, then walks the format itself: phases, question types, pacing, what appears to be scored, common errors, and one original arithmetic item you can reproduce on paper.
After you finish, you should be able to open a study pack, name the base before you touch a calculator, and write a report line that matches the evidence without inventing a trend.
What McKinsey publishes versus what candidates report
Start with the firm's own language. On its Solve careers page, McKinsey calls Solve a gamified assessment meant to showcase problem-solving, says there is no need to prepare, and states that performance is considered with the rest of your application and any other assessments. The same page sets hard integrity rules: no applications, websites, AI tools, or pre-prepared notes; no screenshots or recordings; complete the assessment alone. Violations can mean exclusion, a rescinded offer, or future restrictions.
McKinsey's Solve FAQ adds operational facts that matter for Red Rock prep: the assessment length is in your invitation email; each task starts with an untimed tutorial; Solve runs on a PC or Mac; no prior business or gaming experience is required; scenarios use variation so candidates do not see identical parameter sets. The interviewing overview repeats the same high-level Solve description for most consulting roles.
A McKinsey blog on its game-based innovation lab says Solve reaches roughly 300,000 early-career consultant candidates per year and names Sea Wolf (microbe selection for contaminated sites) as an example of how thinking is assessed in context. Red Rock itself is not specified in that write-up the way Sea Wolf is.
Everything below about Red Rock's internal phases, clock length, mini-case count, and "process" tracking is candidate-reported unless labeled otherwise. Treat your invitation and on-screen tutorial as the authority if they disagree with any guide, including this one.
The format candidates describe
Candidate reports converge on a research-island setting: you act as a junior analyst reading a study pack (tables, charts, methods notes, sometimes a journal), calculating answers, then summarizing. The flow people describe most often has four beats:
| Beat | What you do | What usually goes wrong |
|---|---|---|
| Investigation | Scan sources; note definitions, sample size, period, counted vs estimated | Jump to the question stem and miss that "adults" is not "all animals" |
| Analysis | Percentages, weighted averages, growth, which exhibit answers the ask | Mix bases; treat a subsample as the population |
| Report | Short written or structured summary tied to the evidence | Over-claim causality; ignore the limitation you just read |
| Cases (mini-scenarios) | Several shorter data items after or alongside the main study | Spend eight minutes on the first item and starve the last ones |
Names and order can shift on screen. Some candidates report Investigation–Analysis–Report as a single continuous study, then a run of shorter cases. Others describe the whole sitting as a sequence of mini-cases that get messier. Either way, the skill is the same: lock the unit, pick the right exhibit, calculate once, then say only what the file supports.
McKinsey does not publish a guaranteed module list for every applicant. Candidate reports often place Red Rock next to Sea Wolf on roughly 65-minute invitations, with a longer sitting (often cited near 85 minutes) sometimes adding Sustainable Futures Lab. Confirm the clock and module names on your invite.
Question types worth drilling
You are not being tested on wildlife biology. You are being tested on whether you can keep units honest under time pressure. The patterns that show up again and again in candidate accounts:
| Pattern | Typical trap | Fix before you calculate |
|---|---|---|
| Rate or density | Averaging rates of unequal areas or periods | Prefer totals / total base over mean of rates |
| Percent of percent | Applying a correction to the wrong population | Write "of X" next to every percentage |
| Subsample vs population | Using a chart of adults to answer "all animals" | Match the question's noun to the exhibit's noun |
| Exhibit selection | The flashiest chart is not the one that answers the ask | Start from the question, then hunt the matching table |
| Supported claim | Report language that invents a trend or a cause | If the pack has one pass and no time series, you cannot claim "stable" |
If you can do those five without a stopwatch, timed practice gets a lot cheaper.
Timing (candidate-reported; invite wins)
Candidates commonly cite about 35 minutes of timed Red Rock work, separate from the untimed tutorial McKinsey describes for each Solve task. Mini-case counts in public discussion often land around 5–7 items. A useful pacing heuristic, labeled as Ned's estimate rather than firm doctrine: budget 5–7 minutes per mini-case and hard-stop near 8 so the last scenarios still exist.
Do not bank unused Red Rock minutes for Sea Wolf. Candidate reports treat module clocks as separate. If your invitation shows a different total assessment length, use that number and the tutorial screens, not a blog table.
What appears to be scored
Officially, McKinsey says Solve results are considered with the rest of the application; it does not publish a Red Rock pass mark or a public accuracy/process split. Candidate discussion frequently claims the platform also tracks how you navigate: time on information panels, frantic toggling, skipping sources, or reaching a right answer by spraying clicks. Treat "process score" ratios you see on forums as folklore until the firm publishes them.
What you can take as firm policy from McKinsey's own Solve page: external tools, AI, notes, screenshots, and help from others are out of bounds. Getting a number right while violating integrity is not a win. Use only the calculator and materials the assessment provides.
Worked example: three salt-marsh patches (lock the base)
Invented study pack. Not a McKinsey item.
Investigation. Ecologists counted nesting terns on three marsh patches in one breeding season. Patch A: 12 ha, 96 nests. Patch B: 28 ha, 140 nests. Patch C: 40 ha, 80 nests. A methods note says Patch C was surveyed after a spring flood that washed out an estimated 25 percent of nests before the count. A press clipping claims "average nest density is 5.3 nests per hectare, so the metapopulation is recovering." Sample: one ground count per patch; no prior-year series in the file.
Question. Which statement is best supported?
- (i) The unweighted mean of the three patch densities is 5.3 nests/ha.
- (ii) Area-weighted density across all three patches is 3.95 nests/ha before any flood correction.
- (iii) After correcting Patch C only for the 25 percent washout, total nests are 421.
- (iv) The data prove the metapopulation is recovering.
Analysis.
Densities: A = 96/12 = 8.0. B = 140/28 = 5.0. C = 80/40 = 2.0.
Unweighted mean = (8.0 + 5.0 + 2.0) / 3 = 5.0, not 5.3. (i) is false. (5.3 is a trap: someone averaged 96, 140, and 80 as if they were densities, or divided total nests by something other than total hectares.)
Area-weighted: total nests 96 + 140 + 80 = 316. Total area 12 + 28 + 40 = 80 ha. Density = 316 / 80 = 3.95 nests/ha. (ii) is true.
Flood correction on Patch C only: counted nests are 75 percent of pre-flood nests, so pre-flood C ≈ 80 / 0.75 ≈ 106.7. Total ≈ 96 + 140 + 107 = 343, not 421. (iii) is false. (421 would be a 25 percent bump on all nests: 316 × 1.25 = 395 — close cousin of the classic "applied the caveat to the wrong base" — or another arithmetic mash of the three patches.)
(iv) is false: one season, flood already confounds Patch C, no time series. "Recovering" is not in the file.
Report line. "Area-weighted density is 3.95 nests/ha on 80 ha before flood adjustment. The 5.3 figure is not supported by an unweighted mean of patch densities. Correcting only Patch C raises the total modestly and still does not establish a trend. Limitation: single ground count; flood timing is confounded with the largest, lowest-density patch."
That is the Red Rock muscle: bases, weights, caveats. No ecology PhD required.
Ned's rule. Before any division, write the unit of the denominator in three words. If you cannot name it, you are not ready to calculate. Most Red Rock "math errors" are unnamed bases wearing a calculator costume.
Failure modes that are Red Rock-specific
- Answering from the stem. The lie usually sits in the methods paragraph or the footnote on the chart.
- Mean of rates on unequal bases. The salt-marsh item exists because this miss is routine.
- Causation in the report. "Recovering," "caused by," "therefore relocate breeders" — usually unsupported.
- Eight minutes on mini-case one. You just traded the last two or three items for pride on the first.
- Preparing only the optimization game. Sea Wolf does not train "which base."
- Treating leaked item banks as the test. Integrity risk, plus McKinsey's FAQ states parameters vary. Transfer skills only.
Practice this once, on a clock
One concrete assignment: open CoachNed's Red Rock–style practice at /mckinsey-red-rock, run one timed sitting, and after each miss write the base you should have used in one line. Do not chase a second full run until you have reviewed every wrong base. For the arithmetic habits that transfer into live cases, use /drills/math and /drills/charts. Everything on CoachNed is open for seven days with no card; then $120 for a recruiting season or $49 a month. The sibling Red Rock simulator guide covers how to use the practice product; this page is the format itself.
When Solve is behind you, shift to interviewer-led cases: the McKinsey case interview guide and practice cases. Exhibit honesty from Red Rock helps in the room; it does not replace synthesis under a partner's questions.
CoachNed is independent practice material and is not affiliated with McKinsey & Company. McKinsey changes modules without a press release; if your screen names something else, follow the screen.
Frequently asked questions
How long is McKinsey Red Rock Study?
McKinsey does not publish a fixed Red Rock duration on its public Solve page. It says the assessment length is in your invitation and that each task has an untimed tutorial. Candidate reports commonly cite about 35 minutes of timed Red Rock work; confirm your invite.
Is Red Rock Study the same as McKinsey Solve?
No. Solve is the umbrella assessment. Red Rock Study is the data-interpretation module many candidates report seeing inside Solve, often alongside Sea Wolf and sometimes Sustainable Futures Lab. Module mix varies; your invitation is the source of truth.
Does McKinsey publish a Red Rock passing score?
No. McKinsey states that Solve performance is considered with the rest of your application and other assessments. Public forums invent numeric cutoffs; treat those as unverified.
Can I use a calculator or ChatGPT during Red Rock?
Use only what the assessment provides. McKinsey's Solve page prohibits applications, websites, AI tools, pre-prepared notes, screenshots, and assistance from others, except disability accommodations approved in writing.
Should I memorize wildlife facts for Red Rock Island?
No. McKinsey says Solve does not require prior business or gaming experience. The transfer skill is exhibit math and evidence discipline, not species trivia.
Sources
- Solve, McKinsey's assessment game — official description of Solve, integrity rules, and how results are used. Checked 2026-09-24.
- Solve FAQ (PDF) — invitation length, untimed tutorials, PC/Mac, no business knowledge required, scenario variation. Checked 2026-09-24.
- Interviewing at McKinsey — Solve as a step for most consulting roles; case and personal experience interviews described separately. Checked 2026-09-24.
- From apprenticeship in space to selecting microbes — ~300,000 early-career candidates per year; Sea Wolf example; Imbellus/GBIL context. Checked 2026-09-24.