Redrock Study is the data-interpretation module most candidates see inside McKinsey Solve. The commonly reported length is about 35 minutes. You work wildlife (or similar research) scenarios: read sources, decide which numbers matter, calculate, then write a short report. It is not Sea Wolf, it is not a case interview, and it is not a leaked "real test."
Your invitation controls modules and time. The two-module sitting is often Redrock plus Sea Wolf (~30 minutes). Longer ~85-minute variants exist; see the Solve guide.
Last full pass: 24 August 2026. CoachNed is independent practice material, not affiliated with McKinsey.
The three phases (this is the Redrock skill)
| Phase | What you are doing | How people fail it |
|---|---|---|
| Investigation | Read the study pack. Sample size, period, definitions, what was counted vs estimated | Jump to questions; miss that "adults" ≠ "all animals" |
| Analysis | Percentages, weighted averages, growth, which chart actually answers the ask | Mix bases; treat a subsample as the population |
| Report | Short written or structured summary that matches the evidence | Over-claim causality; ignore the limitation you just read |
You will typically see several mini-cases (candidate reports often say 5–7) that get messier. Budget 5–7 minutes each; hard-stop near 8 so the last scenarios still exist. Solve notes that process (how you navigate) is scored alongside accuracy — published trackers often cite roughly 70/30 accuracy vs process on the data module. McKinsey does not publish a passing score.
Transfer practice (charts, percents, "does this exhibit support the claim"): graph drills and case math. Independent Redrock-style sim: Red Rock practice. Not a clone of McKinsey's item bank.
Worked example: weighted density, not a grocery SKU
Study pack (investigation). Researchers counted desert bighorn on three ranges. Range High: 40 km², 180 animals, counted in a wet year. Range Mid: 70 km², 210 animals, average year. Range Low: 90 km², 90 animals, drought year. A blog post claims "the metapopulation density is 3.6 animals per km², so the species is stable." A methods note says drought counts miss ~20 percent of animals (they stay in canyons). Sample: one aerial pass per range.
Question. Which statement is best supported?
- (i) Unweighted mean density is 3.6 / km².
- (ii) Area-weighted density is about 2.4 / km² before any drought correction.
- (iii) After a 20 percent drought correction on Range Low only, total animals are 570.
- (iv) The data prove the metapopulation is stable.
Analysis.
Densities: High 180/40 = 4.5. Mid 210/70 = 3.0. Low 90/90 = 1.0.
Unweighted mean = (4.5+3.0+1.0)/3 = 2.83, not 3.6. (i) is false. (3.6 is a trap: 180+210+90 = 480 and someone divided by 133 instead of 200, or averaged wrong.)
Area-weighted: total animals 480, total area 200 km² → 2.4 / km². (ii) is true.
Drought correction on Low only: 90 / 0.80 = 112.5 ≈ 113. Total 180+210+113 = 503, not 570. (iii) is false. (570 would be a 20 percent bump on all animals: 480×1.2=576 — the classic "applied the caveat to the wrong base.")
(iv) is false: one pass, mixed weather years, no time series. Stability is not in the file.
Report line. "Area-weighted density is 2.4 animals/km² on 200 km². The 3.6 figure is not supported. Drought undercount on the largest, emptiest range would raise the total modestly, not prove a trend. Limitation: single aerial pass; weather is confounded with range."
That is Redrock: bases, weights, caveats, no extra biology PhD required.
Process score: what the platform can see
From public candidate discussion summarized on the Solve guide: time on information panels before clicking, whether you skip sources, frantic toggling, external calculator / other tabs. Getting a right answer by spraying clicks can score worse than a slightly lower accuracy with a calm read.
Do not screenshot. Do not open another window. Use only what the assessment gives you.
Failure modes that are Redrock-specific
- Answering from the question stem. The lie is in the methods paragraph.
- Unweighted averages on unequal areas. The bighorn item exists because this is the usual miss.
- Causation in the report. "Stable," "caused by," "therefore we should translocate" — usually not supported.
- Eight minutes on mini-case 1. You just failed 4–6.
- Preparing Sea Wolf only. Optimization games do not train "which base."
- Buying leaked Redrock. Integrity plus the module rotates. Transfer skills only.
How this sits next to Sea Wolf and the case
Sea Wolf (~30 min) is constraint matching in an ocean site. Different muscle. Practise the switch: data honesty, then optimization, if your invite has both.
Passing Solve gets you to interviewer-led cases and PEI. Redrock math helps exhibits; it does not teach synthesis in a live room. After the assessment date, move to the case interview guide and practice cases.
Four-day transfer plan
Day 1. Read this page and your invite. Watch nothing that still demos only 2020 Ecosystem Building as "the test."
Day 2. 15 timed "does this chart support the claim?" items. Punish extra-textual leaps.
Day 3. Weighted averages and percent-of-percent until they are boring. Math drills.
Day 4. One full timed Redrock-style sitting on independent practice, then stop clicking and review where you changed bases.
McKinsey changes modules without a press release. If your screen names something else, follow the screen.