Most candidates "practice Red Rock" by doing loose percentage problems on a phone. That is not the skill McKinsey is buying. Red Rock Study sits inside Solve, McKinsey's gamified assessment, and the failure mode I see is not weak algebra. It is losing the plot when reading, evidence selection, calculation, and a short written answer arrive on the same clock.
The thesis is simple: a useful Red Rock simulator must force the full loop under time, then give you a way to score the loop, not only the final digit. Format detail and question families belong in the sibling Redrock study guide. This piece is about how to practice with a simulator, what "good" looks like, and how to use CoachNed's timed Calderas Mining drill at /mckinsey-red-rock without pretending it is McKinsey's item bank.
After reading, you should be able to audit any practice tool in two minutes, score your own runs with a four-line rubric, and run a five-day plan that transfers to the live assessment.
What Solve actually publishes about Red Rock practice
McKinsey's public line on Solve is consistent and deliberately thin. On the official Solve page, the firm describes a gamified assessment of problem-solving ability, says there is no need to prepare, and states that performance is considered with the rest of the application. The interviewing overview places Solve as a step for most consulting roles and repeats the same "curiosity and creativity" framing.
Duration is not a single public constant. McKinsey's Solve FAQ (as of the August 2025 PDF) says the length is outlined in your invitation email, each task starts with an untimed tutorial, and the assessment runs on a PC or Mac after a tech check. A regional careers page for Finland still describes Solve as approximately 60 minutes online. Treat those figures as orientation; your invite is the rulebook.
Integrity rules are not optional folklore. McKinsey's assessment integrity expectations require you to participate alone, use one screen, and refrain from AI tools, outside websites, prepared notes, recordings, or screenshots. Practice happens before the window. During the window, you work alone with what the tutorial allows.
What McKinsey does not publish is a Red Rock scoring matrix, a pass mark, or a guarantee that every invitation includes the same module mix. Candidate reports about Investigation, Analysis, Report, mini-cases, and shared clocks live on forums and prep sites. Useful for orientation. Not primary fact. If a simulator claims "exact McKinsey scoring," treat that as marketing.
The four moves a simulator must force
When I debrief candidates who froze on Solve, the same gap appears: they practiced one move in isolation. A Red Rock-style exercise is a research-and-calculate loop. A simulator earns its keep only if it reproduces pressure on all four moves below.
| Move | What you do | What flimsy practice skips |
|---|---|---|
| Timed reading | Restate the objective, then scan exhibits for inputs that enter the answer | Untimed reading with unlimited rereads |
| Data extraction | Keep only labeled facts; ignore decorative numbers | Highlighting every cell "just in case" |
| Calculation | Write the identity, then compute with units and scale visible | Punching a calculator without an equation |
| Written answer | Commit to a number and a one-line so-what | Stopping at "I got 1.87" with no implication |
If a tool only quizzes "what is 22% of 2.4," it is a mental-math drill wearing a Red Rock hat. Mental math still matters; use /drills/math or /drills/quick-math for that. It is not a substitute for the loop.
The written-answer step is the one most free tools omit. In a live study you do not only compute; you also have to put the result somewhere that looks like a junior-analyst output: a blank, a chart choice, a short claim. Even a one-line "payback is about 3.6 years; take it if the fleet lasts past five" trains the so-what muscle interviewers reward later in cases.
How CoachNed's sampler maps to those four moves
CoachNed's Red Rock practice at /mckinsey-red-rock is an original, independent micro-study. It is not McKinsey's game, not a clone of Redrock Island wildlife content, and not a promise of any score. It is a short stand-in for the research-and-calculate feel: one haul-fleet exhibit, an on-page calculator, four forward-only numeric questions, and a ten-minute clock.
The setting is Calderas Mining. Diesel haul opex is $2.4m a year. Electric cuts opex 22% and adds $1.1m of capex amortized over five years. You see a compact cost table, answer four questions in order, and get a walkthrough when the timer ends or you finish. No account is required to start. Everything on CoachNed is open for seven days with no card; then $120 for a recruiting season or $49 a month.
Honest gaps matter. The sampler compresses Investigation into a single table rather than a multi-exhibit journal. It does not run a full Report with chart selection. It scores correctness of four numbers, not McKinsey's unpublished process model. Use it to rehearse timed reading, extraction, calculation, and a deliberate so-what after the scorecard. Use the sibling study guide for format lore. Use Sea Wolf and SFL simulators for the other Solve skill families.
Worked example: score a Calderas run like an interviewer
Reproduce this on paper in under four minutes, then compare to the sampler's walkthrough.
Diesel opex = $2.4m. Electric opex before capex = 2.4 × (1 − 0.22) = 2.4 × 0.78 = $1.872m. Capex amortization = 1.1 / 5 = $0.22m a year. Fully loaded electric = 1.872 + 0.22 = $2.092m. Net reduction versus diesel = 2.4 − 2.092 = $0.308m, or 0.308 / 2.4 ≈ 12.8%. Simple payback on the $1.1m = 1.1 / 0.308 ≈ 3.6 years.
A candidate who submits 1.872, 2.092, 12.8, and 3.6 has the arithmetic. That is necessary and not sufficient. I still ask: did they write the identity before touching the keypad? Did they keep $m opex and $m capex in the same units? Did they amortize only when the prompt required it? Did they say whether 3.6 years is fast for a haul fleet that lasts past five?
Two common miss patterns on this exact setup:
| Answer pattern | Likely break | Fix on the next rep |
|---|---|---|
| Quotes 22% as if it were dollars | Skipped the identity | Write "opex_e = opex_d × (1 − s)" before any digit |
| Reports 1.872 as the "fully loaded" cost | Forgot amortization | Circle "capex / life" whenever the prompt says annual cost including |
| Uses 22% as the net save after capex | Confused percent save with net reduction | Compute dollars first, then percent |
| Finishes with 3.6 and no recommendation | Treated the number as the decision | Add one sentence: take / wait / need more life data |
That fourth row is why the written-answer step belongs in practice even when the UI only accepts a number. Say the so-what aloud after every submit.
How to score yourself without a published pass mark
McKinsey will not give you a Red Rock threshold. Do not invent one. Score yourself on transfer, not on superstition. After each timed run, fill this four-line card before you look at any walkthrough.
| Line | Grade 0 or 1 | Pass look |
|---|---|---|
| Objective restated in one sentence before opening every cell | 0 / 1 | You can recite the exact quantity asked |
| Every number used appears with a unit and a source label | 0 / 1 | No orphan digits |
| Equation written before the calculator | 0 / 1 | Identity visible on paper or scratch |
| So-what spoken in one sentence after the final number | 0 / 1 | Decision implication, not just the digit |
A run that hits 4/4 on process and misses one tolerance band is better practice than a lucky 4/4 on answers with 0/4 on process. I would rather hire the first candidate: they will catch themselves in a live case when a partner changes an assumption mid-sentence.
Track two series across five days: process score (0–4) and answer score (percent correct). Process should rise first. Answer score often lags by a day because units and amortization only stabilize after the habit sticks.
Ned's rule. If you cannot write the equation without looking at the exhibit a second time, you did not extract; you memorized. Extraction means the inputs live in your notes with labels. Memory is what fails when the tutorial adds a third table.
A five-day practice plan that respects the invite
Do not burn a week on folklore about wildlife islands. Burn it on the loop.
| Day | Assignment | Done when |
|---|---|---|
| 1 | One untimed Calderas pass; stop the clock | You can explain each of the four answers from the identity alone |
| 2 | Twenty minutes of percent, percentage-point, and amortization drills at /drills/math | You never confuse a percent save with a dollar save |
| 3 | One timed Calderas run; fill the four-line process card before the walkthrough | Process score is at least 3/4 |
| 4 | Second timed run; change only one habit (units, amortization, or so-what) | That habit holds for all four questions |
| 5 | Light review, tech check of your own machine, and a read of McKinsey's integrity rules | You know the invite length and will work alone |
If Day 3 process score is still 0–1, do not add Sea Wolf yet. Fix the Red Rock loop first. Solve is multiple games; each has its own clock and skill family. Stacking simulators before the habit sticks is how candidates stay busy and stay stuck.
On test day, complete McKinsey's tech check, use a stable connection, and treat the on-screen tutorial as law. The FAQ notes that scenarios vary by candidate and that you may be asked later to explain your logic. Practice that explanation now: after Calderas, say the so-what without looking at the walkthrough.
Practice this today
Open /mckinsey-red-rock. Start the ten-minute Calderas sampler. Before you submit question one, write the electric-opex identity on paper. After the scorecard, grade yourself on the four-line process card above and pick one habit for tomorrow. If you want a longer case loop after that, use /start for a five-minute typed first rep, or /interview for the live voice case with Ned.
CoachNed is independent and is not affiliated with McKinsey & Company. McKinsey and Solve are trademarks of McKinsey & Company.
Frequently asked questions
Is there a free McKinsey Red Rock simulator?
CoachNed publishes a free, original Red Rock-style sampler at /mckinsey-red-rock: timed exhibit reading, calculator, four numeric questions, and a walkthrough. It is independent practice in the same skill family, not the live McKinsey assessment and not a replica of McKinsey's item bank.
How should I practice for Red Rock if McKinsey says no prep is needed?
McKinsey's public guidance says Solve does not require preparation and that tutorials teach the interface. That is true for content knowledge. What still transfers is reducing interface friction and rehearsing the research-and-calculate loop under a clock. Practice the skills; do not chase leaked scenarios.
What does a good Red Rock simulator include?
Timed reading of an objective and exhibits, forced evidence selection, calculation with units, and a written or spoken so-what. Bonus if it is forward-only under a shared clock and gives a review that diagnoses the first broken link, not only the final digit.
What is a good Red Rock score?
McKinsey does not publish a public Red Rock threshold. Score your practice on process (objective, labeled inputs, equation-before-calculator, so-what) and on answer accuracy separately. Treat any "pass mark" on a third-party tool as that tool's own bar, not McKinsey's.
Can a simulator guarantee I pass Solve?
No. Solve performance is considered with the rest of your application, per McKinsey's own Solve page. A simulator can make your reasoning more consistent. It cannot replace the instructions in your invitation or the integrity rules that require you to work alone.
Sources
- Solve, McKinsey’s assessment game — official description of Solve, no-prep framing, and integrity summary. Checked 2026-09-24.
- Interviewing at McKinsey — places Solve in the consulting recruiting process. Checked 2026-09-24.
- Assessment integrity expectations — alone, one screen, no AI, notes, or recordings. Checked 2026-09-24.
- How to prepare (McKinsey Finland) — regional note that Solve is approximately 60 minutes online. Checked 2026-09-24.
- Solve FAQ (August 2025 PDF) — invitation sets length; untimed tutorials; PC/Mac; performance considered with the rest of the application. Checked 2026-09-24.