McKinsey Solve in 48 Hours: The Honest Last-Minute Plan

Written by Tom Prescott, founder of SolvePrep. Last updated: August 30, 2026.

If your test is tomorrow or the day after, you're in the majority — 70% of candidates in our data don't start practicing until the final 48 hours, and the median candidate's first practice run happens one day before the real thing our study. So no lecture. You're here now, the clock is real, and the useful question is what still moves your score in the time you have. Some things genuinely do. Some popular ones genuinely don't. Here's the split.

What still works with 48 hours

Format familiarity is the highest-value thing you can still buy, because it's the cheapest difficulty to remove. Most of a first-timer's losses aren't intelligence losses — they're "I didn't know the mechanics and the clock ate me" losses. Two days is enough to make test day your second or third time seeing each game instead of your first, and in our data users score 20% higher on their best run than their first. What two days won't fix: the deep process-consistency layer that builds across many runs. Accept that trade and optimize what's optimizable.

The plan, hour by honest hour

Today, session one (90–120 min): one full timed run of each game. All three — Sea Wolf, Red Rock Study, Sustainable Future Lab — in the real format, real clocks start a full run. Don't pause, don't retry mid-game. The point of this session is diagnosis, not scoring: you're finding out which game is your problem.
Today, session two (60–90 min): attack the weakest game. Your score report tells you where the points went. Two more full runs of your worst game beat four scattered ones — the Red Rock Study shows the steepest improvement curve in our data (median +20.3 points from first run to repeat), so if that's your weak one, the reps pay unusually well.
Tomorrow morning (if you have the second day): one run of each, again. You're consolidating mechanics, not chasing a personal best. If a run goes badly, fine — better here than there.
The night before: stop. This is the step people skip, and it's the one with actual evidence behind it everywhere in cognitive performance: the Solve is a processing-speed and judgment test, and sleep-deprived judgment is measurably worse. A midnight cram run buys you familiarity you already have and costs you tomorrow's clock speed. Close the laptop.
Test-day setup (15 min, morning of): Desktop or laptop, never a tablet. External mouse if you have one — several mechanics are drag-heavy. Wired or strong connection, charger in, notifications off, a closed door, water, and the bathroom trip before you launch. Read every in-game instruction screen fully; the format rewards the 40 seconds of reading it punishes skipping.

What to skip entirely

Video walkthroughs instead of runs (watching is not doing — the clock only respects hands-on reps), memorizing "optimal strategies" from forums (several are outdated to retired game versions — the Ecosystem game is gone what replaced it), and any guide promising the answers. The Solve generates scenarios; there's nothing to memorize, only mechanics to have already operated.

The one decision left

You'll take the test either way. The only open question is whether tomorrow is your first time seeing these games or your fourth. One of those walks in knowing exactly what's coming — and if you want every run between now and then plus the breakdown of where your points went, that's Elite, one-time, with the 7-day guarantee unlock unlimited runs. Either way: run each game at least once today. That single decision outperforms everything else on this page.