What score do you need to pass McKinsey Solve?

This is the most common question candidates ask before sitting Solve, and it has an unsatisfying answer: there is no score to hit. Not a percentage, not a number of correct answers, not a threshold McKinsey will confirm. Understanding why is more useful than the fake precision most sites offer, because it changes what you should actually optimize for.

Solve is ranked, not marked

Your product and process results are combined into a single figure, and that figure is then positioned against everyone else in your applicant pool. The output is commonly described as a decile rank.^[1] Scores are benchmarked within the candidate pool rather than against a fixed cutoff.^[2]

That single design choice explains most of the confusion online. A ranked assessment has no stable passing mark by definition. The same performance can clear in a weaker cohort and miss in a stronger one, which is why candidates comparing notes about what they answered correctly often reach contradictory conclusions.

Where the bar sits, as best anyone can tell

McKinsey does not publish cutoffs, so everything below is inference from candidate reports and coaching data rather than firm disclosure. Treat it as a range, not a target.

Reported pass rates for Solve cluster around 20 to 30 percent, which means roughly 70 to 80 percent of candidates are eliminated at this stage.^[3] Some estimates put it nearer 20 percent, or about one in five.^[4] Working backwards from that, the commonly cited rule of thumb is that landing around the 80th percentile, meaning the top fifth of your cohort, is usually enough.^[5]

The boundary is not uniform. The effective threshold is reported to be tougher at competitive, high-volume offices such as New York and London than at smaller markets.^[3] Applying to a large office means being ranked against a deeper field for the same number of interview slots.

How Solve compares to the test it replaced

Useful context for anyone reading older advice: the old Problem Solving Test is generally described as having had a pass rate around 30 percent, against roughly 20 percent for Solve.^[4] The screen got harder, not easier, and prep material written for the PST is aimed at a different assessment.

Why a decile is not the whole decision

Solve is one input into a review, not a gate that opens on its own. McKinsey states on its assessment page that your performance is considered together with the rest of your application and any other assessments you take.^[6]

In practice, that means the same decile produces different outcomes for different candidates. A mid-range result attached to a strong resume, relevant experience or a referral can still advance, while a better result attached to a thinner profile may not.^[1] Coaching write-ups routinely describe candidates at the eighth decile going both ways depending on the rest of the file.

The practical read: if your profile is unusual for the office you are applying to, assume you need to be comfortably inside the pass zone rather than on its edge.

Does everyone even take it?

Mostly, but this is genuinely contested and worth checking against your own invitation rather than a blog. Most accounts describe Solve as near-universal across undergraduate Business Analyst, MBA Associate, advanced professional degree and experienced hire routes, with exceptions limited to very senior hires or offices piloting different processes.^[7] Some experienced-hire coverage notes that certain US offices have begun waiving the assessment for senior candidates.^[8] At least one long-running recruiting guide goes further and states that many US offices have phased out the requirement.^[9]

These accounts do not agree, and none of them is McKinsey. Your invitation email is the only authority on whether you are sitting it and which format you will get.

You will not be told your score

McKinsey does not release score reports, percentiles or feedback by default. You receive a pass or fail notification, typically one to two weeks after you finish.^[10] A pass means your combined product and process result met the threshold for your pool.

One detail that rarely surfaces: some candidates have reportedly been able to request their decile or raw score from recruiting, but only by asking proactively.^[10] It is not offered, it is not guaranteed, and it is worth asking politely rather than assuming. If you are rejected and considering reapplying, knowing whether you missed narrowly or badly is genuinely useful, and the standard wait before reapplying is commonly cited as twelve months or longer.^[10][11]

So what should you actually aim for?

Since you cannot target a number, target the behaviors that move a ranked score. All of these are things you control on the day.

• Budget minutes per task before you start, and abandon anything that overruns, even unfinished.
• Orient before acting. A short read of the objective and constraints is recorded as deliberate behavior, not wasted time.
• Commit to decisions. Revising on new information is fine. Churning without it is logged.
• Get the arithmetic automatic. Percentage change, weighted averages and growth rates should not cost you thinking time.
• Remove interface novelty. Meeting an unfamiliar layout under a hard clock is the most avoidable way to lose rank.

None of this is a hack. In a ranked assessment where most candidates are capable, the margin usually comes from the people who did not waste minutes learning the interface while the clock ran.

Frequently asked questions