Capital One Case Interview vs McKinsey: One Long Problem

The difference that matters most in a Capital One case interview vs McKinsey is how the questions relate to each other. In the Capital One business cases candidates I've worked with describe, each follow-up question built on the answer to the one before, so a number from early in the case could still be in play near the end. McKinsey's published Electro-Light sample case is laid out differently: a later question states the target share it works from instead of taking it from your earlier answer. If your case practice comes from consulting casebooks, that difference should change how you practice.

The Capital One Power Day post makes this point in a single paragraph. This is the long version: what McKinsey publishes about its format, what a connected Capital One case looks like, and what it does to the cost of a mistake.

Capital One case interview vs McKinsey: what McKinsey publishes

McKinsey's careers site describes its case interview, which it calls the Problem-Solving Interview, as a business case McKinsey presents to evaluate your analytical thinking and your approach to solving complex problems. Nothing in that description would be out of place in a Capital One case.

The Electro-Light sample case McKinsey publishes on its careers site is laid out as a client scenario followed by numbered questions. In the Electro-Light sample, Question 2 asks what share of the electrolyte drink market Electro-Light would need to capture to break even, based on the target price and upfront fixed costs. Question 3 then states a 12.5 percent target share of the electrolyte-drinks market in its own wording and asks what would be needed to reach it. Whatever you worked out in Question 2, you start Question 3 holding the number it needs. The same Electro-Light sample's hints for Question 2 say calculators are not allowed and suggest writing calculations out on paper, while one candidate was told on their Capital One recruiter call that a physical calculator or a spreadsheet was fine, as the Capital One case interview calculator post covers.

That is a reading of McKinsey's published sample, not a claim about every live McKinsey interview. If your practice cases hand you each question's numbers the same way, the rest of this post is about the habit that builds.

What a connected Capital One case looks like

A connected Capital One case is one where a later question takes something from earlier in the case as its input. As a coach's way of sorting it, three things can carry forward: numbers from the prompt, a figure you calculated, and a decision you recommended. Candidates I've worked with describe numbers from the Power Day prompt coming back later in the case, and each follow-up building on the answer before it. One candidate saw the same thing in their own Capital One cases.

Carried-forward numbers from the prompt are the easy kind to handle, and the Power Day post covers the fix: write down every number as the interviewer reads the prompt. The other two kinds are harder, because they don't exist until you produce them.

Losing an earlier number costs real time even when it was handed to you. In a mock case I ran, a candidate I worked with had lost track of the total customer count shared earlier in the case. When a later question asked for total revenue, they kept asking how they could calculate it without a customer count, and even after I told them the number was in the material I'd shared earlier, they couldn't find it. Your notes should let you put your finger on any earlier number in seconds.

A calculated figure that carries forward means your arithmetic in one question becomes the given in the next. Prepare as if a Capital One question will use your number without restating it, and you should have that number written down, labeled, where you can find it again.

A recommended decision that carries forward is the kind to watch for if your prep came from consulting casebooks. In one candidate's Power Day case, the first question asked for options. The next question came back with two options, one of them the candidate's own, and asked which to pick. The question after that treated one of the two as decided and asked how to carry it out. None of those questions restated what the one before had settled. If you hedge when a Capital One question asks you to choose, the next question has nothing to build on, so you should commit to a recommendation and say what it rests on.

A decision that carries forward can also be handed to you. In a mock case I ran with the same chain of questions (options, then a pick between two, then how to carry it out), a candidate I worked with picked one of the two, and, playing the interviewer, I said leadership wanted the other one and moved the case on to how to carry it out. That was my call in a mock, not a claim about what Capital One interviewers do. If an interviewer does set the direction with a line like that, you should accept it and answer the next question on that basis rather than re-argue your pick.

A made-up case to show the chain

This case is invented for illustration and is not from any Capital One interview: a pet-supply retailer is weighing a monthly pet-food subscription. It has 200,000 loyalty members, expects 10 percent of them to subscribe, and earns $6 of contribution per subscriber per month. Treat it as pattern practice, not a preview of a real case.

In the made-up pet-supply case, the second question asks for monthly contribution: 200,000 members times 10 percent is 20,000 subscribers, and 20,000 times $6 is $120,000 a month.

In the made-up pet-supply case, the third question adds a new fact, $1.8 million to set up delivery, and asks how long the subscription takes to pay that back. The interviewer doesn't restate the $120,000. It's yours from the question before. $1.8 million divided by $120,000 is 15 months.

In the made-up pet-supply case, the fourth question says leadership wants payback inside a year and asks what would have to change. That question can't be answered without both earlier numbers: you need the 15 months to know how far off you are, and the $120,000 to know how much more contribution a 12-month payback takes.

Now run the same made-up pet-supply case with one slip. At the second question, 10 percent of 200,000 gets keyed as 2,000 subscribers instead of 20,000. Monthly contribution comes out at $12,000, payback comes out at 150 months, and the fourth question starts from a subscription that looks hopeless. One typo early in the case has turned into a recommendation to kill the product by the end.

Compare the made-up chain with McKinsey's published Electro-Light sample, where Question 3 states the 12.5 percent share it works from. In that layout, a wrong number in Question 2 doesn't reach Question 3, because Question 3 supplies its own. In the made-up chain, the same slip costs you everything downstream.

Why a connected case changes what a mistake costs

In a connected Capital One case, a wrong number doesn't stay in the question where you made it. Every question that uses it inherits the error, and the case stops making sense from that point on. On one candidate's Power Day, the interviewer flagged a math error rather than letting it carry forward. A case built as a chain stops working if one link is wrong, so the flag keeps the case running.

The Capital One case interview calculator post has one candidate's version of that exchange, including what made the fix quick: every intermediate value was already on paper. You should read a flag in a connected case as the case being repaired, and go straight to your notes to find the bad link.

A pause is also cheaper than a wrong link. The Power Day post covers the room candidates I've worked with were given to pause before answering. In a connected case, the ten seconds you take to check which earlier number a question depends on is ten seconds well spent.

What transfers from McKinsey prep, and what doesn't

Most of what you build practicing McKinsey-style cases carries over to a Capital One case interview. Laying out a structured approach to the first question, numbering your points so the interviewer can follow, reading an exhibit before talking about it, and setting up a formula before touching the numbers all work the same way.

Aim for close to MECE, not strict MECE; time spent perfecting non-overlapping buckets is better spent keeping track of what the case has already established.

The habit that doesn't transfer from consulting casebook practice is the reset. Drilling questions one at a time, from a book where each question hands you its own numbers, trains you to start every question clean. In a connected Capital One case, starting clean means dropping the one thing the next question needs.

How to practice for a connected case

Run cases start to finish, never as loose questions. If your practice material is a set of separate questions, chain them yourself: before each new question, write down which earlier output it would use.

Keep one page for what the case has established. Prompt numbers, your calculated figures, and the decisions you recommended, each labeled. The calculator post explains why that page matters more when you're working on a one-line display.

Commit when a question asks for a recommendation. A hedged answer leaves the next question with nothing to stand on. In practice, have your partner follow a pick-one question with one that assumes a choice has been made, sometimes the choice you didn't make, and answer it without re-arguing.

Practice recovery on purpose. Have a practice partner tell you, two questions later, that one of your earlier numbers was wrong. Then re-run the chain from your notes, not from memory.

Full case walkthroughs from prompt to final recommendation, showing how each answer carries into the next question, are what the paid guide covers. That part isn't written yet.

If you've been practicing from consulting casebooks, you already have most of what a Capital One case asks for. The one change is to stop treating each question as a fresh start.