Case interview math: explain before you calculate
Interviewers grade the setup more than the arithmetic, and they still expect the arithmetic to be right. The three-step loop that separates offers from dings in case math, plus a five-minute daily drill.
- Interviewers grade the setup more than the arithmetic, and they still expect the arithmetic to be right.
- Three steps: say what you are about to do and what you expect it to mean, get a short buy-in, then compute.
- The buy-in is what lets the interviewer redirect you while redirecting is still cheap.
- Pick one zero system and never mix. Mixing them is the main source of magnitude errors.
On this page
Here is the uncomfortable truth about case math: interviewers care more about how you set up a calculation than about the multiplication itself. They still expect the arithmetic to come out right. But the habit that separates offers from dings, the one almost every serious prep guide converges on, is this. State your approach out loud, get buy-in, and only then calculate.
Candidates who dive straight into arithmetic lose twice. The interviewer cannot follow the thinking, so even a correct number earns partial credit. And if the setup is wrong, they find out two minutes too late, after saying a number that cannot be unsaid.
The three-step loop
Every calculation in a case, from a quick percentage to a full market model, runs through the same loop.
Step 1: announce the approach. Before touching a number, say the equation in words. "To get annual revenue, I'll take daily customers, times average ticket, times 360 days, and then multiply by the store count." If you are on paper, write the formula and turn the page toward the interviewer. In a voice interview, the sentence does the same job. It is the math twin of the headline-first structure presentation.
Step 2: get buy-in. One short question: "shall I run it?" or "am I missing anything?" This is not asking permission out of weakness. It gives the interviewer the chance to redirect you before you burn time, and interviewers routinely do exactly that. Cases also hand you more data than the question needs, the same way real clients do. Announcing your approach is how you find out which numbers matter before spending two minutes on ones that do not. Buy-in is your filter.
Step 3: compute out loud. Narrate each step as you do it. "Two hundred customers times a 6 euro ticket is 1,200 a day, times 360 is roughly 430 thousand a year." Talking through the arithmetic means a slip gets caught at the step where it happens, by you or by the interviewer, instead of surviving into your final answer.
Hear the difference a real conversation makes
Your first full voice interview is free. No sales call, just a case and a live interviewer.
What the difference sounds like
The case: a client runs 50 coffee kiosks and wants to know rough annual revenue.
The candidate who dives in:
The number happens to be defensible, and it still lands badly. The interviewer had to pull the reasoning out afterward, which reverses the entire dynamic.
The candidate who sets up first:
Same skill level, completely different impression. The second candidate also protected themselves. If 300 customers a day had been way off, the interviewer would have said so at "sound reasonable?", before any arithmetic happened.
Round out loud, and manage your zeros
Two mechanical habits carry most of the remaining weight.
Round aggressively and announce it. "I'll round 47 to 50 for a slightly generous estimate, okay?" Clean numbers keep your head free for the logic. Nobody has ever lost an offer for rounding transparently. Plenty have lost one grinding through 47 times 365 in their head.
Pick one zero system and never mix. Thousand, million, billion, or scientific notation, either works, but mixing them is the number one source of magnitude errors. A useful anchor: thousand times thousand is a million, million times thousand is a billion. And before any result leaves your mouth, run the laugh test. If your per-kiosk number implies more customers a day than the neighborhood has residents, a zero slipped somewhere.
When the number is wrong anyway
You will make an arithmetic error at some point. The grading is not about whether it happens, it is about the recovery. The candidates who survive errors do three things. They catch it themselves, usually because they sanity-checked against context. They fix it without drama: "wait, that's off by a factor of ten, it should be 4 million, let me redo that step." And they do not make the same error again ten minutes later. A caught error is nearly free. An uncaught one, repeated, is how math ends interviews.
The so-what is part of the math
A calculation is not finished when the number is out. It is finished when the number means something. "Revenue is about 21 million" is an answer to a school exam. "Revenue is about 21 million, and since the client claims 30 million in costs, this business is structurally losing money, which changes the question" is an answer to a case. Make the translation a reflex: every number gets one sentence connecting it back to the client's question.
Drill this
Case math is a fluency skill, and fluency comes from short daily reps, not weekend marathons.
The five-minute daily rep, for two weeks:
- Take one quantitative prompt per day. Market size of anything, break-even for a made-up shop, revenue of the cafe you are sitting in.
- Speak the setup first, out loud, as if an interviewer were across the table. Approach, buy-in question, then compute while narrating.
- Round out loud, keep zeros in one system, end with a so-what sentence.
- Time-box it to five minutes. Done is the goal, not perfect.
Out loud is non-negotiable. Silent math practice trains the wrong skill, because the interview version of the skill is verbal. After two weeks the setup sentence stops costing effort, which frees your attention for the actual case.
Frequently asked questions
Can I use pen and paper for calculations?
In almost every real interview, yes, and you should. Writing the formula down and turning it toward the interviewer is the classic version of explain-before-calculate. In a voice-only setting, the narration replaces the paper. Say what you are writing as you write it.
What if I make an arithmetic mistake?
Catch it, name it, fix it, move on. Interviewers grade recovery, not perfection. The damaging version is the confident wrong-magnitude answer that you never notice. Sanity-check every result against common sense before saying it.
How fast do I need to be?
Slower than you think. The most recent prep guides are explicit that quality beats speed, and a slow correct answer beats a fast wrong one. Speed comes on its own from daily reps. Rushing is how zeros get lost.
Which mental math is actually worth memorizing?
Percentage decomposition (15% of 400 is 40 plus 20, so 60), turning growth rates into doubling times (roughly 7 percent a year doubles in a decade), dividing by fractions instead of decimals (40 percent of a number is two fifths of it), and a handful of demographic anchors like the US population at roughly 330 million and about 2.5 people per household. That short list covers most case arithmetic. Everything beyond it is nice to have.
Practice this on your own casebook
Upload the casebook PDF you already study with and run its cases as live voice interviews with exhibits and 8-dimension feedback.