Published 9 min read
How to Read Charts and Exhibits in Case Interviews
Case interview charts test whether you can pull the single insight that matters out of a graph in under a minute, not whether you can read every number on the page. When an interviewer hands you an exhibit, scan the title and axis labels first, form a hypothesis about what story the data is telling, check that hypothesis against the actual numbers, then state the so-what out loud before you touch anything else. Most candidates do this backwards. They dive into the numbers before they understand what is even being measured, and they run out of time before they say anything useful.
I conducted well over a hundred interviews at McKinsey, and the chart moment was one of the most reliable filters I had. Not because the math is hard. The math behind a case exhibit is usually simple, often just a comparison or a trend line. The filter is whether a candidate treats the chart as a puzzle to solve calmly or as a threat to react to. Panic looks like silence followed by a recitation of numbers. Competence looks like a fast, structured pass followed by one clear sentence.
Why interviewers hand you a chart at all
An exhibit exists to test something a verbal case cannot test well: can you extract a business insight from data under time pressure, the way you would in an actual client meeting. Consultants get handed messy spreadsheets and slide decks constantly, and a partner does not want to watch you stare at a table for three minutes before saying anything. The exhibit is a compressed version of that real skill.
It also tests something subtler. A chart usually has more information on it than you need. Interviewers build in a decoy axis, a distracting third data series, or a footnote that changes the read entirely. Candidates who grab the first number they see and start talking are showing the interviewer that they will do the same thing with a real client deck: mention a stray fact instead of the takeaway that actually matters.
The four-step method for reading any exhibit
Every chart you get in a case, no matter how it looks, responds to the same four-step read. Do these in order, every time, and you will rarely be caught off guard.
First, read the title and both axes before you look at a single data point. The title tells you what question the chart is answering. The axes tell you the units and the scale, and scale mistakes are the single most common way candidates embarrass themselves, mistaking a chart in thousands of units for one in millions.
Second, take fifteen to twenty seconds of silence to scan the shape of the data. Is it going up, down, flat, or diverging between two series. You are not calculating anything yet. You are forming a first impression the way you would glance at a stock chart.
Third, form a hypothesis about the takeaway and then check it against two or three specific numbers. This is the step people skip. They see a downward line and announce "revenue is declining" without checking whether the decline is driven by one outlier segment or a broad trend across the board.
Fourth, say the so-what out loud in one sentence, tied back to the objective of the case. Not "sales went from 40 to 25." Instead, "specialty coatings revenue fell 37 percent while the rest of the portfolio held flat, which points to a segment-specific problem rather than a company-wide one." That sentence is the entire reason the exhibit exists.
Matching your read to the type of chart
Different chart formats hide different traps, and knowing the traps in advance means you spend your fifteen seconds of scanning looking in the right place instead of everywhere.
| Chart type | What it usually tests | First thing to check |
|---|---|---|
| Bar chart comparing categories | Whether you notice the biggest or smallest bar and explain why it matters | The scale of the y-axis and whether bars are stacked or grouped |
| Line chart over time | Whether you separate a real trend from a one-time blip | Whether one data point is an outlier year (a recession, a launch, a recall) |
| Two-axis chart (dual y-axis) | Whether you catch that the two lines use different scales | Read both axis labels before comparing the lines visually |
| Pie or share-of-total chart | Whether you compare relative size, not raw counts | Whether the slices actually sum to 100, and whether "other" is hiding the real story |
| Table of raw numbers | Whether you can do fast mental math instead of describing every cell | Row and column totals, and any footnote defining the units |
| Waterfall or bridge chart | Whether you can trace what adds and what subtracts from a starting number | The sign of each bar, since a waterfall bar can be a negative you misread as positive |
Waterfall charts and dual-axis charts are the two that trip up the most candidates, because both are built to be misread quickly if you are moving fast. Slow down by exactly five seconds on those two chart types and you will beat most of the field.
The mistakes that give you away
A few habits show up over and over in weak chart reads, and every one of them is fixable with practice rather than raw ability.
Reading every number on the page out loud instead of the two or three that matter. The interviewer already has the chart in front of them. Reciting it back wastes time and signals that you cannot tell relevant from irrelevant.
Ignoring the units or the footnote. A chart labeled in thousands of units that gets read as millions changes every conclusion that follows it, and interviewers plant this specifically to see who checks.
Jumping to a recommendation before finishing the read. Charts often arrive mid-case, and the temptation is to leap straight to "so we should cut costs" before you have actually confirmed what the data shows. Confirm the fact first, then connect it to the objective.
Going silent for too long. Fifteen to twenty seconds of quiet scanning is normal and expected. A full minute of silence in front of a chart reads as being stuck, even if you are actually thinking well. Narrate lightly if you need more time: "give me a second to make sure I'm reading this axis correctly."
Missing that a chart contradicts your structure. If your issue tree assumed the problem was pricing and the exhibit shows pricing has been flat while volume collapsed, say that out loud immediately. Interviewers respect candidates who adjust in real time more than candidates who force the data to fit a structure built two minutes earlier. If your structure needs a refresher, the guide to building structure without memorizing frameworks covers exactly how to build a tree you can flex mid-case.
Multi-exhibit cases: keep a chart log
Interviewer-led cases sometimes hand you three or four exhibits across a single session, and candidates lose the thread of what each one showed by the time they reach the recommendation. Keep a running one-line log on your notes page as you go: exhibit number, the so-what in one sentence, nothing else. When you reach your recommendation, you flip back through three lines instead of trying to remember four charts from memory. This habit alone fixes a huge share of the weak recommendations I heard, where a candidate forgot the second exhibit even existed. For more on organizing a page that survives a forty-five minute case, see how to take notes in a case interview.
Candidate-led cases put more of this burden on you, since you are the one asking for the data rather than being handed it at a set moment. That changes when exhibits appear but not how you read them once they do. The difference between the two formats is worth understanding on its own, which is covered in interviewer-led versus candidate-led cases.
What to actually say out loud
The words matter as much as the read. A strong chart response has three parts: the fact, the comparison, and the implication. "Revenue per store dropped from 2.1 million to 1.4 million over three years" is the fact. "That's a third of stores' revenue disappearing while total store count barely changed" is the comparison. "Which suggests same-store performance, not expansion, is the real problem" is the implication that ties back to the case objective. Say all three, in that order, in under twenty seconds, and move on. Do not linger to prove you noticed more detail than the interviewer expected. One sharp read beats three mediocre observations.
The bottom line
A case interview chart rewards a fast, structured read over a thorough one: title and axes first, a hypothesis you check against two or three numbers, then one sentence tying it back to the objective. Practice the four-step method on old case books until it is automatic, because the chart moment moves fast and there is no time to invent a process while the clock is running. Once you can read an exhibit cleanly, the harder skill is turning that read into a recommendation the client can actually act on, which is where closing a case with a strong recommendation picks up.
Go deeper
Chart reading is one piece of a bigger system. The CaseMap method inside the course walks through the full sequence, from opening the case to reading an exhibit under time pressure to closing with a recommendation that holds up under follow-up questions.
Get the complete Cut to the Case course →
One payment, lifetime access, 30-day guarantee.
Frequently Asked Questions
How much time should I spend reading a chart in a case interview?
Fifteen to twenty seconds of silent scanning before you say anything, then a full read of no more than twenty seconds out loud. Longer than that and the interviewer starts to wonder if you are stuck rather than thinking.
What is the most common mistake candidates make with case interview charts?
Reading every number on the page instead of the two or three that matter, and skipping the axis labels and units before jumping to conclusions. Both habits signal that you cannot separate relevant data from noise, which is exactly what the exhibit is testing.
Do I need to do math when I look at a chart in a case interview?
Sometimes, but usually only enough to check a hypothesis, like confirming a percentage change or comparing two totals. The exhibit is testing insight extraction more than calculation, so save the heavier math for when the interviewer explicitly asks you to compute something.
What if the chart contradicts the structure I already built?
Say so immediately and adjust. Interviewers plant contradicting exhibits on purpose to see whether you defend a structure that no longer fits the data or update your thinking in real time. Updating in real time is the stronger signal.
How do I keep track of multiple exhibits in one case?
Keep a one-line log on your notes page for each exhibit: the number and the single so-what, nothing more. When you reach your recommendation you can scan three or four short lines instead of trying to recall entire charts from memory.
Are dual-axis charts really that common in case interviews?
They come up often enough to be worth specific practice, because the two lines using different scales is an easy trap to miss at normal case speed. Reading both axis labels before comparing the lines visually avoids the mistake almost every time.