How to read a chart or exhibit in a case interview
At some point in a case the interviewer hands you a chart or a table and waits. They are checking whether you can find what matters in a page of data, avoid its traps and say something useful, all while staying calm. This guide gives a routine you can run in about a minute and a half, the traps to look for and a worked example with invented figures. Formats vary, and some interviewers prefer you to talk while you read while others prefer silence at first, so follow their cues.
What is the routine?
Five steps.
- Orient. Read the title, the units (millions or billions, per customer or in total, pounds or euros, nominal or adjusted for inflation), the period and the source or definition.
- Check the axes and scale. Where does the value axis start? Are the intervals even? Is there a second axis? Are the segments in a stacked chart comparable?
- Find the pattern. What is the largest, the smallest, the fastest-moving, the outlier? What changed, and when?
- Calculate what matters. Work out one or two numbers that answer the case question, such as a share, a growth rate or a gap, rather than every number you can.
- Say the so-what and the next step. One sentence on what it means for the client's question, one thing the exhibit cannot tell you, and the next piece of data you would request.
Tell the interviewer what you are doing: "Let me take a moment to read this." That is generally better than silence.
What traps should I look for?
- Truncated axes. A bar chart that does not start at zero exaggerates differences, since bar height no longer matches value.
- Narrow line scales. Small moves look dramatic when the range is tight.
- Two axes. Lines may cross without any meaning.
- Share versus level. A segment's share can rise while its value falls, if the total falls faster. Always say which you are reading.
- Mixed units. Millions beside billions, or per-unit beside total.
- Base effects. A large percentage rise from a tiny starting figure may be small in absolute terms.
- Stacked charts. Only the bottom segment starts at zero, so read the others by subtracting their edges. A 100 percent stack shows share, not size.
- Correlation and cause. Two lines moving together do not show that one caused the other.
- Small samples and outliers. A single unusual point can drive a pattern.
How do I move from numbers to insight?
An exhibit is a claim made visually, so treat your reading as a claim you can defend. Compare segments, periods and cohorts, ask which part of the total drives the change, and then ask why. Your sentence should include a number and a judgement: not "Europe is growing" but "Europe supplied about half of the growth, so the client's results depend on it."
What does a worked example look like?
Case: Calder Appliances, a maker of kitchen equipment, asks where its growth is coming from. The interviewer gives this table of revenue by region, in millions of pounds, for two years.
| Region | Two years ago | This year |
|---|---|---|
| UK | 120 | 126 |
| Europe | 80 | 104 |
| Asia | 40 | 60 |
| Total | 240 | 290 |
Computed before writing:
- Total growth: 290 / 240 - 1 = 20.8% over two years, which is about 9.9% a year compounded.
- Growth by region: UK 5.0%, Europe 30.0%, Asia 50.0%.
- Contribution to the 50m increase: UK 6m (12%), Europe 24m (48%), Asia 20m (40%).
- Share of revenue: UK 50.0% falling to 43.4%, Europe 33.3% to 35.9%, Asia 16.7% to 20.7%.
What to say aloud: "This shows revenue by region in millions of pounds over two years. Total revenue grew from 240 to 290, about 21%, or roughly 10% a year. The UK is still the largest region but grew only 5%, so it contributed 6 of the 50 million increase. Europe grew 30% and supplied about half of the growth. Asia grew fastest, at 50%, but from the smallest base, so it supplied 40%. The table does not show profit or whether the growth came from price or volume, so my next question is how margins differ by region, because growth in a lower-margin region could mean profit grows more slowly than revenue."
Notice the structure: unit and period, headline total, the segment that matters most, a caution about base effects, what is missing and the next question. It takes under a minute and avoids reading out every cell.
How do I handle a chart I cannot fully read?
Say what you can see and what you cannot, and ask. "The y-axis does not appear to start at zero; may I assume it does not?" "Is this nominal or adjusted for inflation?" Asking is better than guessing silently. If the chart conflicts with your hypothesis, say so and revise it.
What if the exhibit contains a lot of data?
Do not try to read everything. Decide what question the client is asking and scan for the figures that answer it. Look at totals first, then the largest items, then outliers. Say what you are ignoring and why.
What mistakes should I avoid?
- Reading the numbers aloud without a conclusion.
- Calculating before checking the units.
- Stating a cause the data cannot support.
- Missing the most obvious pattern while computing an obscure one.
- Giving percentages without absolute numbers, or the reverse.
- Forgetting to link the exhibit back to the case question.
What to do this week
- Take five charts or tables from reports you read and write a one-sentence so-what for each.
- For each, list the unit, the axis start and one possible trap.
- Practise calculating a contribution to growth from a table, mentally.
- Write down three follow-up data requests you might make for each.
- Ask a friend to hand you an exhibit with no warning and time yourself.
About this guide: it is original to Charta Learning and has not yet been reviewed by an experienced practitioner. Names in examples are invented. All figures are invented for practice, and exhibits in real cases vary widely in format and difficulty.