Data Interpretation
Read the chart, then do the arithmetic. Most DI questions are just percentages, ratios, or differences hiding behind a bar or pie.
What you'll learn
- The four chart types you will actually see: bar, pie, line, table
- Read axis labels and units FIRST, then the legend, then the question
- Percentage growth is (new − old)/old, never (new − old)/new
- Spot the axis-scale trick where a chart starts at 100 instead of 0
Before you start
Picture a 2-mark DI question. You glance at the chart, your eyes find the tallest bar, and you start computing — only to realise the axis was in thousands of crores, not crores, and the question wanted a ratio, not a difference. Two marks gone.
Data Interpretation rewards a different reflex: read the chart before you read the question, and read the question before you touch the arithmetic. The arithmetic itself is almost always a percentage, a ratio, or a difference. It’s also the single most transferable GA skill: every analyst dashboard and quarterly review is a DI question in disguise, axis tricks included.
The four chart types you will actually see
- Bar chart — categories on one axis, heights on the other. Best for comparing discrete groups (quarterly sales, marks per subject).
- Pie chart — parts of a whole, summing to 100%. Best for proportions, not absolute values.
- Line chart — a value tracked over time. Best for trends and growth rates.
- Table — raw numbers, often the most information-dense. Skim the column headers first.
Whatever the format, the read-it-right routine is the same: axis labels and units first, then the legend, then the question, then arithmetic. Skipping any step is how marks leak.
A small bar chart, read end-to-end
What is the percentage growth in sales from Q1 to Q4?
Read it right.
- Axis + units. Y-axis is in lakhs of rupees, starts at 0.
- Legend. Single series — sales per quarter.
- Question. % growth from Q1 to Q4. So Q1 is the old value, Q4 the new.
- Arithmetic.
growth % = (Q4 − Q1) / Q1 × 100
= (200 − 120) / 120 × 100
= 80 / 120 × 100
= 66.67 %So 66.67%. Notice the denominator is the old value — that single choice is what most students get wrong under time pressure.
How GATE asks this
Almost always one 2-mark question per paper. The shape is a NAT or MCQ: a chart or small table appears, and you must compute a single percentage, ratio, or difference from it. The chart is rarely the difficulty — the difficulty is reading the question carefully (“growth”, “share”, “ratio”, “by how much more”) and matching it to the right arithmetic.
A second habit worth building: when a pie chart gives percentages and the question asks for absolute values, you need the total somewhere in the question text (or another chart). Pie shares alone never give you absolute numbers.
Quick check
Quick check
Practice this in an interview
All questionsPie charts work only when you have two to three parts whose proportions differ substantially and sum to a meaningful whole. Beyond that, humans compare angles and arc lengths poorly, making slices of similar size indistinguishable. A sorted bar chart almost always communicates the same information more accurately.
Match the chart to the relationship in the data: comparison across categories calls for bars, trends over continuous time call for lines, correlation between two numeric variables calls for a scatter plot, and distribution shape calls for a histogram or box plot. The question you are answering — not aesthetics — drives the choice.
Coined by Edward Tufte, the data-ink ratio is the proportion of ink (or pixels) in a chart that encodes actual data, divided by the total ink used. Maximizing it means removing every element — gridlines, borders, tick marks, legends, decorative shading — that does not carry information the viewer cannot infer from the remaining ink.
A bar chart displays counts or aggregates for distinct categories separated by gaps; a histogram displays the distribution of a single continuous variable by dividing it into adjacent bins with no gap; a KDE (kernel density estimate) plot is a smoothed, continuous approximation of the same distribution without requiring a bin-width choice.