Choosing the Right Chart for the Question: Interview-Ready Visualisation Framework
Why do two dashboards built from the same data tell completely different stories? Because the chart is not a decoration - it is the argument. Pick a pie chart when the question is about trend, or a line chart when the question is about ranking, and the viewer works harder while trusting you less.
- Start with the question, not the chart. Ask: am I comparing, tracking over time, showing composition, distribution, relationship or geography?
- Match chart to data shape. Categories usually need bars; time series need lines; part-to-whole needs stacked bars or 100% bars; relationships need scatter plots.
- Use bar charts for ranked comparisons. They are the safest business chart because length is easy to compare accurately.
- Use line charts only when order matters over time. A line implies continuity, so avoid it for unrelated categories.
- Avoid pies when there are many slices. Use them only for simple part-to-whole stories with 2-4 clearly different parts.
- One chart should answer one primary question. If it tries to show trend, ranking, composition and correlation together, split it.
- The best chart reduces thinking time. The viewer should know what changed, where the problem is, and what to do next.
The big picture is simple: a good chart is selected backwards from the decision the viewer must make. The sequence is question - data shape - chart family - design check. If you skip the first step, you end up with attractive but unhelpful visuals.
Core Explanation: The Chart Picker Framework
A chart is a visual encoding of data using position, length, area, colour or shape. The right chart is the one whose encoding makes the required comparison easiest.
Business users usually ask six types of questions. Each question type has a natural chart family.
The most reliable mental shortcut is this: comparison uses length, trend uses position over time, relationship uses points, distribution uses bins, and geography uses location.
The Six Chart Choices You Must Be Able to Defend
1. Bar chart - for comparison
Use a bar chart when the business question is “which is bigger, smaller, better or worse?” Sort bars when ranking matters. Use horizontal bars when category names are long.
Example: comparing revenue by Indian state is usually clearer as a sorted horizontal bar than as a map, unless the spatial pattern itself is the point.
2. Line chart - for trend over time
Use a line chart when the x-axis is ordered time: days, months, quarters or years. It shows direction, seasonality, volatility and turning points.
Rule: if the x-axis categories can be rearranged without changing meaning, do not use a line chart.
3. Stacked or 100% stacked bar - for composition
Use stacked bars to show how parts contribute to a total. Use 100% stacked bars when the question is about share, not absolute size.
Example: if a payments company wants to compare UPI, cards and net banking mix across quarters, a 100% stacked bar shows mix shift better than three separate pies.
4. Histogram or box plot - for distribution
Use a histogram when you need to show how values are spread. Use a box plot when comparing distributions across groups.
Example: average delivery time hides whether most orders are on time but a few are extremely delayed. A histogram reveals that operating risk.
5. Scatter plot - for relationship
Use a scatter plot when asking whether two numerical variables move together: discount vs margin, ad spend vs leads, income vs default risk.
Interview caution: a scatter plot can suggest correlation, but it does not prove causation. You still need business logic, controls or experimentation.
6. Map - for geography
Use a map only when geography is part of the answer. If the task is simply to rank cities by sales, a sorted bar is often faster and more precise.
Stephen Few defines data visualization as “the graphical display of abstract information for two purposes: sense-making and communication.”
A Small Worked Example: Same Data, Different Question, Different Chart
Assume a hypothetical FMCG brand has quarterly sales by channel:
If the question is “Which channel is larger in Q4?”, use a simple bar chart. If the question is “How did e-commerce catch up over the year?”, use a two-line chart. If the question is “How did channel mix change?”, use a 100% stacked bar.
The data did not change. The question changed. That is the whole discipline.
Definitions You Should Say Clearly
- Chart: a visual representation of data designed to reveal comparison, trend, composition, distribution, relationship or location.
- Encoding: the visual method used to represent data, such as length, position, colour, area or shape.
- Dimension: a categorical field used to slice data, such as region, channel, gender or product.
- Measure: a numerical field that can be aggregated, such as sales, margin, orders or conversion rate.
- Data-ink ratio: Edward Tufte’s term for the share of ink used to show data rather than decoration.
Case Study: Razorpay Payments Insights and Chart Choice
Razorpay turns complex Indian digital payments data into business-readable reports by matching each question to a chart form.
Situation: Indian digital payments data is multi-dimensional. A payments platform must explain time trends, payment-method mix, regional patterns, merchant categories and festival spikes without overwhelming readers. A spreadsheet can hold that information, but it cannot make the pattern instantly visible.
The move: Razorpay’s public Payments Insights style of reporting uses the logic every analyst should follow: use trend charts for time movement, comparison charts for payment methods or sectors, and geography only when regional variation is the story. The primary driver is question-led visualisation. Supporting drivers include clean aggregation of transaction categories, consistent time periods, business-friendly labels and a narrative around what changed.
The lesson: A payments report is not better because it has more charts. It is better when each chart answers a distinct question - what grew, where it grew, which method gained share, and which merchant categories moved.
So what: Razorpay is a strong Indian example because payment data is naturally messy - time, geography, categories and transaction values all coexist. The disciplined analyst separates these questions instead of forcing them into one “dashboard-looking” visual.
How AI Changes Choosing the Right Chart for the Question
AI does not remove chart judgment. It makes chart generation faster, which makes judgment more important.
- Natural-language BI is becoming normal. Tools can now generate charts from prompts like “show monthly repeat purchase trend by city.” The risk is that the tool may choose a default chart that looks plausible but answers the wrong question.
- AI chart recommendation can speed up exploration. Auto-insights features can suggest outliers, trends and relationships. Use them for discovery, then manually check whether the recommended chart matches the audience’s decision.
- LLMs help critique charts. You can ask an AI assistant to identify whether the chart shows comparison, trend, composition, distribution or relationship - and whether a simpler chart would work better.
Paste your dataset column names and the business question into ChatGPT: “I have columns: month, city, channel, orders, revenue, discount, margin. For the question ‘Which channel is driving margin decline?’, suggest 3 chart options, explain the best one, and name one misleading chart to avoid.” Then verify the answer using the chart picker framework above.
Interview Relevance
“You are given monthly sales, profit margin, region, channel and customer segment data. How would you decide which chart to use in a management dashboard?”
In interviews, say the chart choice and the reason together: “I would use a sorted horizontal bar because the question is ranking regions, and length is easier to compare than area or angle.” That one sentence sounds analytical.
Common Mistake
The biggest mistake is choosing a chart because it looks sophisticated instead of because it answers the question. It costs candidates because it signals weak business thinking, even if the chart is technically correct. Fix: always state the question first, then the chart, then why that chart makes the comparison easiest.
What to Revise Next
Once you can choose the right chart, revise how charts can mislead and how design choices affect readability. The natural next topics are: