"The best chart isn't the fanciest — it's the one that answers your question fastest."
Learning Objectives
By the end of this module, you will be able to:
- Match a question to the right chart type
- Use a simple decision guide
- Avoid picking charts by habit
1. Start With the Question
Before choosing a chart, name what you're trying to show. Your question points to the chart:
| If you want to show… | Use a… |
|---|---|
| Compare amounts | Bar chart |
| How values are spread | Histogram / density |
| Parts of a whole | Pie / stacked bar |
| Relationship between two things | Scatter plot |
| Change over time | Line chart |
| Patterns across categories | Heatmap |
Key idea: Pick the chart to fit the question, not the other way around. "I need a chart" isn't a plan; "I need to compare sales across regions" points straight to a bar chart.
Explain like I'm new: Think of chart types as different tools in a toolbox. You wouldn't grab a hammer just because it's shiny — you grab it because you have a nail. Likewise, you pick a line chart because you have a trend over time, not because line charts look nice.
Most questions in data work fall into a handful of shapes, and each shape has a natural chart. "How do these groups compare?" is a comparison. "How is this one thing spread out?" is a distribution. "How much does each piece contribute to the total?" is a composition or parts-of-a-whole question. "Do these two things move together?" is a relationship. Once you learn to hear which shape a question has, the chart almost picks itself — the table above is really just a translation guide from question-shape to chart.
2. A Quick Decision Guide
- Comparing categories? → bar chart
- Showing a trend over time? → line chart
- Showing a share of a total? → pie or stacked bar
- Exploring how two numbers relate? → scatter plot
- Seeing the spread of one number? → histogram
Notice that this guide runs on the shape of the question, not the shape of the data. The very same spreadsheet of daily sales can feed all five charts, depending on what you ask of it. That's the mental shift this module is trying to build: you don't start by staring at your columns wondering what to plot — you start by finishing the sentence "I want to show that…" and let the chart follow.
Concept: A chart is an answer, so it needs a question first. If you can't state your chart's question in one sentence, you're decorating rather than communicating — and the result will usually confuse your audience instead of guiding them.
Simple example: "Which month had the most website visits?" is a comparison → bar chart. "How did visits change across the year?" is a trend → line chart. Same data, different question, different chart.
Real-world use case: A store owner has one year of daily sales. Asking "are weekends busier than weekdays?" is a comparison — a bar chart of average sales per weekday answers it instantly. Asking "is business growing?" is a trend — a line chart across the months answers that. The data never changed; the question chose the chart.
3. Match Chart to Audience
An expert might read a dense scatter plot; a general audience needs something simpler. Always design for who's looking. A board meeting, a scientific paper, and a social-media post can all show the same finding — but each deserves a different level of detail and polish. Ask yourself how much time the viewer has and how much they already know.
The rule of thumb is that the less familiar your audience is with data, the simpler your chart should be. A room of fellow analysts can happily read a box plot or a small-multiples grid; a general public audience reads bars, lines, and simple pies far more reliably. When in doubt, err toward the chart your audience already sees every day. A slightly less precise chart that everyone understands beats a technically superior one that only you can read.
Common mistake: Reaching for a pie chart with a dozen slices. Once a pie has more than about five parts, the human eye can no longer judge which slice is bigger. A ranked bar chart almost always communicates "parts of a whole" more clearly when there are many categories.
Data scientist tip: When stuck, default to the bar or line chart. They're the clearest, most universally understood charts — you rarely go wrong with them, and almost everyone reads them correctly.
Key Takeaway: Choose charts by starting from your question: bar (compare amounts), histogram (spread), pie/stacked (parts of a whole), scatter (relationships), line (over time), heatmap (patterns). Match the chart to your audience too — and when unsure, the humble bar or line chart is almost always the safe, clear choice.
Further Learning
Adapted from Microsoft's Data Science for Beginners (MIT License).