Introduction
Choosing a chart type is not a matter of visual preference; each chart form is built to answer a specific kind of question about data, such as comparing categories, tracking a trend over time, showing composition, or revealing the relationship between two numeric variables, and picking the wrong form can obscure the very message the data is meant to convey.
Cricket analogy: A commentator wouldn't show a run-rate line graph to explain who scored the most sixes in a tournament; a bar chart comparing players side by side answers that instantly, the same principle behind matching a chart type to what relationship in the data you actually need to show.
Explanation
The three most common chart families map directly onto three common questions: bar charts answer 'how do these categories compare,' line charts answer 'how has this value changed over a continuous sequence such as time,' and scatter plots answer 'do these two numeric variables move together.'
Cricket analogy: Bar charts are the tool for comparing discrete categories side by side, like runs scored by each batsman in an innings; line charts are the tool for showing a trend over a continuous sequence, like a team's win percentage across seasons; and scatter plots reveal whether two numeric measurements relate, like strike rate against average.
Composition charts such as pie charts or stacked bars answer 'what share of the whole does each part represent,' but they degrade quickly once the number of categories grows large or the proportions between slices become close, since the human eye is poor at comparing angles or areas precisely.
Cricket analogy: A pie chart showing each bowler's share of total wickets works fine with four bowlers, but once a tournament's twenty bowlers are crammed into one pie, the slivers become impossible to compare, the same reason composition charts lose their usefulness once categories multiply or proportions get close.
Example
# Choosing a chart type based on the question, illustrative example data
import matplotlib.pyplot as plt
categories = ['A', 'B', 'C', 'D']
values = [42, 37, 55, 29]
# Comparison question -> bar chart, axis starts at zero
plt.bar(categories, values)
plt.ylim(0, max(values) * 1.2)
plt.show()Analysis
Before trusting any comparison chart, it is worth checking whether its axis has been truncated or manipulated, because a bar chart with an axis that does not start at zero can visually exaggerate small differences into what looks like a dominant gap, which is one of the most common ways a technically accurate chart still misleads a reader.
Cricket analogy: Checking a chart for honesty means looking at whether the run-total bar axis starts at zero or is truncated partway up, since a bar chart with a truncated axis can make a small lead look like a dominant one, exactly the kind of distortion you must check for in any comparison chart before trusting it.
Key Takeaways
- Match the chart type to the question: comparison, trend, composition, or relationship.
- Bar charts compare discrete categories; line charts show trends over a continuous sequence; scatter plots reveal relationships between two numeric variables.
- Composition charts like pie charts work best with few categories and clearly distinct proportions.
- Always check whether a comparison chart's axis starts at zero, since truncation can exaggerate small differences.
- The wrong chart type can obscure or distort the very message the underlying data is meant to convey.
Practice what you learned
1. Which chart type is best suited to comparing values across discrete categories?
2. Which chart type is best for showing how a value changes over a continuous sequence like time?
3. What problem commonly occurs when a pie chart has too many categories?
4. Which chart type is best suited to revealing whether two numeric variables are related?
5. Why should you check whether a bar chart's axis starts at zero?
Was this page helpful?
You May Also Like
What Is EDA
What exploratory data analysis is, why analysts use summary statistics and visuals to understand a dataset before modeling, and what issues it uncovers.
Correlation vs Causation
Why two variables moving together does not mean one causes the other, how confounding variables create misleading correlations, and how causation is tested.
What Is Tableau
What Tableau is, how its drag-and-drop model builds interactive dashboards from connected data sources, and how finished dashboards are shared with others.