Choosing the Right Chart for Your Data

The Chart Selection Framework

Selecting the right chart type starts with understanding what relationship in the data you want to communicate. There are five core data relationships, and each maps to a specific set of chart types.

This lesson goes deeper on each relationship type with practical guidance on when to use each chart and when to avoid common alternatives.

1. Change Over Time

Time series data needs to communicate trend, rate of change, and seasonal patterns.

Line Chart

Best for: Continuous data with many time points. When the slope (rate of change) matters as much as the values themselves.

Monthly revenue trend (12+ data points)
Daily website visitors
Hourly sensor readings

Design guidance:

• Start y-axis close to the minimum value (not necessarily zero) to show trend clearly
• Use solid lines for actual data, dashed lines for forecasts or targets
• Annotate notable peaks, troughs, or events directly on the chart
• Keep fewer than 5 lines on the same chart or it becomes unreadable

Column Chart for Time

Best for: Discrete time periods where individual values matter as much as the trend, and you have fewer than 15 time points.

Quarterly earnings per share
Annual headcount
Monthly budget vs actual (side by side)

Area Chart

Best for: Showing volume over time, particularly for stacked comparisons showing how composition changes over time.

Avoid area charts for more than 3-4 series -- they become unreadable when stacked.

2. Comparison Across Categories

Horizontal Bar Chart

Best for: Comparing a single metric across many categories (7+), especially when category labels are long.

Always sort the bars. An unsorted bar chart is rarely useful. Sort by value descending (or ascending for a top-to-bottom reading order) unless there is a natural ordering (e.g., age bands, satisfaction levels).

Revenue by country (top 15)
Product ratings from highest to lowest
Time spent on each task type

Column Chart (Vertical Bar)

Best for: Fewer categories (under 7) or when categories represent time periods.

Grouped Bar Chart

Best for: Comparing two or three metrics across the same categories.

Limitation: Becomes crowded with more than 3 groups or more than 6 categories. Consider small multiples instead.

Dot Plot (Cleveland Dot Chart)

Best for: Comparing many categories with a focus on ranking or precise values rather than magnitude. Less visual noise than bars.

3. Composition and Part-to-Whole

Donut Chart

Best for: Showing proportional breakdown of 3-5 categories where the total matters (e.g., revenue breakdown, market share).

Use a donut (not a pie). The hole in the middle can display the total value, making it more informative. Avoid when categories are numerous (use a bar chart instead) or when exact values matter (use a stacked bar or table).

Stacked Bar Chart

Best for: Composition across multiple groups (e.g., revenue breakdown by category, per quarter). Can show how composition changes over time.

Challenge: Segments other than the first (at the base) are hard to compare because they do not share a common baseline. Consider using 100% stacked bars to focus on proportions rather than absolute values.

Treemap

Best for: Hierarchical composition where the hierarchy itself is meaningful (category > sub-category > product). Good for showing proportional breakdown at multiple levels.

4. Distribution

Histogram

Best for: Understanding the shape of a single continuous variable's distribution. Are values concentrated in the middle? Skewed right? Bimodal?

Design guidance: Choose bin width carefully. Too many bins create a jagged chart that obscures the distribution shape; too few lose detail. A rule of thumb for bin count: square root of the number of data points.

Box Plot

Best for: Comparing the distribution of a variable across multiple groups. Shows median, quartiles, and outliers compactly.

Order value distribution by customer segment
Salary distribution by department
Delivery time by carrier

Box plots require the audience to understand quartiles. For general business audiences, consider violin plots (show full distribution shape) or stripplots (show individual points) alongside the box.

Violin Plot

Combines box plot with kernel density estimate (shows the full distribution shape). More informative than a box plot but requires statistical literacy.

5. Relationships Between Variables

Scatter Plot

Best for: Showing the relationship between two continuous variables. Reveals correlation, clusters, and outliers.

Design guidance:

• Add a trend line (linear or LOESS) to quantify the relationship
• Colour-code points by a categorical variable to show group differences
• Use alpha (transparency) when many points overlap

Bubble Chart

Adds a third variable encoded as bubble size. Effective for up to 3 variables. Bubble area (not radius) should encode the value.

Heatmap

Best for: Correlation matrices and cross-tabulations where you want to show value intensity across a matrix.

Correlation between 10 variables (10x10 matrix)
Revenue by hour of day and day of week
Website clicks by page and device type

Small Multiples

When you need to compare patterns across many groups and a single chart becomes cluttered, use small multiples: the same chart type repeated for each group.

Revenue trend line chart, one per region (6 small panels)
Scatter plot of orders vs revenue, one per product category

Small multiples allow direct comparison of patterns across groups by maintaining the same axes across all panels.

Chart Anti-Patterns to Avoid

Pie chart with more than 5 segments: Angles are hard to compare. Use a ranked horizontal bar chart instead.

3D charts: Always misleading. 3D perspective distorts values. Never use.

Dual-axis charts: Two y-axes on the same chart create visual confusion. Use two separate charts or normalise the data.

Radar/spider charts: Misleading due to area distortion and non-intuitive axis arrangement. Use a bar chart or table instead.

100-bar chart without sorting: An unsorted comparison chart forces the viewer to do all the analytical work.

Key Takeaways

• Match chart type to the data relationship: time trends use line/column, category comparisons use bar, distributions use histogram/box, relationships use scatter, and composition uses donut/stacked bar.
• Always sort bar charts by value unless there is a natural order -- unsorted bar charts make comparison significantly harder.
• Use small multiples when a single chart becomes cluttered with multiple groups -- keep the same axes across all panels for fair comparison.
• Avoid 3D charts, dual-axis charts, and pie charts with more than 5 segments -- these mislead more than they inform.
• Donut charts are preferable to pie charts: the hole allows displaying the total, and the arc is easier to read than sector area.

Practice Exercise

For each of the following analytical questions, identify the most appropriate chart type and justify your choice:

1. How has monthly revenue changed over the last two years?
2. Which 15 sales representatives have the highest quarterly targets, and how close are they to their goal?
3. What proportion of revenue comes from each product category?
4. Is there a relationship between the number of items in an order and the order value?
5. How does the distribution of delivery times compare between three different shipping carriers?
6. How does the revenue contribution of each category change from Q1 to Q4?