13 Chart Types and Their Use Cases

13.1 Why Chart Choice Matters

The chart type is the grammar of the visual sentence. The wrong choice obscures the very point the analyst is trying to make.

The same dataset can support a dozen different chart types. Most of them will obscure the message; one or two will reveal it. The discipline of effective visualisation therefore begins with chart selection — choosing the type that aligns the visual encoding with the question the reader is trying to answer.

The choice is not a matter of aesthetic taste. It is determined by two things: the question the reader needs answered and the type of data available to answer it. Every well-known chart type sits at the intersection of one or two questions and one or two data structures. The practitioner classics by Stephen Few (2012) and Claus O. Wilke (2019) organise the field around exactly this principle.

13.2 A Framework for Chart Selection

A short set of questions, asked before any chart is drawn, will narrow the choice to a handful of candidates:

What is the question? Is the reader comparing values, looking at composition, examining a distribution, exploring a relationship, tracking a trend, or finding a location?
How many variables? One, two, three, or many?
What kind of data? Categorical, ordinal, quantitative continuous, time, or geographical?
How many categories or points? A handful, a few dozen, or thousands?
Static or interactive? Will the reader hover, filter, drill down, or only look?
Who is the reader? A board member with one minute, an analyst with an hour, or a public audience with no analytical context?

The answers to these six questions usually leave only two or three sensible chart types in contention. The task is then to choose between those, not between every chart in the encyclopaedia.

13.3 The Seven Categories of Chart

flowchart TD
    Q["What is the question?"]
    Q --> C1["Comparison<br>How do values<br>differ across<br>categories?"]
    Q --> C2["Composition<br>What are the parts<br>of a whole?"]
    Q --> C3["Distribution<br>How are values<br>distributed?"]
    Q --> C4["Relationship<br>How do two or<br>more variables<br>relate?"]
    Q --> C5["Trend<br>How does a value<br>change over<br>time?"]
    Q --> C6["Ranking<br>What is the order<br>of magnitude?"]
    Q --> C7["Geographical<br>How does a value<br>vary across<br>locations?"]
    style Q fill:#e3f2fd,stroke:#1976D2
    style C1 fill:#e8f5e9,stroke:#388E3C
    style C2 fill:#fff8e1,stroke:#F9A825
    style C3 fill:#fff3e0,stroke:#EF6C00
    style C4 fill:#fce4ec,stroke:#AD1457
    style C5 fill:#ede7f6,stroke:#4527A0
    style C6 fill:#f3e5f5,stroke:#6A1B9A
    style C7 fill:#eceff1,stroke:#455A64

The Seven Categories at a Glance

Category	Question Answered	Typical Charts
Comparison	How do values differ across categories?	Bar, column, grouped bar, dot plot, bullet chart
Composition	What are the parts of a whole?	Pie, donut, stacked bar, 100 % stacked bar, treemap, waterfall
Distribution	How are values distributed?	Histogram, density plot, box plot, violin plot, strip plot
Relationship	How do two or more variables relate?	Scatter plot, bubble chart, correlation heatmap, network diagram
Trend	How does a value change over time?	Line, area, sparkline, slope chart, candlestick
Ranking	What is the order of magnitude?	Sorted bar, Lollipop chart, dot plot, ordered cleveland dot plot
Geographical	How does a value vary across locations?	Choropleth map, dot-density map, flow map, cartogram

13.4 Comparison Charts

Comparison charts show how a numerical value differs across discrete categories. The bar chart, in its various forms, is the workhorse of this category and almost always the right starting point.

Vertical Bar (Column) Chart: Use for a few categories where category names are short. The eye reads height as quantity directly.
Horizontal Bar Chart: Use when category labels are long, or when there are many categories. Sorting from largest to smallest aids readability.
Grouped (Clustered) Bar Chart: Use when comparing values across two categorical dimensions — for example, sales by product across years. Limit to two or three groups; more becomes unreadable.
Stacked Bar Chart: Use to compare totals across categories and show their composition. The eye reads the bottom segment accurately and the upper segments less so; reserve the bottom slot for the most important segment.
100 % Stacked Bar Chart: Use when only the proportions matter, not the totals. Each bar reaches the same length and shows percentage shares.
Dot Plot (Cleveland Dot Plot): Use for many categories where space is constrained. A single point per category along an axis. Reads more cleanly than a tall thin bar chart.
Bullet Chart: Use to compare a single value against a target and a range — a compact alternative to the gauge or speedometer chart, with far better data-ink ratio.

13.5 Trend Charts

Trend charts show how a value changes over time. The horizontal axis carries time; the vertical axis carries the value. The line chart is the workhorse; areas, sparklines, and slope charts handle special cases.

Line Chart: Use for continuous time series. The eye reads slope as rate of change.
Multi-Line Chart: Use for several series over the same time range. Limit to four or five series; for more, use a small-multiple grid of line charts.
Area Chart: Use to emphasise the cumulative magnitude as well as the change. Useful when the value is genuinely a quantity that fills space (revenue, output) and less appropriate for rates and ratios.
Stacked Area Chart: Use to show how composition changes over time. Avoid if individual category trends matter more than the total.
Sparkline: Use as a tiny embedded line chart inline with text or in a table cell. Conveys trend at a glance without dedicating chart space.
Slope Chart: Use to compare two time points across many categories — for example, market shares by region in 2020 and 2025. Each category is a single line connecting two points.
Candlestick Chart: Use for financial price series where open, high, low, and close are all relevant. The chart is dense and rewards a financially trained reader.

13.6 Composition Charts

Composition charts show how a whole is divided into parts. The reader’s question is what fraction of the total does each part represent? The pie chart is the famous-but-flawed default; bar-based alternatives are usually clearer.

Pie Chart: Use only for two to four categories with substantially different shares, where the message is one slice dominates. For more than four categories, the angles become indistinguishable. Few notes that pie charts are over-used in business reporting precisely because their flaw — angle encoding — is invisible to the casual reader.
Donut Chart: A pie with the centre removed; carries the same encoding as the pie. The central space can carry the total or the headline figure.
Stacked Bar Chart: Already covered under comparison; equally useful for composition.
100 % Stacked Bar Chart: Often the better choice than a pie when the question is what fraction. Multiple 100 % stacked bars allow comparison of composition across categories — something a single pie cannot do.
Treemap: Use for hierarchical composition with many small categories. Areas are easy to compare in aggregate but harder for precise quantitative reading. Useful for showing structure rather than for precise comparison.
Sunburst Chart: A radial treemap with concentric rings showing levels of hierarchy. Often beautiful, occasionally informative; use sparingly.
Waterfall Chart: Use to decompose a starting value into the additions and subtractions that produce an ending value. Standard in finance for revenue-to-profit walks and for variance bridges.

13.7 Distribution Charts

Distribution charts show how values are distributed across a numerical range. The reader’s question is what is the shape, centre, and spread of this variable, and where are the outliers?

Histogram: Use to show the frequency distribution of a continuous variable. The choice of bin width is consequential; too few bins hide structure, too many produce noise.
Density Plot: A smoothed alternative to the histogram. Use when comparing distributions of several groups overlaid on one chart.
Box Plot (Box-and-Whisker): Use for compact comparison of distributions across many groups. Shows median, quartiles, and outliers in a small footprint. Reads cleanly side by side.
Violin Plot: Combines a box plot with a density plot. Useful when distributional shape matters and a box plot alone would hide bimodality or skewness.
Strip Plot / Dot Plot: Each observation as a point on a single axis. Useful for small datasets where every value should be visible.
Beeswarm Plot: A strip plot where points jitter to avoid overlap. Useful for moderate datasets where individual values still matter.
Q-Q Plot: A specialist statistical chart that plots quantiles of one distribution against quantiles of another. Used to assess normality or compare distributions formally.

13.8 Relationship Charts

Relationship charts show how two or more variables relate. The reader’s question is do these variables move together, and if so, in what pattern?

Scatter Plot: The fundamental chart for the relationship between two continuous variables. Each observation is a point at \((x, y)\). The eye reads correlation, clusters, and outliers easily.
Bubble Chart: A scatter plot in which a third variable is encoded by the area of each point. Useful for three-variable comparison; remember that area is low on Cleveland’s hierarchy and reads less precisely than position.
Scatter-Plot Matrix (SPLOM): A grid of pairwise scatter plots covering many variables. Useful in exploratory data analysis to spot relationships across a multivariate dataset.
Correlation Heatmap: A coloured matrix of pairwise correlations among many variables. Useful for high-level structure; not a substitute for scatter plots when the precise shape of relationship matters.
Connected Scatter Plot: A scatter plot in which the points are connected in time order. Reveals trajectories — useful for showing how two variables co-evolve.
Network Diagram (Graph): Nodes connected by edges. Used for relational data such as social networks, supply-chain links, or citation patterns.

13.9 Specialised Business Charts

Several chart types serve specific recurring business questions:

Funnel Chart: Visualises a sequence of stages with successive drop-off — sales pipeline, recruitment funnel, conversion funnel. Each stage is shown as a horizontal bar narrowing from top to bottom.
Gantt Chart: Visualises project tasks, their durations, and their dependencies on a horizontal time axis. Standard in project management.
Sankey Diagram: Visualises flows between categories with band widths proportional to flow magnitude. Useful for budget walks, customer-journey flows, and energy-balance diagrams.
Radar (Spider) Chart: Plots multiple quantitative variables on radial axes from a common centre. Effective for comparing entities across a small set of attributes; degrades quickly past five or six axes.
Heatmap (Matrix Heatmap): A grid in which each cell is coloured by a numerical value. Useful for time-of-day-by-day-of-week patterns, calendar heatmaps, and table replacements where colour aids scanning.
Bubble Map (Proportional Symbol Map): A geographical map in which point sizes encode magnitude at locations.
Bullet Chart: Compact alternative to gauges and dials, already covered under comparison.

13.10 Geographical Charts

When the data has a geographical dimension, the reader’s question is often where?

Choropleth Map: Regions shaded by a numerical value. Standard for density and rates. Use rates, not absolute counts, to avoid the larger-region-bias problem.
Dot-Density Map: One dot per unit of magnitude, distributed across the geography. Communicates raw counts honestly without the area bias of a choropleth.
Flow Map: Lines connecting origin and destination, with thickness encoding flow magnitude. Used for migration, trade, transport, and supply-chain flows.
Cartogram: A distorted map in which areas are scaled by a variable other than geography (population, GDP). Useful for elections and demographics.
Heat Map (Geographical): A continuous shaded surface across geographic space, distinct from the matrix heatmap.

13.11 A Practical Chart Chooser

flowchart TD
    Q["What does the<br>reader need<br>to see?"]
    Q --> A["Compare<br>values"]
    Q --> B["Show parts<br>of a whole"]
    Q --> C["Show a<br>distribution"]
    Q --> D["Show a<br>relationship"]
    Q --> E["Show change<br>over time"]
    Q --> F["Show a<br>geography"]
    A --> A1["Bar / column<br>(few categories)<br>Dot plot or sorted<br>bar (many)"]
    B --> B1["100% stacked bar<br>Treemap (many)<br>Pie (only 2–4)<br>Waterfall (delta)"]
    C --> C1["Histogram<br>Density plot<br>Box plot<br>Violin plot"]
    D --> D1["Scatter plot<br>Bubble chart<br>Correlation heatmap"]
    E --> E1["Line chart<br>Area chart<br>Sparkline<br>Slope chart"]
    F --> F1["Choropleth (rates)<br>Dot-density (counts)<br>Flow map<br>Cartogram"]
    style Q fill:#e3f2fd,stroke:#1976D2
    style A fill:#e8f5e9,stroke:#388E3C
    style B fill:#fff8e1,stroke:#F9A825
    style C fill:#fff3e0,stroke:#EF6C00
    style D fill:#fce4ec,stroke:#AD1457
    style E fill:#ede7f6,stroke:#4527A0
    style F fill:#eceff1,stroke:#455A64

The flowchart above is a starting point, not a rule book. The right chart almost always depends on a property of the specific dataset that no generic chooser can capture: the number of categories, the spread of values, the importance of precise reading versus pattern detection, and the medium in which the chart will appear. Use the chooser to narrow the field; let the data finish the choice.

13.12 Common Pitfalls

Pie Chart Default: Reaching for a pie chart whenever the question involves “share of”. For more than four categories, almost any alternative is better.
Line Chart for Categorical Data: Connecting points with a line when the x-axis is unordered categorical (regions, products). The line implies an order that does not exist.
Bar Chart for Continuous Distribution: Using a bar chart where a histogram is needed. Categories along the x-axis must be discrete.
Stacked Bar Confusion: Asking the reader to compare middle segments of stacked bars; only the bottom segment can be read accurately on a baseline.
Dual-Axis Mismatch: Putting two unrelated series on a single chart with two y-axes. The implied correlation is whatever the designer chose by scale.
Three-Dimensional Effects: Adding 3D to a chart whose data is two-dimensional. The effect distorts proportions and adds no information.
Choropleth of Counts: Shading regions by absolute count rather than rate. Larger regions appear extreme regardless of incidence.
Treemap Without Hierarchy: Using a treemap for flat categorical data where a sorted bar chart would be clearer.
Radar Chart for Many Attributes: Pushing a radar chart past five or six axes; the area becomes meaningless and comparisons unreliable.
Funnel Chart Without Drop-Off: Using a funnel for any sequence even when the rates are flat; the visual implies attrition that is not in the data.
Showing Off the Chart Library: Using exotic chart types because they are available, rather than because they answer the reader’s question.

13.13 Illustrative Cases

The following short cases illustrate how the right chart-type choice clarifies a business question. They describe common analytical situations and the reasoning behind the chart selection.

Sales Performance by Region: Sorted Horizontal Bar

A regional manager wants to see which of twelve regions is over- or under-performing against target. The original design is a column chart sorted alphabetically. The redesign uses a horizontal bar chart sorted by performance, with the target as a thin reference line. The pre-attentive ordering does the work of communicating the ranking.

Product-Mix Evolution: 100 % Stacked Bar over Time

A retailer wants to show how the product-mix share has shifted over five years. The original draft is five pie charts side by side. The redesign uses one 100 % stacked bar chart with a time axis, in which each colour band represents a category. The reader reads the change in share as the change in band width over time.

Customer Acquisition Funnel: Funnel Chart with Conversion Rates

A digital marketing team tracks visitors through five conversion stages — visit, sign-up, activation, first purchase, repeat purchase. A funnel chart shows the absolute counts at each stage, while small annotations beside each step report the stage-to-stage conversion rate. The combined chart answers two questions in one image.

Regional Health Outcomes: Choropleth of Rates, Not Counts

A national health agency maps malaria cases across districts. The first draft is a choropleth shaded by absolute case count; large rural districts dominate. The redesign shades by cases per ten thousand population using a perceptually uniform palette, and the geographical pattern of incidence becomes visible at a glance.

Project Timeline: Gantt Chart with Dependencies

A programme office must communicate a multi-track project plan covering eighteen months and four parallel workstreams. A Gantt chart with workstream rows, task bars, and dependency arrows answers the questions of what, when, and on what does this depend without requiring the reader to read paragraphs of narrative.

Summary

Concept	Description
Foundations and Selection Framework
Why Chart Choice Matters	The chart type is the grammar of the visual sentence; the wrong choice obscures the analyst's point
Chart Selection Question	Is the reader comparing, looking at composition, distribution, relationship, trend, or location
Number of Variables	One, two, three, or many variables to show in the same chart
Kind of Data	Categorical, ordinal, quantitative continuous, time, or geographical
Number of Categories	A handful, a few dozen, or thousands of categories or points
Static or Interactive	Will the reader hover, filter, drill down, or only look
Audience	Board member with one minute, analyst with an hour, or public audience without context
The Seven Categories
Comparison Category	How do values differ across categories
Composition Category	What are the parts of a whole
Distribution Category	How are values distributed across a numerical range
Relationship Category	How do two or more variables relate
Trend Category	How does a value change over time
Ranking Category	What is the order of magnitude across categories
Geographical Category	How does a value vary across locations
Comparison Charts
Vertical Bar Chart	Use for a few categories with short names; eye reads height as quantity directly
Horizontal Bar Chart	Use for long category labels or many categories; sort largest to smallest
Grouped Bar Chart	Use for two categorical dimensions; limit to two or three groups
Stacked Bar Chart	Compare totals across categories and show their composition; bottom slot reads accurately
100% Stacked Bar Chart	Use when only proportions matter; each bar reaches the same length
Cleveland Dot Plot	One point per category along an axis; clean alternative for many categories
Bullet Chart	Compare a single value to a target and a range; compact gauge alternative
Trend Charts
Line Chart	Use for continuous time series; eye reads slope as rate of change
Multi-Line Chart	Use for several series over the same time range; limit to four or five
Area Chart	Emphasises cumulative magnitude as well as change; for genuine quantities, not rates
Stacked Area Chart	Show how composition changes over time; mind individual category trends
Sparkline	Tiny embedded line chart inline with text or in a table cell
Slope Chart	Compare two time points across many categories; each category a line
Candlestick Chart	Financial price series with open, high, low, and close; rewards trained readers
Composition Charts
Pie Chart	Use only for two to four categories with substantially different shares
Donut Chart	Pie with centre removed; central space can carry total or headline
Treemap	Hierarchical composition with many small categories; areas read in aggregate
Sunburst Chart	Radial treemap with concentric rings; beautiful, occasionally informative
Waterfall Chart	Decompose a starting value into additions and subtractions to an ending value
Distribution Charts
Histogram	Frequency distribution of a continuous variable; bin width is consequential
Density Plot	Smoothed alternative to histogram; useful for overlaid group comparison
Box Plot	Compact comparison of distributions across many groups; reads cleanly side by side
Violin Plot	Box plot plus density plot; useful when distributional shape matters
Strip / Dot Plot	Each observation as a point on one axis; for small datasets
Beeswarm Plot	Strip plot with jitter to avoid overlap; for moderate datasets
Q-Q Plot	Plots quantiles of one distribution against another; specialist statistical chart
Relationship Charts
Scatter Plot	Fundamental chart for relationship between two continuous variables
Bubble Chart	Scatter with third variable encoded by point area
Scatter-Plot Matrix	Grid of pairwise scatter plots; useful in exploratory analysis
Correlation Heatmap	Coloured matrix of pairwise correlations across many variables
Connected Scatter Plot	Scatter where points are connected in time order; reveals trajectories
Network Diagram	Nodes and edges; for social networks, supply chains, citations
Specialised Business Charts
Funnel Chart	Sequence of stages with successive drop-off
Gantt Chart	Project tasks with durations and dependencies on a horizontal time axis
Sankey Diagram	Flows between categories with band widths proportional to magnitude
Radar Chart	Multiple quantitative variables on radial axes; degrades past five or six
Matrix Heatmap	Grid coloured by numerical value; useful for time-of-day-by-day-of-week patterns
Bubble Map	Geographical map with point sizes encoding magnitude at locations
Geographical Charts
Choropleth Map	Regions shaded by numerical value; use rates, not absolute counts
Dot-Density Map	One dot per unit of magnitude across the geography; honest counts
Flow Map	Lines connecting origins and destinations with thickness encoding flow
Cartogram	Distorted map with areas scaled by a variable other than geography
Common Pitfalls
Pie Chart Default	Pitfall of reaching for a pie whenever the question involves share of
Line Chart for Categories	Pitfall of connecting points with a line when the x-axis is unordered categorical
Bar for Continuous Distribution	Pitfall of using bars where a histogram is needed for continuous distribution
Stacked Bar Confusion	Pitfall of asking the reader to compare middle segments of stacked bars
Dual-Axis Mismatch	Pitfall of putting unrelated series on a single chart with two y-axes
Three-Dimensional Effects	Pitfall of adding 3D to a chart whose data is two-dimensional
Choropleth of Counts	Pitfall of shading regions by absolute count rather than rate
Treemap Without Hierarchy	Pitfall of using a treemap for flat categorical data where a sorted bar would be clearer
Radar With Many Attributes	Pitfall of pushing a radar chart past five or six axes
Funnel Without Drop-Off	Pitfall of using a funnel where the rates are flat and no attrition exists
Chart-Library Showmanship	Pitfall of using exotic chart types because they are available, not because they answer the question