Chart reference guide

Information on over 30 of the most commonly

used charts including what they are, when to
use, examples, and some design tips.

900

800

700

600

500

400

300

200

100

copyright © Sara Dholakia for Smashing Magazine 2023

Just getting into the world of data visualization? Don’t worry, below are descriptions of commonly
used terms and phrases you’ll find throughout the guide.

Nomenclature

Chart purpose Types of variables

Comparison: How are the elements similar Quantitative variable: This is any numerical
or different? This can be among items and/or variable (ex: population size or temperature)
over time. Example: comparing the sales of two and includes continuous and discrete variables.
different brands of dog food in a single retail
location. Continuous variable: This is a type of
quantitative variable that can take any
Composition: What parts make up the whole? value between two numbers (ex: weight or
The composition can be static or change over temperature).
time. Example: showing the breakdown of the
diet of Pallas cats. Discrete: This is a type of quantitative variable
that, unlike continuous variables, has a limited
Distribution: Where do the values in a number of possible values. Discrete variables
data set fall? Are there outliers? Example: are finite, numeric, countable, non-negative
communicating the distribution of grades within integers (ex: the number of people who have
a middle school class, including the average and been to space).
outliers.
Ordinal variable: This is a non-numeric variable
Relationship: How do the elements relate to that has a natural order (ex: days of the week
each other? Is there a correlation? Example: or spiciness levels on a menu (mild, spicy, very
showing how colder temperatures are spicy)).
correlated with a decrease in ice cream sales.
Categorical (aka nominal) variable: This
describes variables that don’t have an inherent
Chart vs. graph order or numerical values (ex: oak, ash, and elm
trees; pink, purple, and blue).
Graph: This refers to visual representations of
data on a Cartesian plane (they often look like a
grid and have an x-, y-, and sometimes a z-axis).

Chart: This is used as a catchall word for visual

representations of data and includes graphs,
diagrams, tables, etc.
Alluvial diagram and marine animal populations from 1970-
These are diagrams that 2020
were created to show flow Distribution Relationship
or change in a system over
time. They’re made of two
or more vertical axes. Area graph – stacked
The flow moves from left to right. Categorical These are like simple area
variables are placed on each axis and their graphs. The difference is
value is represented by their height on the axis. that they show two or more
The amount of the component moving from one data series. Each data
axis to another is represented by the height of series after the first one
the line connecting the axes. starts where the one before it ended. In other
The height or stroke weight of the horizontal words, instead of having all of the variables
lines within the diagram are proportional to the start at the zero point, each point on the area
flow. They help to communicate how much of a graph starts where the previous one stopped.
variable is moving. When to use: They can be useful to compare
When to use: If you’re communicating change how multiple variables change over a period of
or flow within a simple system and want to time. However, they shouldn’t be used if you’re
communicate relative proportions alluvial trying to communicate specific values.
diagrams can be a good choice. If absolute With multiple independent categories, a line
values are an important component of your graph will usually be a better choice.
message, label each line or arrow with its value.
Example: Change in sales of a company over
Example: Change in profession chosen by the last 5 years, displayed by department
college graduates over a ten year period
Comparison Distribution
Comparison Distribution Relationship

Area graph – simple Bar graph – simple

Area graphs are similar These are one of the
to line graphs except that simplest and most common
the area under the line is data visualizations. Bar
filled in with a color and/ graphs are made of either
or pattern. Like line graphs, vertical or horizontal
they’re used to show data over a period of time. bars and are used for numeric comparison of
This makes it easy to identify change. categories.
When to use: While they’re similar to line When to use: They’re usually a great option
graphs, area graphs are generally used to when you’re comparing numerical data across
display overall trends rather than specific multiple categories. Because they’re common
values. They can also be useful to communicate and easy to read, they work with all audiences.
large differences between data points or plot
data over a very long period of time. However, bar graphs may not be a good choice
if: the data is continuous, you want to show
Example: The worldwide decline in terrestrial change over time, or you’re comparing parts of a
larger set of data. Bar graph – grouped aka multi-set bar graph
These are used to compare
Additionally, you should opt for a lollipop graph
data values across two or
instead of a simple bar graph if most of the
more components of two
variables you’re dealing with have high values
or more categories. Each
(in the 80-100% range) because having a lot
component is assigned
of tall columns like you would with a bar graph
a color and/or pattern fill
can be visually aggressive and cause an optical
and grouped with the other components of its
effect called a Moiré pattern which can confuse
category. Grouped bar graphs can be displayed
the audience.
vertically or horizontally.
Example: Comparison of profits from each of
When to use: Grouped bar graphs can be used
five different products sold at a store in the last
to successfully show within group and between
week
group comparisons.
Design tips: If you have long labels on one axis
Example: Imagine we’re looking at dollars
and have space, try displaying the bar graph
spent by a company’s customers in France,
horizontally. If that doesn’t work, you can set
Germany, and Italy over the course of the
the axis labels at an angle to the axis line. Also,
previous three years. If we plotted the data in a
try to stick to 10 bars or fewer.
grouped bar graph, we’d be able to see multiple
Comparison comparisons including:
1. Dollars spent by customers in France over
the three years
Bar graph – floating aka waterfall graph 2. Dollars spent by customers in Germany over
They are used to compare the three years
categories across a 3. Dollars spent by customers in Italy over the
numerical variable like three years
simple bar graphs, but 4. In year 1, dollars spent by customers in
instead of showing a single France vs Germany vs Italy
value for each category, 5. In year 2, dollars spent by customers in
they can show a range of values. For this France vs Germany vs Italy
reason, the bars don’t start at the same place 6. In year 3, dollars spent by customers in
on the x-axis; some don’t touch the x-axis so France vs Germany vs Italy
they “float.” This allows you to show data with 7. If you’re willing to do some math, you
different minimum or starting values. These can
can compare the total dollars spent by
be displayed vertically or horizontally.
customers in all 3 countries across 3 years
When to use: Floating bar graphs can work Comparison
well when the data has different minimum or
starting values.
Example: High and low temperature for a period Bar graph – stacked
of seven days Stacked bar graphs are
Comparison Distribution used to show data for two
or more components within
a single variable. Because
each bar in the graph
breaks the variable down into its components, a Bubble graph
stacked bar graph shows you the total for each They’re multivariate
instance of the variable in addition to showing graphs that are like a cross
the part-to-whole relationship of the variable. between a scatterplot and
A stacked bar graph can consist of a single a proportional area chart.
category or multiple categories. “Multivariate” means that
there are multiple variables. For the graph, that
When to use: Stacked bar graphs work well
means that there are three dimensions instead
when you’re both: comparing the totals for a
of the usual two (that are plotted on the x and
variable and want to give insight into what is the
y axes). The third dimension is represented by
largest contributing component to those totals.
the area of the circle or bubble. This is a dual
Example: A household’s total expenses for one purpose data viz designed for comparison and
month with the different component expenses to show relationships.
called out (i.e. utilities, groceries, recreation,
When to use: They’re better for showing
etc.)
relationships rather than exact data values.
Design tips: If your data has long labels and The addition of the third variable makes
you have the space, try displaying the bars reading these more difficult than reading either
horizontally. Also, it’s generally a good idea to a scatterplot of a proportional area chart,
limit the number of components within a bar to especially when there are too many bubbles.
four. Use this sparingly. They can work with limited
Comparison Composition datasets and when the audience is familiar with
this type of chart.

Example: Showing the relationship among life

Box & whisker plot aka box plot expectancy (x- or y-axis) and GDP per capita (x-
These are used to show or y-axis) by population size (circle area)
variation and distribution
Design tip: To help prevent misinterpretation of
of sets of data. Each bar
the data, base the size of each circle on its area
contains an upper quartile
rather than its radius or diameter.
(where the upper 25% of
data falls), a lower quartile (where the lower Comparison Distribution Relationship
25% of data falls), and a median (shown as a
dividing line between the quartiles). If the data
has outliers, then the bar will have a line, or Bullet chart
“whisker,” extending from either side out to the Bullet charts are usually
value of the high and/or low outlier. used to show performance
data. They look like a
When to use: They’re a good choice to
thermometer or progress
communicate spread & skewness and/or
bar. Bullet charts are used
outliers. They can be displayed vertically or
to show a primary measure as a rectangle or
horizontally.
bar placed on a larger bar that’s split up into
Example: Height distribution of students in five different ranges (for example, poor, acceptable,
middle school classes and good). There’s often a goal or target value
Comparison Distribution on the bar as well for a comparative measure.
When to use: Bullet charts are a great choice showing, Circle packs can be a good choice.
when you want to show a performance measure However, if your space is limited, you should
in relation to a target or targets. They were consider a tree map because they’re more
created as an alternative to meters and gauges, space efficient (lots of empty space within the
and are better choice. Meters and gauges take circles).
up a lot of space while communicating limited
Example: A visual representation of theHubble
information. Bullet charts take up fairly little
telescope’s photo targets (one of the largest
space and let you communicate a lot with that
circles could represent galaxy with one of the
space.
circles within that being spiral galaxies and a
Example: One person’s LDL cholesterol (primary circle within spiral galaxies could be Milky Way).
measure) laid out on a bar with ranges (low, Composition Distribution Relationship
good, high) with the goal
Comparison Distribution
Connecting line map aka connection map
They’re used to show
Chloropleth map geographic relationships
Chloropleth maps use color and connections. They
and/or pattern to show data can also be used to show
values for a certain variable a route. Connecting line
across a geographic area. maps are created by placing lines from point to
They’re similar to heat point on a geographic map.
maps but are visualized on geographic areas When to use: These can be a good choice
rather than in a matrix. to map out a route and to show both spatial
When to use: These are a good choice when patterns and the distribution of connections.
you want to show data for a variable across a Example: A map with a route for a road trip
geographic area.
Distribution
Example: A county map showing predicted
amounts of rainfall for the following day
Distribution Donut chart
They’re a variant of pie
charts. Both pie and
Circle pack aka circular tree map donut charts are used to
They’re similar to tree show the proportions of
maps in that they show components that make up
hierarchy, but they a whole. However, donut charts have the center
use circles instead of removed, hence the name.
rectangles to communicate When to use: Donut charts make it easier to
components. Optionally, the area of each circle compare different sections than pie charts do,
can be used to represent a value. so if that’s your goal, use a donut chart. Use
your best judgment when deciding how many
When to use: If you have a lot of room and don’t
section/components is too many. Remember
need to include labels for each thing you’re
that your audience should be able to quickly
read and understand the chart. they’re not a great choice when you want
to communicate exact values. You could
Example: Showing the results of a five question
counteract that by including the value in each
survey
cell, but if you have more than a handful of rows
Design tip: You can use the center of the chart and columns, it’ll become crowded and more
to place labels or other info. difficult to read.
Comparison Composition
Example: Showing temperature changes across
five cities over the course of a year (rows could
be labeled with city names, columns could be
Gauge chart
labeled with months of the year, and the colors
They display a single value would correspond to temperatures)
with qualitative context
Relationship
in the form of ranges of
numbers (for example:
poor 0-15, good 16-30,
best 31-35). The shape is usually a half circle or Histogram
a three-quarter circle (like a speedometer). Histograms are used to
show fistribution of data.
When to use: Gauge charts are easy to read so
This can include frequency
they can provide a good at-a-glance view when
distribution and time series
you’re showing a single data value and want to
data. Histograms look
give it some context. However, they aren’t good
like bar graphs because they’re made up of
for communicating complex data.
rectangles plotted along an axis, but there are
key differences. Bar graphs show at least one
Example: A gauge that communicates the
categorical variable and one numerical variable.
average response on a customer satisfaction
Histograms don’t use categorical variables
survey. The gauge is broken into three sections:
at all. Instead, they use ranges or intervals.
dissatisfied, satisfied, expectations exceeded.
Usually, the intervals are plotted on the x-axis,
Comparison and the frequency of the interval is plotted on
the y-axis. Histograms often show continuous
data, and this is why they’re often displayed
Heat map with the rectangles or bars touching.
They’re visual, color-
coded representations of When to use: Histograms are one of the best
multivariate data. Heat and most often used tools to visualize frequency
maps are made up of a distribution of a set of data because they
table, and usually, all of the can generally quickly show where the data is
rows show one category and all of the columns concentrated and what the extremes are.
show another category. Table cells are colored Example: Showing the frequency distribution of
according to their value. students’ mid-semester grades (the intervals
could be something like: 0-69, 70-79, 80-89,
When to use: Heat maps provide a way to 90-100)
visualize complex data at-a-glance, and
Distribution Relationship
they can show change over time. However,
Line graph “parent” nodes can be made larger or shown in
Line graphs show a different color.
information as a series of
When to use: Network diagrams are a great
data points. Often, these
way to visualize the components and layouts
are represented as small
of a network. However, at a certain point when
dots on a graph. The points
there are too many nodes, the visualization
are connected by line segments. They show
becomes overwhelming, so be thoughtful when
values for something over a period of time. The
considering how much detail is helpful.
x-axis usually shows a time scale while the
y-axis tends to show quantitative values. Example: Relationships between characters in a
book
When to use: Line graphs are a great tool to
Comparison Composition Relationship
show patterns and trends as well as change
over time. They’re also one of the few data
visualizations that can show negative values.
They can have multiple lines, each representing Parallel coordinates plot
a separate series of data if you want to compare They’re used to display
trends, for example. However, if there are too multivariate numerical
many lines, particularly when they’re plotted data. In other words, they
close together, it can be difficult to read. allow for comparison of
multiple variables so you
Example: The recorded population of two types can see relationships.
of marine animals over the span of five years
Each variable is placed on its own vertical line
Design tip: If you have multiple lines, try using or axis. This means that each axis can use a
both color and different line types (dashed, different scale of measure. For example, one
dotted, etc.) to code the different categories axis can represent fuel efficiency (measured in
for better accessibility. This will ensure that miles per gallon), another axis can represent
individuals with color impairment can still easily cylinders (measured by number of cylinders),
read the graph. and a third can represent cost (measure in tens
Distribution Relationship of thousands of dollars). The data is plotted as
lines connecting across all of the axes.
When to use: They can be a good choice for
Network diagram showing the relationships across and among
multiple variables. However, they can become
They’re a visual
overly cluttered and difficult to read, so use your
representation of the
best judgment when it comes to the amount of
relationships between
data being shown at one time
things, often within a
system. Each component Example: Comparing different variables of car
of the system is represented with a node, performance
and the connections between the nodes are Design tip: If used with an audience that’s less
represented with line segments or arrows. familiar with data visualizations, provide extra
Additional context and information can be context and visual cues.
provided in different ways. For instance, if Comparison Relationship
elements of the network exist in a hierarchy, the
Pie chart relative size.
They’re circular graphs When to use: Proportional area charts are
used to show or compare a good method of conveying proportions.
the components of a However, they aren’t a great choice when you
category. They get their have a lot of items nad/or long labels.
name from their pie-like Example: Relative sizes of rodents in North
appearance. Each slice of the pie represents America
one component of the whole and when all of
Comparison
the pie chart’s components are added up, they
must equal 100%.
Proportional symbol map
When to use: (TLDR; use thoughtfully and
They use symbols (usually
rarely) They can be used with any content
a simple geometric shape
that can be broken down into its individual
like a circle) overlaid on
components. However, it’s generally best not
a map to communicate
to use a pie chart when there are more than
amount or proportion of a
six categories because then the pie chart can
variable in different geographic locations. The
become difficult to read.
area of the symbol is proportional to its value.
Additionally, they’ve rightly gotten some
Additionally, the symbols can be labeled with
criticism for being unhelpful. If multiple pie
the actual value if room permits.
slices are similar in size, it’s hard to compare
them because most people can’t calculate the
When to use: These are a good choice if you
area of a arc or slice of pie in their heads.
want to show locations where a certain variable
is present as well as provide at-a-glance info on
Pie charts can be a workable solution if you
the proportion of the variable in each location
have only a few components and they’re
relative to other locations.
labeled. That being said, I still find that donut
charts are almost always a better alternative
Example: World map showing living nobel
because with donut charts, all the segments
laureates
have the same width so you’re just comparing
the lengths rather than the area of a part of a Comparison Distribution
circle.

Radar chart
Example: Revenue split up by different sources
They’re used to showing
Comparison Composition
multivariate data that
allows you to see which – if
any – variables have similar
Proportional area chart values and/or if there are
A proportional area chart outliers. They can also be used to visualize
can be made up any shape performance since they allow you to see which
but the most frequently variables scored higher and lower within a set.
used shapes are squares All of the variables have an axis that starts in
and circles. They’re used the center (of a circle or some other shape) and
to communicate an at-a-glance comparison of grows outward. All of the axes are arranged
radially and are equidistant. The scale is the Scatterplot
same on all of the axes, and they’re connected They’re used to show
by lines. data for two different
quantitative variables. Data
When to use: Radar charts can be useful
values are usually plotted
for simple comparisons (especially of
as dots. Sometimes, there
performance), but they can very quickly become
will also be a line among the dots, and it’s
too cluttered and hard to read. At most, try to
called the “line of best fit.” This helps to show
stick with no more than four or five variables
correlations that may exist between the two
They’re not a good choice to compare a single variables. The types of correlations that you
value across variables because their circular can see are: positive (as one variable increases,
nature makes them more difficult to read than a so does the other), negative (as one variable
linear visualization. Radar charts work best for decreases, the other increases), null (no
audiences that are familiar with them. correlation), linear (the dots basically form a
straight line (the y-values increase by a constant
Example: Comparing performance of two makes
amount as the x-values increase), exponential
of car across five performance measures
(the y-values have equal ratios, not equal
Comparison differences), and u-shaped (the data points
create a ‘u’ shape).

When to use: Scatterplots are a great choice

Radial bar chart when you have a pair of quantitative variables
They’re like bar graphs that you want to show either a correlation or a
but the bars are circular lack of correlation for. Super important note:
instead of rectangular and remember that correlation doesn’t necessarily
aren’t plotted on a graph. imply causation!
Because of this, they aren’t
placed on a grid. Radial bar charts are used to Example: The length of cats’ tails and the
show comparisons between categories. average length of their whiskers
Distribution Relationship
When to use: While radial bar charts can
be aesthetically appealing, they don’t
communicate exact data values very well.
Tree diagram
Because these charts are circular, as you move
from the center outward, each bar gets longer Tree diagrams are used
because the radius for the circle they’re plotted to show hierarchy and
on keeps increasing. That being said, they can relationships. They start
be used for at-a-glance, simple comparisons with a root node and from
between very different values and to show there, they branch out into
progress or motivate the audience (think of the the next level of nodes and continue until they
“rings” on the fitness app on the apple watch). reach the end of the hierarchy where the leaf
nodes live.
Example: To show how much progress you’ve When to use: Tree diagrams are a wonderful
made toward your steps goal vs. your cardio tool for showing hierarchy and are used in a
goal variety of cases like taxonomy, ancestry, etc.
Comparison
Example: Navigational structure for a website Word cloud
Composition Relationship They’re a cluster of words
aquamarine space in different sizes. The larger
cyan sky cerulean
the word, the more often
azure midnight indigo
turquoise tiffany
Tree map
ocean that word shows up with a
royal prussian cobalt

They’re a way to display data set. They’re used to

part-to-whole relationships communicate the relative frequency of words or
between a lot of phrases.
hierarchical categories. The When to use: Word clouds don’t provide precise
size of each component info but they can be a way to quickly depict
represents its proportional size, and the tree simple information about word frequency in a
map is made up of rectangles grouped and/or way that adds some visual interest.
nested together.
Example: Keyword metadata for a website
When to use: Tree maps are useful when Comparison
trying to communicate general part-to-whole
relationships but not for conveying precise
information. If your main goal is to show
hierarchy though, tree diagrams are a better
choice than tree maps.

Example: College majors within Liberal Arts

Comparison Relationship

Venn diagram
They’re made up of two
overlapping shapes –
usually circles – and are
used to display similarities
and differences between
two things, sometimes referred to as sets. The
circles contain a list of traits that apply to each
of the two circles in addition to a list in the area
of overlap or intersection. Those traits are true
for both sets.

When to use: Venn diagrams are useful for

displaying similarities and differences between
two sets.

Example: Creatures of the land and of the sea

Comparison Relationship