what is a bar chart?

Feb 19
Written By Alex Velez

In this article, we’ll explore the bar chart: what it is, how to interpret it,
common variations, and tools to build bar charts, plus bonus tips, tricks, and
real-world examples. Let’s jump in and start laying the groundwork for
creating compelling visualizations in the form of bars!

What is a bar chart?

A bar chart is a common chart type for graphing categorical data or data
sorted into groups. It consists of multiple rectangles aligned to a common
baseline. The length of each is proportional to the value it represents—in
other words, in a bar chart, the data is encoded by length. Our eyes are very
good at comparing lengths when objects are aligned, making this graph easy
to interpret—just one reason bar charts are common.

Our eyes start at the base and scan towards the end of each bar. We
measure the lengths relative to both the baseline and the other bars, so it's a
straightforward process to identify the smallest or the largest bar. We can
also see the negative space between varying heights of bars to compare the
incremental difference between them.

Not only are these graphs easy to read, but they are also widely recognized.
Chances are, you've already encountered a standard horizontal or vertical
bar chart. But bars come in many shapes and sizes. I'll list below a few of the
most common variations, with links to examples.
What are common types of bar charts?

VERTICAL BAR CHARTS

Standard vertical bars (also known as
column charts) are one of the most
common varieties, and pretty much all
tools have built-in functionality to create
them. Just because they are common
doesn’t mean they are boring. Read more
on how basic vertical bars can be used to
communicate multiple stories.

HORIZONTAL BAR CHARTS

Horizontal bars are similar to the vertical
version flipped on its side. This is a great
option for long category names because
space on the left-hand side enables
horizontally positioned text (an
alternative to diagonal text on the x-axis
of a vertical bar chart). Horizontal bars
are comfortable to read since the
structure mimics how we process
information (in western cultures), where
we read from left to right, starting at the top. Check out this article for a
simple horizontal bar makeover.

CLUSTERED BARS
Both vertical and horizontal variations
can display more than one series. Be
aware that adding many data series to a
bar chart can make it challenging to
focus and identify key takeaways. It's
also important to note that there is a
natural grouping, so be mindful of the
order of series within each cluster. Read more to see how clustered bars can
be used to compare multiple series.

STACKED BARS
Stacked bars come in both vertical and
horizontal forms. They allow you to
compare the totals by scanning the ends,
but also show the subcomponent pieces
within each category. The challenge with
stacked bars is that it's hard to compare
the subcomponents across categories
that aren't aligned to the same baseline.
Here is a post with more examples of
when stacked bars work.

100% STACKED BARS

An alternative to standard stacked bars
is a 100% stacked bar chart (horizontal
variation shown at left), where the
relative percentages are displayed
rather than the absolute values. The
benefit of this version is it gives you two
consistent baselines for comparison:
one on the left and another on the right.
Read more to see why this can be a
great way to visualize survey data.

WATERFALL CHART
The waterfall chart is a more advanced
variation of a bar chart that isn’t as
widely used. This chart can sometimes
be useful to visualize change, where the first bar represents a starting value,
and the last bar shows the ending quantity. The smaller bars in between
illustrate the incremental change to go from start to finish. Learn how to
create a waterfall chart in Excel.

What tools make bar charts?

Most visualization and analytical tools have built-in functionality to easily
create standard bar charts. Below are a few common ones and source links
to get you started.

 Excel
 Datawrapper
 Flourish
 Tableau
 PowerBI
 PowerPoint
 Data studio
 Infogram

For more tool-specific bar chart examples, check out Exercise 2.2 from Let’s
Practice! The solution features the same vertical bar chart created in a
number of different tools: Excel, Datawrapper, Flourish, Google Data Studio,
PowerBI, R, Python, and Tableau.

What are tips for designing bar charts?

Here are a few formatting steps to consider when designing bar charts.

 Adjust the gap between bars. There is no specific rule here, but
generally speaking, the gap should be thinner than the width of the
bars. In Excel, I’ll typically choose a gap width percentage between
30%-40%. The goal is to strike a healthy balance: you don’t want to
have too much white-space so that the data isn’t prominent, but you
also don’t want the bars so close together that they appear as a single
shape rather than distinct elements. Use good judgment here! (An
exception to this would be if you are creating a histogram, which is a
similar but different chart type for showing a dataset’s underlying
distribution. Learn more about the differences between histograms and
bar charts.)
 Intentionally order bars. Be thoughtful about the order of the bars
and how the reader will interpret them. If there is an inherent ordering
in your data—for example ordinal categories, like age ranges or a
Likert scale—consider making use of this natural order. If this isn’t the
case, consider arranging the bars based on the data, for example from
largest to smallest, or smallest to largest, depending on what you are
trying to convey.
 Consider a legend. I often find myself monkeying around with bar
chart labels more so than other chart types—especially when I’m not
using an axis to label categories. Bar charts don’t typically have a ton
of whitespace, so to prevent clutter, I prefer to label the category
name within a bar. This is sometimes hard to do with vertical or
stacked bars, where horizontal space is limited. For this reason, it’s not
uncommon that I add a legend to a bar chart if I can’t fit the label
inside the bar or directly to the side in a visually appealing way. When I
use a legend, I like to position it beneath the title, so that the reader
encounters how to read the chart before coming across the data. In the
below visual, you can see how I’d typically label series in a horizontal
vs. vertical bar chart.

 Move direct data labels inside bars. Bars can feel very heavy and
take up a lot of ink. If labeling the values of each bar directly, consider
 moving data labels inside the ends of the bars to prevent clutter. This
minor change in position can trick our eyes into thinking there is less
while preserving the same amount of detail.
 Use a zero-baseline. This is an important rule to follow. Because data
is encoded by length, the entire bar range needs to be present;
otherwise, our relative measurement will be inaccurate. If you need to
alter the baseline to communicate smaller differences, chances are
there's an alternative chart type that is better suited for your needs.
Still not convinced? Read more to see an example and a little math on
why this is key.

Where can I see more bar chart examples?

 100% stacked bar makeover: show the full picture
 Another 100% stacked bar makeover: declutter! (and question default
settings)
 Horizontal bar makeover: US prison population revisualized
 Small improvements to a bar chart: power-pairing of color and words
 Alternatives to bars: novel verse the boring old bar chart
 Another bar chart variation: bullet graphs
 what is a line graph?
Mar 24

Written By Mike Cisneros

This article is part of our back-to-basics blog series called what is…?, where
we’ll break down some common topics and questions posed to us. We’ve
covered much of the content in previous posts, so this series allows us to
bring together many disparate resources, creating a single source for your
learning. We believe it’s important to take an occasional pulse on
foundational knowledge, regardless of where you are in your learning
journey. The success of many visualizations is dependent on a solid
understanding of basic concepts. So whether you’re learning this for the first
time, reading to reinforce core principles, or looking for resources to share
with others—like our new comprehensive chart guide—please join us as we
revisit and embrace the basics.

Line graphs (or line charts) are best when you want to show how the value
of something changes over time, or compare how several things change
over time relative to each other. Whenever you hear that key phrase “over
time,” that’s your clue to consider using a line graph for your data.

Line graphs are common and effective charts because they are simple, easy
to understand, and efficient. Line charts are great for:

 Comparing lots of data all at once

 Showing changes and trends over time
 Including important context and annotation
 Displaying forecast data and uncertainty
 Highlighting anomalies within and across data series

On the other hand, they are not necessarily your best choice for:

 Displaying quantities of things

 Working with categorical data
 Making part-to-whole comparisons
 Showing sparse data sets

In this post, we’ll talk about how a line graph works, and how to build one
that provides meaningful information and context to your audience.

Then, we’ll answer some commonly asked questions about line graphs:

 What kind of data can be shown on a line graph?

 How many lines can I show at one time?
 Does my y-axis have to start at zero?
 How much space should I leave at the top and bottom of my graph?

In future articles, we’ll go into more detail about some special variants of line
graphs, like the slopegraph and the area graph.
How does a line graph work?
Think of the lines on a line graph as little histories.

 First, we measure the value of something we want to keep track

of at different points in time. For example, at the end of every
month, our store wants to record how much money we brought in
through sales.
 Once we have lots of these measurements, we plot those points on
our graph, and then draw a line that connects those points, to
make it easier to see how those values rose, fell, or stayed the same
over time.

But that’s not all we can do with line graphs!

In addition to drawing that line, we can include other important contextual

information on our graph. We can choose to mark, label, and/or
annotate individual points on our line; we can add reference lines and
regions; we can even include forecast data points, and distinguish
between actual and projected data by changing the look of our line and
our data markers.

Here’s what a chart like that could look like.

Our collected sales data is still shown on the chart, as the solid black line,
but now we’ve been able to put that data into context.

 A gray reference band shows what the range of monthly sales

totals have been, and we’ve labeled those specific values as well.
 A green line shows us what our sales goal is, to make it easy to see if
our sales numbers were worth celebrating in any given month.
 To reduce some visual clutter, the data labels and most of the
markers of individual points on our line have been removed.
Only the markers for the months in which we hit our maximum and
minimum sales values remain, because they are important to the story
we plan to tell with this graph.
 A dotted line shows the projected sales over the next three months.
 Annotations at key moments in time explain the reasoning for the
numbers, actions we took in response to those numbers, and our
expectations.

Remember: line graphs work by emphasizing how our measured values

change over time, or as a specified independent variable changes. We can
take advantage of the white space in our line graphs to include informative
context that helps an audience understand why these specific changes
matter.
What kind of data can be used on a line graph?
A typical line graph will have continuous data along both the vertical (y-
axis) and horizontal (x-axis) dimensions. The y-axis usually shows the value
of whatever variable we are measuring; the x-axis is most often used to
show when we measured it, either chronologically or based on some
independent variable (e.g., as we rev our old car’s engine, we measure
the decibel level at different RPM).

While some line graphs do not use continuous data on the x-axis (particularly
slopegraphs and parallel coordinates diagrams, which are specialized
variants of line graphs), what we absolutely can’t use on our x-axis is data
that doesn’t have any meaningful relationship among the categories shown.

Let’s say we have a list of the first six months of 2020: January,
February, March, April, May, June. It would feel wrong to list them in any
other order, because they are continuous and have an intrinsic order. January
2020 leads to February 2020, which leads to March 2020, and so on.

Let’s also say we have a list of types of produce: apples, pears, limes,
lemons, dates, grapes. Unlike our list of months, one kind of produce doesn’t
necessarily lead to the next. We could order them alphabetically, by size, by
color, or randomly, and it wouldn’t feel unusual, because they have no
intrinsic order, and are not continuous—they are categorical.
 DO: Show a line of sales by month.

DON’T: Show a line of sales by category.

Line graphs only make sense when there is a meaningful

relationship between successive points on the line. Therefore, a
grocery store could reasonably use a line graph to show their produce sales
data by month, but using one to show sales differences among various types
of produce isn’t recommended. That would be better done in a bar chart.
How many different lines can I put in one chart?
People have a hard time keeping track of more than four or five things
simultaneously, so consider using that as your guideline. If you absolutely
need to show more lines than that, try to use color, line weight, and labeling
to focus on the most important lines or points.

This graph above shows five years’ worth of data for annual diabetes rates of
patients at our local medical center, Anytown Medical, as well as the rates
for 12 other centers in our region. Since what we are likely to care about the
most is our own center (particularly because our rates are low), we are using
color, weight, and labeling to bring focus to Anytown and to the regional
average, while also including the values for all other centers.

Does the y-axis of a line graph have to start at zero?

When we look at line graphs, we compare the lines to each other more than
their height from the x-axis. By their nature, line graphs are better at
showing the change of values over time than an exact quantity at each
measured point.

Line graphs, therefore, are not required to have a zero baseline; but in
most cases, it is still advisable. It is less confusing to an audience if the y-
axis starts at zero, since that is the normal expectation when seeing a chart
for the first time. There should be compelling reasons to subvert those
expectations.

Here are some specific cases for which a zero baseline might not be
optimal:

The range of measurements is small, but the distance from the

minimum value of those measurements to zero is large.

If our y-axis went all the way down to zero, these meaningful fluctuations
would be impossible to perceive.

DO: Start the axis at a logical value.

 DON’T: Include zero if that obscures the data.

The relationship to zero is not meaningful.

On some scales, like on a Fahrenheit scale, the position of “zero” is

somewhat arbitrary (50 degrees isn’t twice as warm as 25 degrees), so if no
data point would ever come close to zero, we shouldn’t feel obliged to
include that as our baseline. Our graphs should include a range of reasonable
values, providing the most meaningful context to a viewer.
 DO: Use a reasonable range of values.

DON'T: Include zero if it's a meaningless value.

Some of your measurements are negative.

It might be important to include a horizontal line as a reference point for

zero, but no data series on our graphs should cut across either of our axes.
It’s preferable to make the minimum value of our y-axis lower than the
minimum measured value of our data, so that our baseline is a boundary line
for our data.
 DO: Keep the axis below the minimum data.

DON’T: Let the axis cut through the data.

IMPORTANT! In situations where your line graph has a non-zero baseline,

it’s important to make that fact obvious to your viewer, so they are not
misled by a zoomed-in perspective on your data.

How much space should I leave at the top and bottom of

my graph?
The specific answer to this question will vary with each graph you make. A
useful starting point is to make the range from your smallest value to
your largest value take up about 70-80% of your graph’s available
vertical space, with a roughly equal amount of white space above and
below it.

From there, you can modify your axis range, and the position of the line or
lines plotted on the graph, to suit your specific needs. The relative slopes of
your lines won’t change, so do your best to select an axis range that lets you
have round numbers for your axis labels, a reasonable amount of white
space in the chart to include data labels and annotations, and is wide enough
(if necessary) to show reference lines, like goals or projected values, that
might be bigger or smaller than your smallest or largest measured value.
In every case, it’s important to be reasonable and thoughtful in how you
present the data, in order for your audience to understand your message
clearly.

Now that we’ve talked about all these different elements of line graphs, the
next step is to make some of your own, using whatever tool or tools you
prefer. The storytelling with data community would be a great place to start!
Try any or all of these quick exercises:

 color + words
 alternatives to pies
 declutter

You could also look back on some of our prior #SWDchallenges, like this one
about annotated line graphs, to see what some of your fellow communication
enthusiasts have created.

Or, continue your journey through our full what is…? chart series, by
browsing other common visuals like bars and pies, or explore our
comprehensive chart guide page for additional chart types.
 what is a pie chart?
May 14
Written By Elizabeth Ricks

This article is part of our back-to-basics blog series called what is…?,
where we’ll break down some common topics and questions posed to
us. We’ve covered much of the content in previous posts, so this series
allows us to bring together many disparate resources, creating a single
source for your learning. We believe it’s important to take an
occasional pulse on foundational knowledge, regardless of where you
are in your learning journey. The success of many visualizations is
dependent on a solid understanding of basic concepts. So whether
you’re learning this for the first time, reading to reinforce core
principles, or looking for resources to share with others—like our new
comprehensive chart guide—please join us as we revisit and embrace
the basics.

Pie charts are one of the most common types of data visualizations.
They are also one of the most widely condemned and misused. The
good news is they do have a proper use case and it’s easy to use them
appropriately.

In this post, we’ll discuss:

 how a pie chart works

 two specific use cases for a pie
 how to identify whether your data is better served as something
other than a pie
 design tips for creating an effective pie

What is a pie chart?

A pie chart expresses a part-to-whole relationship in your data.

Imagine an actual pie (I’ll let you choose your favorite variety!). Each
slice represents one component and all slices added together equal the
whole. Pies are perhaps so commonly used because, as Manuel Lima
writes about in The Book of Circles, humans gravitate to circles as the
universal symbol of unity, perfection, and infinity. We like things to be
whole—and understand how the respective pieces relate to the whole.
A pie chart provides us this comfort of a part-to-whole relationship.

Here’s a brief overview of how we consume a pie. Each pie slice has
three visual components: its central angle, area, and arc length. Recent
research by Robert Kosara and Drew Skau demonstrated that we read
pie charts by area—or possibly in combination with arc length—which
debunked some previously held beliefs that pies are read by angle. You
can read more about their research in an illustrated tour of the pie
chart study results.

Let’s look at an example. Imagine you have a portfolio of five products

and you measure sales (in millions of dollars) for each. The sum of
each individual product’s sales equals the total sales amount. You can
see the part-to-whole relationship of each product’s contribution to the
total in the pie below. In this instance, I put the total dollar volume at
the top and labeled each segment with the percent of the total.
Alternatively, I could have labeled each segment with the respective
dollar amount. Which you choose is largely dependent on what you
want your audience to see or do with the graph.

When should I use a pie chart?

Pie charts are probably better than any other visual for expressing a
part-to-whole relationship. When you hear “percent of…” or “part of…”
that’s one indication a pie chart could meet your needs.

There are two primary use cases for a pie chart:

1. If you want your audience to have a general sense of the part-

to-whole relationship in your data and comparing the precise
sizes of the slices is less important.
2. To convey that one segment of the total is relatively small or
large.

Our sales example supports the first use case. Suppose we want to
focus on how much revenue a given product—or set of products—
contributes to the total. Given our natural construct of how we read a
clock, it’s fairly easy to tell that Product A makes up roughly a quarter
of total revenue and that Products A, B, and C combined are roughly
two-thirds of total revenue. For these general part-to-whole
relationships, a pie chart could effectively convey that information.
The two pie charts below are examples of the second use case. With
each of these, I can easily make one broad conclusion—one slice of the
pie is relatively larger or smaller than the others.

For applications with different datasets, check out Lisa Charlotte Rost’s
examples in what to consider when creating pie charts.

When should I pass on the pie?

Generally speaking, pies should not be used for evaluating the relative
sizes of categories, comparing data across pies, and visualizing
percentages that do not sum to 100%. Let’s look at an example of
each.

Comparing the size of categories: Pie charts break down pretty

quickly if we need our audience to have a more nuanced
understanding of our data than simply big, small, or about the same. To
illustrate, consider the following example. If I asked you to compare the
size of slices C and D, which would you conclude is larger? Taking it a
step further, how easily can you estimate how much bigger?
This example demonstrates one limitation of data encoded as pie
slices: humans’ eyes aren’t well-equipped to compare areas. The
challenge becomes even more pronounced without a consistent
baseline—like a y-axis in a horizontal bar chart—against which to make
the visual comparison. As illustrated in the following, comparing the
size of C vs D is much easier in the graph on the right. We’re also
comparing one dimension (length) across two categories, instead of
irregularly-shaped areas.

Comparing data across pies: this is a difficult task, because the

things you aim to compare are separated visually and are located in
different places on each pie. To illustrate, check out the pair of pies
below, excerpted from Exercise 2.1 of Storytelling with Data: Let’s
Practice!

Let’s say the primary comparison you want to highlight is how Tier B
varies across the two breakdowns. Tier B is in different places across
the pies—as a direct result of the variation between the two datasets.
Because of this spatial separation, different positioning within the pie,
and the irregular shapes of the segments, it’s hard to accurately judge
the difference in magnitude.

Showing data that doesn’t sum to 100%: if the slices sum to

something other than 100%, it is wrong. By definition, the pie must
depict a meaningful whole.
I want to use a pie: how do I make a good one?

When using pie charts, here are a few quick design tips to make them
more effective.

Avoid 3D and exploding effects. These elements introduce

unnecessary clutter and can distort the data. This inability to make an
accurate visual comparison was validated in the aforementioned
Kosara and Skau study, where study participants misjudged the values
in exploding pie charts to a higher degree than the flat pie.
Don’t have a ton of slices. There isn’t an exact rule here, so
approach with careful consideration towards your dataset and
audience’s needs. You can improve readability by either de-
emphasizing small categories with similarity of color (as shown in the
following middle pie with grey) or by aggregating into an “All other”
category, as illustrated on the right.

Sort your data meaningfully. Check out the two pies below. On the
left, the data is sorted alphabetically (beginning with apples at zero
degrees). The right pie is sorted in descending order starting with the
largest category (pineapple). This takes into account our natural
construct of reading around a circle.

Eliminate the legend and label the data directly. Check out the
difference in the following graphs. Labeling directly, as shown on the
right, reduces the task of going back and forth between the legend and
the data. Whenever possible, aim to label your slices directly with the
value so long as there’s sufficient space to make the labels legible.

Use color strategically. When used sparingly, color is arguably the

most important decision we make to focus our audience’s attention.
Let’s assume we’d like to highlight Product A’s contribution to our sales
portfolio. Which of the following pies more effectively draws your
attention to Product A?
For another example of effectively designing a pie, check out how to
make a better pie chart.

Where can I learn more about pie charts?

Here are some related resources to further your learning:

 Practice with the exercises which graph would you choose? or

alternatives to pies
 Listen to the podcast episode it depends…
 Discuss in the community conversation alternative part-to-whole
approaches
 Read the ongoing discussion and debate about the use cases for pies:
o Alternatives to pies (SWD)
o The great pie debate (SWD)
o An updated post on pies (SWD)
o Who made that pie chart? (New York Times)
o Save the pies for dessert (Stephen Few)
o F**k it, let’s use pies (Jon Schwabish)

Now that we’ve taken a tour through the ins and outs of pie charts, be sure
to continue your journey through the rest of our “what is...?” chart series,
by browsing other common visuals like bars and lines, or explore our
comprehensive chart guide page for additional chart types.

what is a scatterplot?
May 27
Written By Alex Velez
This article is part of our back-to-basics blog series called what is…?,
where we’ll break down some common topics and questions posed to
us. We’ve covered much of the content in previous posts, so this series
allows us to bring together many disparate resources, creating a single
source for your learning. We believe it’s important to take an
occasional pulse on foundational knowledge, regardless of where you
are in your learning journey. The success of many visualizations is
dependent on a solid understanding of basic concepts. So whether
you’re learning this for the first time, reading to reinforce core
principles, or looking for resources to share with others—like our new
comprehensive chart guide—please join us as we revisit and embrace
the basics.

A scatterplot is a niche chart, but it’s one of my favorites! If you are a

statistician or work in a technical field, a scatterplot might be your go-
to graph type. However, if you don’t perform a lot of statistical
analysis, then these charts may be unfamiliar. Regardless of your
current comfort level, scatterplots are extremely useful to focus on the
relationship between two series—a scenario that is common in both
technical and non-technical fields. Let’s explore some of the basics of
scatterplots via an example; I’ll also cover tips for designing more
effective ones and discuss common variations (bubble charts,
connected scatterplots, etc.), too!
What is a scatterplot?

A scatterplot shows the relationship between two numerical variables

plotted simultaneously along both the horizontal and vertical axis.
They are common in scientific fields and often used to understand data
rather than to communicate with it. This is not to say you should never
communicate with one, but you should take explicit steps to make sure
your chart is clear to an unfamiliar audience (something you should do
with all charts!). Let’s look at a scenario where a scatterplot works
nicely to communicate a finding.

Imagine you’re an analyst in the beauty industry and your company

wants to formulate a new lip care product. Before embarking on this
endeavor, you are asked to do some research to see if there is a hole in
the market. You uncover that lip care products are polarizing between
male and female buyers, so there is an opportunity to create a new
product that bridges the gap.

You might consider showing the relationship between male and female
rating scores using a scatterplot, like the one below.

Here are a couple of things worth noting about the above chart.
  It’s exploratory in nature. A scatterplot was likely used to
uncover this finding. Many statistical software packages output
scatterplots to test the correlation between two variables.
Although the graph used in the discovery phase is not always
ideal for communicating final insights, it works in this case. I can
see both sets of rankings simultaneously and also emphasize the
hole in the market. I’ve categorized and labeled the points to
make this graph legible, but an alternative chart—one that is
more familiar—could work as well (check out this post to see
three alternatives for comparing metrics). It all depends on your
audience and specific scenario.
 Be mindful of the variable placement on the axes. In this
example, it doesn’t matter which variable is along the horizontal
or vertical axis, but that won’t always be the case. Sometimes
you’ll have both an independent and a dependent variable. An
independent variable is exactly what its name implies: it’s not
affected by the other variable. A dependent variable is likely the
thing you are trying to measure, meaning it is affected by your
independent variable. It’s common practice to place the
independent metric along the horizontal or x-axis and the
dependent variable along the y-axis. It’s not wrong to invert
these, but it might be unexpected causing an initial bit of
confusion.
 Scatterplots show dots, not lines. Scatterplots are very
similar to line charts in that they both display two numerical
values; however, scatterplots tend to focus on individual data
points (depicted with a dot) rather than aggregating multiple
points into one distinct line.

In its standard form, as seen above, scatterplots show the relationship

between two things, but it’s not uncommon to display more than two
dimensions, especially when exploring your data. We’ll look at a few
common multi-dimensional variations in a moment, but before we do,
let’s break down how to read a scatterplot.

How do you read a scatterplot?

The bulk of this post is dedicated to how you can use scatterplots for
explanatory purposes despite them being a more technical graph type.
As mentioned before, if you do use a scatterplot, be aware that you
may need to break it down for your audience, and if it's their first time
looking at one of these charts, you may need to explain how to read it.
I'll step you through the process I take when examining scatterplots as
well as link to additional resources.

1. Check the axes. One of the first things I do when reading any
graph is to scan each axis. It's especially important when reading
scatterplots as they contain multiple variables, so it's necessary
to determine which variable is along which axis. I should also
mention that not all scatterplots will display similar variables;
some might compare different metrics with varying units of
measure and scale. Understanding this early on will make it
easier to interpret the relationship once you start looking at the
data.
2. Visualize sections. In the above example, I created sections by
grouping the points into quadrants. This is helpful not only for
communicating the finding in the lip care scenario but also for
uncovering it. The next time you look at a scatterplot, ask
yourself what lines you can draw or what natural breaks and
groupings exist that will help you make sense of the comparison.
3. Identify the shape. I often take a step back from the graph, so I
can attempt to summarize the individual points into a unified
shape. Are they all moving in a single linear direction? Or maybe
it's more of an exponential curve? Do the points increase as my
eyes move along the axis? Having a description in mind helps me
uncover and explain the relationship. For instance, in this cost per
mile example, we can see that the data points form a u-shape.
This means that the cost is relatively high for both shorter and
longer uses, but as we drive an average amount, the cost is more
manageable. Keep in mind that there may not be a discernible
shape, which is a perfectly valid finding (and suggests a weak or
non-existent relationship between the variables).

I'd be remiss to not share any warnings at this point. I can still hear my
stat professors drilling this mantra into my head, "correlation does not
imply causation." Just because you uncover a relationship doesn't
mean you've identified the underlying cause. For instance, in the lip
care example, we found that brands preferred amongst men are less
preferred amongst women. Does this mean that if a male likes a brand,
a female won't? Absolutely not, or there wouldn't be an opportunity to
create a bridge product. We can use findings to inform and make
educated decisions, but to generalize the insight would be a misuse of
data. This applies to other chart types as well, however I find that
scatterplots are more susceptible to misreadings. I think it's because
they focus on the relationship, making it easy to both speculate and
extrapolate. (For more data pitfalls check out our review of Ben Jones’
latest book, Avoiding Data Pitfalls.)

If you are interested in reading more about interpreting scatterplots,

check out Lisa Charlotte Rost's post, where she takes you through her
process with a multi-dimensional scatterplot, the bubble chart.
Speaking of which, let's discuss these variations next.

What are common variations of scatterplots?

Bubble charts: Let’s imagine the lip care example includes another
dimension (such as packaging type, price, the active ingredient, etc.).
We can encode this third variable by altering the size, color, or even
the shape of the data points. This is a bubble chart. Bubble charts are
useful for showing multi-dimensional relationships, but this comes at a
cost since they are tough to read. Once we change the size of the
circles, we start encoding information by area. Our eyes are not very
good at measuring area, so specific comparisons are harder to make. I
should also mention that many graphing applications don’t offer
canned bubble chart templates. This means someone has to calculate
the correct area of each circle, which can introduce human error. Also,
as we play with color and shapes, we ask our audiences to do a lot of
mental work to process the information. People can typically only hold
about four pieces of visual information in their short term memory at
any given time, so be mindful of the cognitive effort placed on the
reader! That said, there are fantastic examples of explanatory bubble
charts, like Hans Roslin’s famous BBC video. What makes Hans'
explanation so effective is his willingness and enthusiasm to step his
audience through his animated bubble chart. If you are considering a
bubble chart—or any multi-dimensional graph for that matter—
consider your audience and how much effort you want to spend
explaining how to read your graph.

Connected scatterplots: I think of a connected scatterplot as a

hybrid between a traditional scatterplot and a line chart. There are still
two dimensions along the x- and y-axis, but a third dimension, usually
time, is layered on with lines. The challenge is that we are accustomed
to reading time from left to right, so to see it moving in every direction
from point to point can be unsettling. As with all variations, each graph
has an ideal use case, and with thoughtful annotations, labeling, and
focused attention can be made clear to any audience. To learn more,
check out Dan Zvinca’s guest post on the many roles of lines and look
at the example of Bill Rapp’s January 2018 #SWDchallenge submission
(below).
2x2 grid (conceptual scatterplot): This isn’t a multi-dimensional
view per se, but one that can be practical in a business setting. We
often want to understand relationships between two things or explain
how one decision might impact another. A conceptual scatterplot also
known as a 2x2 grid, can help your audience make sense of these
comparisons, because our visual system is much faster at processing
information than our verbal system. For example, we might use one to
explain the differences between presenting your data live verse
communicating via a written document. The two variables are the
amount of control you want to have and the amount of detail needed
to get your point across. Notice how quickly it is to grasp when one
medium is ideal over the other.

What are tips for designing scatterplots?

Here are a few formatting steps to consider when designing

scatterplots.
 Declutter by removing trendlines. During the analytical
process, you might fit a model to describe the relationship shown
in your scatterplot and display this model as a line through the
data points. My preference—and I realize this may be
controversial—is to remove the fitted line when communicating
data. It adds clutter and if the underlying trend isn’t obvious then
implanting a line might lead to contention or confusion (as seen
in this hurricane example).

 Make overlapping data points transparent. If many data

points overlap, it may become hard to see each value or the
volume of points in a particular section. One trick might be to
play with the opacity of the data markers so that the individual
data points are visible.
 You don’t need a zero baseline. Similar to line charts,
scatterplots encode data by position along the axis. This means
that it isn’t necessary for your baseline to start at zero in the
same way it is for bars. With that said, you should still be mindful
anytime you deviate from the zero baseline as this may introduce
confusion.
 Create sections and add labels for clarity. Adding
annotations and categorizing data points can make scatterplots
easier to consume. In the lip care example, notice how the pithy
descriptions not only work well in constrained spaces but they
also make this graph easier to discuss. I can refer to the
opportunity quadrant and you likely know what I mean! Without
implementing a clear construct to read the scatterplot, the graph
is highly exploratory and requires the reader to repeat the
analysis to uncover the relationship. Check out the difference in
the image below or read these two posts for additional examples
of how to label and categorize scatterplots: an example showing
reach and engagement by country and another example showing
manager performance.
Where can I see more scatterplot examples?

For additional examples, check out our SWD challenge featuring

scatterplots.

Now that we’ve taken a tour through the ins and outs of scatterplots,
make sure to continue your journey through the rest of the “what
is...?” chart series, by browsing other common visuals like bars and
pies, or explore our comprehensive chart guide page for additional
chart types.

Master complex charts in our NEW! on-demand learning

course

Learn in-depth techniques for creating, designing, and effectively

presenting more novel views in PowerPoint. Through step-by-step video
lessons, you'll master radar charts, box plots, scatterplots, horizon
charts, and bubble graphs. With optional exercises to reinforce your
learning and bonus materials covering additional graph types, this
course will elevate your data presentation skills. Enroll today!

what is a Gantt chart?

Nov 25

Written By Amy Esselman

This article is part of our back-to-basics blog series called what is…?,
where we’ll break down some common topics and questions posed to
us. We’ve covered much of the content in previous posts, so this series
allows us to bring together many disparate resources, creating a single
source for your learning. We believe it’s important to take an
occasional pulse on foundational knowledge, regardless of where you
are in your learning journey. The success of many visualizations is
dependent on a solid understanding of basic concepts. So whether
you’re learning this for the first time, reading to reinforce core
principles, or looking for resources to share with others—like our new
comprehensive chart guide—please join us as we revisit and embrace
the basics.

What is a Gantt chart?

A Gantt chart is used to visually present tasks or events over time. It

consists of a horizontal axis of time and horizontal bars representing
the duration of each item arranged along the vertical axis.

The visual was popularized in the early 1900s by American engineer

and management consultant Henry Gantt, for whom the chart is
named. Most commonly, this view is used to see the start date and
duration of each task in a larger project to better understand timelines,
milestones, and dependencies.
How to read a Gantt chart

To interpret a Gantt chart effectively, start by identifying the time scale

along the horizontal axis. Then, scan down the task list on the vertical
axis. Observe the length and position of each bar to understand the
duration and timing of each activity. Tasks are typically arranged
vertically from top to bottom, reflecting their chronological start dates
to visually represent project sequence and progression.

Any dependent tasks are usually connected by vertical lines,

illustrating how one task's completion directly impacts the start or
progression of subsequent tasks to help identify relationships and
potential bottlenecks within a project's workflow. In the example
above, Test A requires the completion of Task 2.

Take note of any milestones or key dates—usually represented by a

shape like a circle or a diamond. In the example above, the go-live
date is January 30 and there is a vertical dashed line to indicate the
current date for reference.

When could you use a Gantt chart?

Gantt charts are particularly useful during the project planning phase,
as the detailed visual schedules help managers identify when or where
there might be bottlenecks or a need for additional resources. Project
managers and teams often refer to Gantt diagrams throughout the
project lifecycle to monitor and control progress and assess risks.

Gantt-style charts can be used for more than just project scheduling.
The combination Gantt and line chart below was used to show how the
evolution of an eLearning system affected employee satisfaction over
time. The Gantt elements denoted the timing of certain events (when a
survey was deployed when a tagging initiative began, and when
courses were split), while the line chart revealed the concurrent
positive impact these events had on employee sentiment.

SOURCE: [Link]
velez/gallery/5101035f-d3df-431e-bcdb-ddda87f83c6b/images
Although commonly shown horizontally, a Gantt chart can also be
flipped vertically. In the example below, SWD Community member
Chris Smith used a vertical Gantt-style visual resume to show their
professional experience over time. Since, in the Western world, time is
most naturally read moving from left to right, take care when switching
the orientation, as it can feel uncomfortable or require a bit of mental
gymnastics to understand.

SOURCE: [Link]
2021/inspired-by-coles-book-acknowledgement-section

Limitations of Gantt charts

However useful Gantt charts are in theory, there are drawbacks to

relying on them in practice. They can become complex and cluttered
for very large projects, may require frequent updates in fast-paced
environments, and don't always capture the full complexity of task
relationships. Creating and maintaining Gantt charts manually can be
challenging and time-consuming.

While project teams often use Gantt charts, not everyone understands
how to interpret them accurately or quickly. Consider your audience's
background and potential learning curve when selecting this
visualization. Choosing a different display altogether may avoid
confusion for audiences who do not know this graph or how to interpret
the information it depicts.

To make a Gantt chart more accessible and engaging in a presentation,

use animation to build the chart gradually. First, walk your audience
through the structure to explain the timeline and tasks, then layer on
each horizontal bar in a meaningful way to communicatey our
message. By breaking down the Gantt chart's complexity and guiding
your audience through its components, you'll help them understand
the project timeline more effectively.

Tools for creating Gantt charts

Some graphing applications, like Excel or Tableau, have built-in

templates to help you create a Gantt chart. If your tool of choice lacks
this capability, you can still build one using a combination of a
horizontal bar chart to show the task duration and a dot plot to depict
milestones.

There are also software packages that focus specifically on task or

project management. Tools like Microsoft Project, Asana, Trello, and
Smartsheet not only have Gantt templates built in but also offer
additional features like collaboration and automated updates, making it
easier to create and maintain accurate charts throughout the project
lifecycle.

Best practices for designing a Gantt chart

To create effective Gantt charts, consider the following design tips:

 1. Choose an appropriate time scale (days, weeks, months)
based on the duration. Keep in mind that providing too granular
of a view could overcomplicate the visual.
2. Limit your color palette to reduce visual complexity. Use color
strategically to focus on specific aspects, like the critical path or
milestones at risk.
3. Add meaningful context, like key milestones and a current
date reference line, to enhance the chart's utility.
4. Simplify the view by removing extraneous elements, like
borders and gridlines, to make it easier to consume.

Final thoughts

In summary, Gantt charts are invaluable tools for visualizing project

tasks across a timeline. However, before using them to communicate
your project’s progress, assess your audience's familiarity with this
chart type and the specific insights you wish to convey to ensure the
visual fits the purpose.

To further enhance your understanding of various graph types, we

invite you to explore the rest of our “What Is...?” chart series. Explore
our comprehensive chart guide page for additional views, like bars and
lines.

Do you want to learn to create and communicate a powerful

data story? Join our upcoming 8-week online course: plan, create, and
deliver your data story. Data storytellers Amy and Simon will guide you
through the world of storytelling with data, teaching a repeatable
process to plan in helpful ways (articulating a clear message and
distilling critical content to support it), create effective materials
(graphs, slides, and presentations), and communicate it all in a way
that gets your audience’s attention, builds your credibility, and drives
action. Learn more and register today.