June 2020

Fundamentals of
Visualization of
Data
and why you should avoid
graphics with food names
WHAT
YOU GO
FIND
HERE

#DATAVIZ
This presentation is an introduction to science and art.
from data visualization, which will feature...

Less is more, minimalism in communication

Do not use pie charts
Monoaxialism, or the sin of the double axis
Conscious use of colors and palettes
Most common types of charts
And more...

Digital Transformation Suzano | Achieve the Unthinkable

Who we are
we

Edgar Italo
Data Scientist Researcher
Digital Forest Technology

Digital Transformation Suzano | Achieve the Unthinkable

Art +
Science
Data visualization is part art,
part science
The challenge of communicating information through graphs is
common in the daily routine of most companies.
The biggest challenge is to convey information clearly and without
distortions for your target audience

Delete
Choose one noises Focus
effective visual attention
Understand the
context Tell a
Digital Transformation Suzano | Achieve the Unthinkable
history
 Less is more:
The focus on the message

Digital Transformation Suzano | Achieve the Unthinkable

Less
it is more
In our daily lives, it is common to find
graphs like this

At first glance, it doesn't seem to have

no problem.

Can you point out what the

problems of this graph?

Source: Storytelling with Data (available at[Link]

Less
it is more
In our daily life, it is common to find
graphs like this
Some questions:

Redundancy on the y-axis (label + grid)

What do the colors mean?

What message am I trying to send?

to pass?

Source: Storytelling with Data (available in[Link]

Less
it's more
See the same information with
another guise
Some positive changes:

- O título explica qual o objetivo

from that visualization;

Lines instead of bars highlight the

growth of the difference between
tickets processed and received;

The information is highlighted

dismissal of two employees in
May as the cause of the problem.

No one needs to introduce this

graph
Source: Storytelling with Data (available at[Link]
Less
it's more
Another example that can
to be improved
Let's exercise:

What would you change about this

graph?

What is missing?

What is the message of this?

visualization?

Source: Storytelling with Data (available at[Link]

Less
it's more
Our improvement suggestion highlights
the success of the pilot
Now we know what it is about
graph, research was conducted beforehand
and after a pilot program of
stimulus to science.

The colors were used to emphasize the

positive results. Supporting texts are
essential for conveying a message
cohesive.

Source: Storytelling with Data (available at[Link]

How to pass
a good
A good visualization should Each element must have
message?
to have company a a reason to exist.
good story. Don't forget Legends, colors, title. Use
what your chart should have with wisdom.
a reason to exist

SOME GUIDES:

Less is more Who is the audience? Know your message

Very care with Probably the level of Your view has to
exaggerations. Redundancy, detail for a to have a reason to exist.
excess of colors and excess executive and for a team Know which a exact
information is a shot the technician will be different. message of each graph
at the heart of good communication Adapt your content to what you put. If not
your audience there is a message, maybe it
should not exist

Digital Transformation Suzano | Achieve the Unthinkable

 Do not use graphs
of pizza

Digital Transformation Suzano | Achieve the Unthinkable

Interpreting
areas

Humans are naturally bad at

interpret areas
Look at the graph on the side.

Try to sort the parts by size.

Source: Fundamentals of Data Visualization–Claus Wilke (available at[Link]

Comparing
graphics

Not convinced?
Now compare the three graphs below.
Try to understand the evolution of values among the three.

Source: Fundamentals of Data Visualization - Claus Wilke (available at[Link]

Comparing
graphics

What if we change it to a bar graph?

Now compare the three graphs below.
Easier, right.

Source: Fundamentals of Data Visualization – Claus Wilke (available at[Link]

 Alternatives to pizza

Digital Transformation Suzano | Achieve the Unthinkable

Alternatives
to the pizza

Let's start with an example:

9%

9% Besides the areas, what do you see as

problems here in this graph?

9%

54%

19%

Quick Win Moonshot EY Nice to Have Park it

Digital Transformation Suzano | Reach the Unthinkable

Alternatives
to the pizza

A bar chart allows

direct comparison of heights and elimination
60%
54%
of excess colors
50%

40%

30%

19%
20%

9% 9% 9%
10%

0%
Quick Win Moonshot EY Nice to Have Park it

Digital Transformation Suzano | Achieve the Unthinkable

Alternatives
the pizza

Data labels and background lines are

redundant. It is possible to choose just one

54% 60%

50%

40%

30%
19%
20%
9% 9% 9%
10%

0%
Quick Win Moonshot EY Nice to Have Park it Quick Win Moonshot EY Nice to Have Park it

Digital Transformation Suzano | Reach the Unthinkable

Alternatives
the pizza

It is possible to use colors to highlight the

most important data (if any)
60%

50%

40%

30%

19%
20%

10%

0%
Quick Win Moonshot EY Nice to Have Park it
Digital Transformation Suzano | Achieve the Unthinkable
Alternatives
to pizza Bar charts are not always useful
Bar chart does not allow to see the
data as parts of a whole
54%

19%

9% 9% 9%

Quick Win Moonshot EY Nice to Have Park it

Digital Transformation Suzano | Achieve the Unthinkable

Alternatives
the pizza
Bar charts are not always useful.
Another option is a bar chart
stacked

Moonshot
Quick Win EY Nice to Have Park it
19%

0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%

Digital Transformation Suzano | Achieve the Unthinkable

Alternatives
the pizza
Bar charts are not always useful
To see proportions, another
the alternative is the treemap

Digital Transformation Suzano | Achieve the Unthinkable

 Good examples (can be used
pizza–in moderation)

Digital Transformation Suzano | Achieve the Unthinkable

Others
Examples
Starting with a
bad example
Others
Examples
Two significant parts
Cores used appropriately
Support text
Others
Examples
Division of chairs of
German parliament (1961-63)
Shows the SPD+FDP coalition as being the majority
Focus on parts of a whole with simple analysis
Use of colors refers to the German flag
Others
Examples
Comic purposes
The sin of
double axis

Digital Transformation Suzano| Achieve the unthinkable

Axis
Duplo Using a second axis seems
useful, but equally treacherous
Changing the scales, the
silent conclusion
drastically

Source: Why not to use two axes, and what to use instead - Lisa Charlotte (available at
[Link]
Axis
Duplo Using a second axis seems
useful, but it is equally treacherous

Changing the scales, the

silent conclusion
drastically

•Induces a
correlation...
probably
nonexistent

Source: Why not to use two axes, and what to use instead - Lisa Charlotte (available at
[Link]
Axis
Duplo Using a second axis seems
useful, but equally treacherous
Changing the scales, the
silent conclusion
drastically

Induces a
correlation...
probably
nonexistent

•Axes of different
magnitudes will go
to confuse the reader

Source: Why not to use two axes, and what to use instead - Lisa Charlotte (available at
[Link]
Axis
Duplo
Changing the scale and range of the axes to
interpretation changes completely!

Orange remained stable, Orange increased, Both started from the same level, but
Blue has increased a lot Blue remained Orange increased much more than Blue

Both started from the same level, Both grew up, but Azul had Both remained stable.
but Azul increased a lot more result greater than Orange
what Orange
Source: Why not to use two axes, and what to use instead – Lisa Charlotte (available at
[Link]
Axis
Duplo
Research shows that using an axis
it doubles the worsening of interpretability

Conclusion:

(...) our results discourage the use of the graph with

double axis, as the participants had a worse performance
with this graph and also classified it as the lowest
in terms of subjective preference.

Source: A Study on Dual-Scale Data Charts–Isenberg et al., 2011. (available at[Link]

Axis
Duplo
To avoid the double axis,
we can
Separate into two graphs and clearly indicate the magnitude of each.
case

Source: Why not to use two axes, and what to use instead - Lisa Charlotte (available at
[Link]
Axis
Duplo
To avoid the double axis, we can:
Index the values and show the relative change.
ignoring the absolute values

Source: Why not to use two axes, and what to use instead – Lisa Charlotte (available at
[Link]
Axis
Duplo
To avoid the double axis, we can:
Show the main series and only highlight
comments related to the second series

Source: Why not to use two axes, and what to use instead - Lisa Charlotte (available at
[Link]
 Conscious use of colors and
popsicles

Digital Transformation Suzano | Reach the unthinkable

Hearts
Good taste and empathy above all
Cores

Let's conceptualize the 'types' of

colors/palettes and the 'purpose' of use

Types: Purpose:
Sequential Distinguish
Divergent Represent
Qualitative Highlight
Cores

What is the nature of your data?

Sequential Palettes
Ordered data that
progress from low to high
Only one of the extremes
deserves highlight
Use in cases: 'the greater the
value, better
Source: COLOR BREWER 2.0 - Brewer et al. (available at[Link]
Hearts

What is the nature of your data?

Divergent Palettes
The two extremes deserve
highlight, with the same
emphasis
•Use in cases: 'bad
good
Source: COLOR BREWER 2.0 - Brewer et al. (available at[Link]
Cores

What is the nature of your data?

Qualitative Palettes
•Does not induce order nor
greatness
Distinction of attributes
Use in cases: 'n
different groups
Source: COLOR BREWER 2.0 - Brewer et al. (available at[Link]
Cores

What is the purpose of the graph?

Distinguish groups
Use qualitative palettes
Be careful with the quantity,
there comes a time when it just doesn't work

distinguish between a group of

other
Groups can have colors
themes

Source: Fundamentals of Data Visualization - Claus Wilke (Adapted from[Link]

Hearts

What is the purpose of the graph?

Represent values
Use sequential and divergent palettes
Choose the shade according to the
intention of the graph (e.g. blue = good,
vermelho = ruim)
Be careful with outlier values
The legend can be continuous or in
classes

Source: Fundamentals of Data Visualization - Claus Wilke (Adapted from[Link]

Cores

What is the purpose of the graph?

Tool to highlight
Use qualitative palettes about
neutral colors
Use highlight color on
transparency or neutral colors
Be careful not to abuse it
highlight

Source: Fundamentals of Data Visualization - Claus Wilke (Adapted from[Link]

Hearts

Tips to keep in your heart

Evaluate if it is better to put

the focus on the color scale or
one of the axes

If there are many groups,

avoid the color legend

Source: What to consider when choosing colors for data visualization - Lisa Charlotte (available at
[Link]
Cores

Tips to keep in the heart

Maintain consistency in
choice of colors for the
variables and groups

Less is more, remember?

Highlight what is important
and make another graph for
show the general context, if
necessary
Source: What to consider when choosing colors for data visualization - Lisa Charlotte (available at
[Link]
Cores

Tips to keep in the heart

Intuitive cores help in

interpretation and association

•Do not use palettes with

gradient to represent
groups

Source: What to consider when choosing colors for data visualization – Lisa Charlotte (available at
[Link]
Hearts
Remember that the little friend
can be colorblind
8% of men and 0.5% of the
women are colorblind

Source: How to Use Color Blind Friendly Palettes to Make Your Charts Accessible – Rachel Cravit (available at[Link]
friendly-palette/)
Cores
Color blindness occurs in
various forms

Source: How to Use Color Blind Friendly Palettes to Make Your Charts Accessible – Rachel Cravit (available at[Link]
friendly-palette/)
Hearts

Prefer palettes
colorblind safe!
Avoid these color combinations

Red & green

Green & brown
Green & Blue
Blue & gray
Blue & purple
Green & gray
Green & black

Source: How to Use Color Blind Friendly Palettes to Make Your Charts Accessible - Rachel Cravit (available at[Link]
friendly-palette/)
Cores

Prefer palettes
colorblind safe!

Source: How to Use Color Blind Friendly Palettes to Make Your Charts Accessible – Rachel Cravit (available at[Link]
friendly-palette/)
Cores

Prefer palettes
colorblind safe!

Source: How to Use Color Blind Friendly Palettes to Make Your Charts Accessible – Rachel Cravit (available at[Link]
friendly-palette/)
Cores

An example of palette
safe is the 'viridis'

Source: How to Use Color Blind Friendly Palettes to Make Your Charts Accessible – Rachel Cravit (available at[Link]
friendly-palette/)
Hearts

An example of a palette
Deuteranopia safe is the 'viridis'

Source: How to Use Color Blind Friendly Palettes to Make Your Charts Accessible – Rachel Cravit (available at[Link]
friendly-palette/)
Cores

An example of a palette
Red Blind - Protanopia safe is the 'viridis'

Source: How to Use Color Blind Friendly Palettes to Make Your Charts Accessible – Rachel Cravit (available at[Link]
friendly-palette/)
Cores

An example of a palette
Blue Blind - Tritanopia safe is the 'viridis'

Source: How to Use Color Blind Friendly Palettes to Make Your Charts Accessible – Rachel Cravit (available at[Link]
friendly-palette/)
 Most common types of
graphs (and where to use them)

Digital Transformation Suzano | Achieve the Unthinkable

EACH
GRAPH
IN HIS
PLACE
Types of graphs
Each chart has its strengths and
weaknesses, know how to use them to your advantage

There are large groups of views, knowing what you want

communicating is the most important
Some types of visualizations are:
Quantities
Distributions
Proportions
Relations x-y
Geospatial data
Uncertainty
Digital Transformation Suzano | Achieve the Unthinkable
Types of
graphics
Visualizing quantities
Bars are the most common way to
represent quantities.

On the left are the ways more

common to represent when there is
a class. Below are the ways to
represent more than one class of
quantities.

Source: Fundamentals of Data Visualization – Claus Wilke (available at[Link]

Types of
graphics
Visualizing Distributions
Histograms and density graphs are
the most intuitive way to visualize
distributions, but both have
parameters that can be misleading.

Below are ways to visualize several

classes: boxplots and their variations
(violins, sine plots) are quite useful
to visualize general differences between the
distributions.

Source: Fundamentals of Data Visualization – Claus Wilke (available at[Link]

Types of
graphs
Visualizing Proportions
Please do not use pie charts
(I think you already understood).

Pie charts and stacked bars

emphasize parts of a whole while
bars facilitate comparison between
individual parts.

When proportions are given accordingly

with multiple groupings arise the
mosaic chart, the treemap and the
parallel sets.

Source: Fundamentals of Data Visualization – Claus Wilke (available at[Link]

Types of
graphics
Visualizing Relations x-y

Next to it, various ways to show

relationships between two variables. The most
well-known and used is the graph of
scatter (scatterplot) but, besides it,
there are several options, applicable
as the amount of data changes
presented and the relationship that is desired
emphasize.

Source: Fundamentals of Data Visualization - Claus Wilke (available at[Link]

Types of
graphics
Visualizing Spatial Data

There are ways to highlight data

spatial elements that go beyond the simple map.
By highlighting with colors, we have the
choropleth, or distorting areas or
simplify them in a square we have the
cartogram.

Source: Fundamentals of Data Visualization – Claus Wilke (available at[Link]

Types of
graphics
Visualizing uncertainty

Error bars indicate the variation

expected of some estimate or
measurement.

For more focused visualizations on

uncertainty, we have the options of the second
line, where we can see the
distribution of probabilities.

For line graphs, the equivalent of

error bar is the "confidence band."

Source: Fundamentals of Data Visualization – Claus Wilke (available at[Link]

Interpreting
areas
In case of doubt, a site groups the
various options
The site From Data to Viz (data-to-
[Link] helps you explore the main
view options according to
the available data.

It's great for having new ideas on how to

present your data (with tutorials in
R, Python eFlourish

Source: From Data to Viz (available at[Link]

Interpreting
areas
Almost all graphs can be
made in Excel, but there are more options
If you program, R and Python have
extremely robust libraries for
generate all kinds of graphs, like ggplot2
And seaborn.

For those who do not write in any

language, PowerBI is a good option. Besides
Yes, it is possible to use Flourish, a
free online tool for generation
of beautiful graphs.

Source: Flourish Examples (available at[Link]

 Bonus:
show of horrors

Digital Transformation Suzano | Achieve the Unthinkable

 Horror
:C
The best-selling combo

Pizza + Sliced + 3D + Soma > 100%

Park it 9% Park it 9% Nice to Park it
Nice to Park it
Nice to Nice to Have 9% 9% Have 9% 9%
Have 9% Have 9%
EY 9% EY 19%
Quick Quick
EY 9% EY 9% Win 54% Win 54%
Moonsh Moonsh
Quick Win Quick Win at 29%
at 19%
54% 54%
Moonshot
19%
Moonshot
19%
Worse

Donut chart
flawed version of the pie chart

Unlike pizza, it is not made

in area, but in segment of
arch (not that it helps)

Digital Transformation Suzano | Reach the Unthinkable

Others
Examples
Spaghetti chart
Completing the food triad

Digital Transformation Suzano | Achieve the Unthinkable

Special
America
Captivating the audience
Sad

Hurricane chart
God of heaven
Visualization
of data is
science!

#REFERENCES & USEFUL LINKS

it's definitely worth checking out!
From Data to Viz–website:[Link]

Fundamentals of Data Visualization–Book:[Link]

Storytelling with Data - Book and Blog:[Link]

The worst chart in the world–Business Insider:[Link]

Color Brewer 2.0 - website:[Link]

Datawrapper - Blog:[Link]

Venngage - Blog:[Link]

Digital Transformation Suzano | Achieve the Unimaginable

THANK YOU!
#acompanystolearn
digital@[Link]
 To reinforce:
Please do not make 3D graphs

Digital Transformation Suzano | Achieve the Unthinkable