DSCI 5360

Visualization Design
Principles
DSCI 5360
Data Visualization for Analytics
Russell R. Torres
 DSCI 5360

““Design is a funny word. Some people think design

means how it looks. But of course, if you dig deeper,
it’s really how it works.”

Steve Jobs

2
 DSCI 5360

Design and Data

• Four levels of visualization design

• Tufte’s Principles
• Integrity
• Design

Edward Tufte
Professor Emeritus of Political Science, Statistics, and Computer Science at Yale
University
The above image, “Tufte during his one-day course in Dallas, May 21, 2015” by Keegan Peterzell, is
licensed under CC BY-SA 4.0. All other content herein is licensed and copyright under different terms
and by different parties.
3
 DSCI 5360

The Four Levels of Visualization Design

Design Model: Describes levels of

design inherent to and that should be
considered in the creation of a
visualization

Munzner 2009

4
 DSCI 5360

The Four Levels of Visualization Design

• Domain Situation – Describing a group of target users, their domain of interest,
their questions, and their data
• Data/Task Abstraction – Abstracting the specific domain questions and data from
the domain specific form into a generic, computational form
• Encoding/Interaction Technique – Deciding on the specific way to create and
manipulate the visual representation of the abstraction
• Algorithm – Crafting a detailed procedure that allows a computer to
automatically and efficiently carry out the desired visualization goal

5
 DSCI 5360

Edward Tufte’s Lessons – Graphical Integrity

Graphical Integrity
• Clear, detailed, and thorough
labeling should be used to defeat
graphical distortion and ambiguity

6
 DSCI 5360

What’s Wrong with this Vis? Missing Scales

$521,943
$397,747
$351,341
$3,594,385 ($11,410)

Tufte 2001
-$4,200,000 7
 DSCI 5360

What’s Wrong with this Vis? Scale Distortion

Volume

Year 8
 DSCI 5360

Edward Tufte’s Lessons – Graphical Integrity

Graphical Integrity
• Clear, detailed, and thorough
labeling should be used to defeat
graphical distortion and ambiguity
• The representation of numbers, as
physically measured on the surface
of the graphic itself, should be
directly proportional to the numeric
quantities represented

9
 DSCI 5360

The Lie Factor

𝑆𝑖𝑧𝑒 𝑜𝑓 𝑡h𝑒 𝐸𝑓𝑓𝑒𝑐𝑡 𝑆h𝑜𝑤𝑛𝑖𝑛 𝐺𝑟𝑎𝑝h𝑖𝑐

𝑇h𝑒 𝐿𝑖𝑒 𝐹𝑎𝑐𝑡𝑜𝑟 =
𝑆𝑖𝑧𝑒 𝑜𝑓 𝑡h𝑒 𝐸𝑓𝑓𝑒𝑐𝑡 𝑖𝑛 𝐷𝑎𝑡𝑎

10
 DSCI 5360

What’s Wrong with this Vis? Distortion

This line, representing 18 miles per
gallon in 1978 is 0.6 inches long.

Tufte 2001
This line, representing 27.5 miles per
gallon in 1985 is 5.3 inches long

11
 DSCI 5360

The Lie Factor

5.3 −0.6
𝐺𝑟𝑎𝑝h𝑖𝑐 𝐸𝑓𝑓𝑒𝑐𝑡 = ∗100=783 %
0.6
27.5 −18.0
𝐷𝑎𝑡𝑎 𝐸𝑓𝑓𝑒𝑐𝑡 = ∗100=53 %
18.0

783
𝑇h𝑒 𝐿𝑖𝑒 𝐹𝑎𝑐𝑡𝑜𝑟 = =14.8
53
12
 DSCI 5360

A Better Approach

Tufte 2001 13
 DSCI 5360

Edward Tufte’s Lessons – Graphical Integrity

Graphical Integrity
• Clear, detailed, and thorough
labeling should be used to defeat
graphical distortion and ambiguity
• The representation of numbers, as
physically measured on the surface
of the graphic itself, should be
Distractions
directly proportional to the numeric
quantities represented
• Show data variation, not design
variation

14
 DSCI 5360

What’s Wrong with this Vis? Unintended Size Coding

Bad Bar Chart Practices, or Send in the Clowns

15
 DSCI 5360

A Better Approach

Bad Bar Chart Practices, or Send in the Clowns

16
 DSCI 5360

Edward Tufte’s Lessons – Graphical Excellence

Graphical Excellence
• Maximize the data ink ratio

17
 DSCI 5360

The Data-Ink Ratio

𝐷𝑎𝑡𝑎 𝐼𝑛𝑘
𝑇h𝑒 𝐷𝑎𝑡𝑎 𝐼𝑛𝑘 𝑅𝑎𝑡𝑖𝑜=
𝑇𝑜𝑡𝑎𝑙 𝐼𝑛𝑘

18
 DSCI 5360

What’s Wrong with this Vis? Distraction

19
 DSCI 5360

A Better Approach

20
 DSCI 5360

The Data-Ink Ratio

Tufte 2001

21
 DSCI 5360

Counter-Point

22
 DSCI 5360

Counter-Point

• Experiment that asked participants

to choose boxplot with largest range
from a set
• Captured EEG brain waves to
measure cognitive load during
evaluation
• Results find that the simplest plot is
the hardest to interpret

23
 DSCI 5360

Edward Tufte’s Lessons – Graphical Excellence

Graphical Excellence
• Maximize the data to ink ratio
• Avoid Chartjunk

Tufte 2001

24
 DSCI 5360

What’s Wrong with this Vis? Distraction

25
 n Even DSCI 5360

A Better Approach

26
 DSCI 5360

The Work of Nigel Holmes

27
 DSCI 5360

Counter-Point

28
 DSCI 5360

Counter-Point

• Experiment to determine whether

visual embellishments cause
comprehension problems and whether
they provide valuable information
• No differences for interpretation
accuracy or recall accuracy after 5
minutes
• Better recall accuracy for Holmes
charts after 2-3 weeks
• Participants saw value messages in
Holmes charts and found them more
attractive, enjoyable, and
memorable.
29
 DSCI 5360

Chartjunk

So What Do We Make of This? What Do We Do About Chartjunk

• Intuitive findings – color and • Chartjunk may aid
human recognizable object enhance • Persuasion
memorability • Memorability
• Non-intuitive findings – common • Engagement
graphs are less memorable that • But may detract from
unique visualization types • Unbiased analysis
• Trustworthiness
• Interpretability
• Space efficiency

30
 DSCI 5360

Layering and Separation

31
 DSCI 5360

Layering and Separation

Tufte 2001

32
 DSCI 5360

Data Density

𝑁𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝐸𝑛𝑡𝑟𝑖𝑒𝑠 𝑖𝑛 𝐷𝑎𝑡𝑎 𝐴𝑟𝑟𝑎𝑦

𝐷𝑎𝑡𝑎 𝐷𝑒𝑛𝑠𝑖𝑡𝑦=
𝐴𝑟𝑒𝑎 𝑜𝑓 𝐷𝑎𝑡𝑎 𝐺𝑟𝑎𝑝h𝑖𝑐

33
 DSCI 5360

Shrinking Graphics and Small Multiples

• Shrinking graphics can allow you to

present general trends among a large
number of factors
• Doing so allows you to maximize data
density while retaining interpretability
(at least from a comparative point of
view)
• Sparklines, (miniature line charts), small
multiples (panels of repeated graphs),
and other small format approaches may
be useful in some circumstances where
relative understanding is more important
than specifics
Meyer et al. 2010
34
 DSCI 5360

Small Multiples and High Data Density

The Double Think: Tufte’s Principles

35
 DSCI 5360

Counter-Point

36
 DSCI 5360

Counter-Point
• We overestimate breadth
• Belief that viewers can take in all (or most) of the details of a scene at once
• Adding extra visual features makes it harder to find specific bits of information
• We overestimate countenance
• Belief that user will attend to a higher proportion of the display than they do
• User typically have expectations about where in a display to look
• We overestimate depth
• Belief that attending to an object leads to more complete and deep understanding than is the
case

37
 DSCI 5360

A Summary of Tufte’s Design Principles

• Maximize the data-ink ratio
• Avoid chart junk (sometimes)
• Use multifunctioning elements
• Layer information
• Maximize data density
• Shrink graphics
• Maximize the amount of data shown (sometimes)

38
 DSCI 5360

References
Content adapted from Visualization Analysis and Design, Tamara Munzner, CRC
Press 2015
http://www.sci.utah.edu/~miriah/cs6630/lectures/L02-design.pdf

39