Designing Great Visualizations

1
Designing Great Visualizations

• Jock D. Mackinlay
• Director, Visual Analysis, Tableau Software

2
Outline

• Examples from the history of visualization
• Computer-based visualization has deep roots
• Human perception is a fundamental skill
• Lessons for designing great visualizations
• Human perception is powerful
• Human perception has limits
• Use composition and interactivity to extend
  beyond these limits
• Finally, great designs tell stories with data
• Image sources
• www.math.yorku.ca/SCS/Gallery
• www.henry-davis.com/MAPS

3
Visual Representations are Ancient

• 6200 BC Wall image found in Catal Hyük, Turkey
• Painting or map?

4
Two Common Visual Representations of Data

• Presentations Using vision to communicate
• Two roles presenter audience
• Experience persuasive
• Visualizations Using vision to think
• Single role question answering
• Experience active

1999 Morgan Kaufmann
5
Maps as Presentation

• 1500 BC Clay tablet from Nippur, Babylonia
• Evidence suggests it is to scale
• Perhaps plan to repair city defenses

6
Maps as Visualization

• 1569 Mercator projection
• Straight line shows direction

7
William Playfair Abstract Data Presentation

• 1786 The Commercial and Political Atlas (Book)
• 1801 Pie chart

8
Dr. John Snow Statistical Map Visualization

• 1855 London Cholera Epidemic
• It is also a presentation

Broad StreetPump
9
Charles Minard Napoleons March

• 1869 Perhaps the most famous data presentation

10
Darrell Huff Trust

• 1955 How to Lie With Statistics (Book)
• Trust is a central design issue
• Savvy people will always question data views
• Does a data view include the origin?
• Is the aspect ratio appropriate?

11
Jacques Bertin Semiology of Graphics (Book)

• 1967 Graphical vocabulary
• Marks
• Points
• Lines
• Areas
• Position
• Statistical mapping

• Retinal
• Color
• Size
• Shape
• Gray
• Orientation
• Texture

12
Jacques Bertin (continued)

• Visual analysis by sorting visual tables
• Technology

13
Jock Mackinlay Automatic Presentation

• 1986 PhD Dissertation, Stanford
• Extended and automated Bertins semiology
• APT A Presentation Tool

14
Scientific Visualization

• 1986 NSF panel and congressional support

15
Richard Becker William Cleveland

• 1987 Interactive brushing

Related marks
Selection
16
Information Visualization

• 1989 Stuart Card, George Robertson, Jock
  Mackinlay
• Abstract data
• 2D 3D interactive graphics
• 1991 Perspective Wall Cone Tree

17
Book Readings in Information Visualization

• 1999 Over a decade of research
• Card, Mackinlay, Shneiderman
• An established process of visual analysis
• Involves both data and view
• Interactive and exploratory

18
Chris Stolte

• 2003 PhD Dissertation, Stanford
• Extended the semiology from Bertin Mackinlay
• VizQL connected visualizations to databases
• Accessible drag-and-drop interface

VizQL
View
Query
Data Interpreter
Visual Interpreter
19
Visual Analysis for Everyone

• 2008 Tableau Customer Conference

20
Human Perception is Powerful

• How many 9s?

21
Human Perception is Powerful

• Preattentive perception

22
Traditional Use Negative Values

• However, mental math is slow

23
Cleveland McGill Quantitative Perception
More accurate
Position
Length
Angle
Slope
Area
Volume
Color
Density
Less accurate
24
Exploiting Human Perception
25
Bertins Three Levels of Reading

• Elementary single value
• Intermediate relationships between values
• Global relationships of the whole

26
Global Reading Scatter View

• Bertin image A relationship you can see during
  an instant of perception

27
Effectiveness Depends on the Data Type

• Data type
• Nominal Eagle, Jay, Hawk
• Ordinal Monday, Tuesday, Wednesday,
• Quantitative 2.4, 5.98, 10.1,
• Area
• Nominal Conveys ordering
• Ordinal
• Quantitative
• Color
• Nominal
• Ordinal
• Quantitative

28
Ranking of Tableau Encodings by Data Type
Nominal Position Shape Color hue Gray ramp Color
ramp Length Angle Area

• Quantitative
• Position
• Length
• Angle
• Area
• Gray ramp
• Color ramp
• Color hue
• Shape

Ordinal Position Gray ramp Color ramp Color
hue Length Angle Area Shape
29
Human Perception is Limited

• Bertins synoptic of data views
• 1, 2, 3, n data dimensions
• The axes of data views
• ? Reorderable
• O Ordered
• T Topographic
• Network views
• Impassible barrier
• Below are Bertins images
• Above requires
• Composition
• Interactivity
• First a comment about 3D

30
3D Graphics Does Not Break the Barrier

• Only adds a single dimension
• Creates occlusions
• Adds orientation complexities
• Easy to get lost
• Suggests a physical metaphor

31
Composition Minards March

• Two images

32
Composition Small Multiples
33
Composition Dashboards
34
Interactivity Bertins Sorting of Data Views
35
Interactivity Too Much Data Scenario
36
Interactivity Aggregation
37
Interactivity Filtering
38
Interactivity Brushing
39
Interactivity Links
40
Telling Stories With Data

• What are the good school districts in the Seattle
  area?
• Detailed reading
• One school or school district at a time

41
Telling Stories With Data (continued)

• I needed a statistical map

42
Telling Stories With Data (continued)

• Positive trend views online
• Easy to see that the district is stronger than
  the state
• Harder to see that reading is stronger than math
• Found the source data, which is a good thing
  about public agencies

43
Telling Stories With Data (continued)

• Reading is clearly better than math

44
Telling stories with data (continued)

• Moral Always Question Data

45
Telling Effective Stories

• Trust a key design issue
• Expressive convey the data accurately
• Effective exploit human perception
• Use the graphical vocabulary appropriately
• Utilize white space
• Avoid extraneous material
• Context Titles, captions, units, annotations,

46
Stories Involve More Than Data

• Aesthetics What is effective is often affective
• Style Include information about who you are
• Playful Allow people to interact with the data
  views
• Vivid Make data views memorable

47
Summary

• Visualization presentation
• Human perception is powerful limited
• Coping with Bertins barrier
• Composition
• Interactivity
• Sorting
• Filtering
• Aggregation
• Brushing
• Linking
• Telling stories with data
• Trust is a key design issue
• Always question data

48
Resources

• My email jmackinlay_at_tableausoftware.com
• Edward Tufte (www.edwardtufte.com)
• The Visual Display of Quantitative Information
• Beautiful Evidence
• Jacques Bertin
• Semiology of Graphics, University of Wisconsin
  Press
• Graphics and Graphic Information Processing,
  deGruyter
• Colin Ware on human perception visualization
• Information Visualization, Morgan Kaufmann
• William S Cleveland
• The Elements of Graphic Data, Hobart Press