package radiantadvisors visual design series
DESCRIPTION
http://radiantadvisors.com/wp-content/uploads/2015/11/Package-RadiantAdvisors_Visual-Design-Series.pdfTRANSCRIPT
© 2015 Radiant Advisors. All Rights Reserved.
It’s a cognitive truth: our visual system is enormously influential in
our brains. As an information learning and processing mechanism,
our brains are our best tool in decoding and making sense of
information presented visually. The best data visualizations are
designed to properly take advantage of “pre-attentive features”
– a limited set of visual properties that are detected very rapidly
(typically 200 to 250 milliseconds) and accurately by our visual
system, and are not constrained by display size.
Through pattern recognition, color recognition, and counting,
data visualization exploits elements of our brain’s intrinsic
horsepower to help us better see and understand data. With self-
service and user-oriented data visualization tools, we can leverage
our natural hardwiring to layer visual intuition on top of cognitive
understanding to interact with, learn from, and reach new insights
from our data. We can also enter a visual dialogue with our data
to find new answers and ask new questions.
This first brief in a four-part series will take a high-level look at:
• A distilled introduction to the science of data visualization
• Key cognitive ingredients to have a visual dialogue with data
• How to curate meaning in the data through visual cues
DATA VISUALIZATION IS A SCIENCE
The brain is a remarkable organism. Even with recent advances in
cognitive science and perceptual psychology, our understanding
of the brain is still somewhat primitive. There are many questions
on how our brains work that we still have yet to answer – and
even more questions that we have yet to figure out even how to
ask. However, after much research we do know that the brain is
a highly visual mechanism. We know that areas like Wernicke’s
Area and Broca’s Area are designed for the comprehension and
processing of language, and we know that the Visual Cortex – the
bi-hemispherical part of the brain responsible for processing visual
information – lights up when presented with colors and shapes.
Being able to see and understand data requires more than simply
drawing up a collection of graphs, charts, and dashboards. Data
visualization is a creative process, and we can learn to enrich it
by leveraging years of research on how to design for cognition
and perception. Think of successful data visualization from a visual
science perspective, and consider the careful balance of the art
of visual design and curation alongside the observations and
insights of data science. The most meaningful data visualizations
will be the ones which correctly present complex information in a
way that is visually meaningful, memorable, and actionable.
Data visualization should work to establish visual dialogue –
to leverage our cognitive visual hardwiring and the power of
perception to have a “conversation” with the data to glean new
information in salient, memorable, and lasting ways. The data
visualization is the tangible byproduct of when art and science
come together to facilitate a visual discussion of data.
KEY COGNITIVE ELEMENTS IN VISUALIZATION
Any work of art relies on core visual principles and elements. Three
key building blocks of visual analysis are pattern recognition, color
use, and counting. They are interrelated and can be integrated to
create meaning visually from data.
Patterns and Organization
The way we perceive patterns is one of our most interesting
cognitive functions. Patterns – the repetition of shapes, forms, or
textures – are a way of presenting information to help our brains
discriminate what is important from what is not. There are patterns
around us every day that we may not even recognize – for example,
the way television show credits list actors in a series (generally the
top star first and the second last, making the first and final data
points in the pattern the most significant). Patterns are how our
brains save time decoding visual information: by grouping similar
objects and separating them. The Gestalt principles of design
emphasize simplicity in shape, color, and proximity and look for
continuation, closure, and figure-ground principles. The German
word gestalt translates to “shape form,” or pattern.
I N S I G H T S E R I E S
THE SCIENCE OF DATA VISUALIZATION
© 2015 Radiant Advisors. All Rights Reserved.
When we look at any data visualization, one of the first things that
the brain does is look for patterns. We discriminate background
from foreground to establish visual boundaries. Then we look to
see what data points are connected and how (otherwise known as
perceptual organization) – whether it is through categorical cues
like dots, lines, or clusters, or through other ordinal visual cues
like color, shapes, and lines. Typically, there are five core ways to
apply pattern recognition:
Proximity – Objects that are grouped together or located close
to each other tend to be perceived as natural groups that share
an underlying logic. Clustered bar graphs and scatter plots utilize
this principle.
Similarity – This principle extends proximity to also include items
that are identical (or close). This gives the brain two different
levels of grouping: by the shared, common nature of objects, as
well as how close they are. Geospatial and other types of location
graphics utilize this principle.
Continuity – It is easier to perceive the shape of an object as part
of a whole when it is visualized as smooth and rounded – in curves
– rather than angular and sharp. Arc diagrams, treemaps, and
other radial layouts use this principle.
Closure – Viewers are better able to identify groups through the
establishment of crisp, clear boundaries that help isolate items
and minimize the opportunity for error (even if items are of the
same size, shape, or color). This effect would be applied, for
example, in a clustered bar chart to add additional organization
to the pattern by alternating shading the area behind groups of
bars to establish boundaries.
Patterns – These help to establish clear visual organization,
composition, and layout. Once we can see patterns in information,
we can next layer visual intuition on top of cognitive understanding
to come to new conclusions. This is where color and counting
come in.
Color Use
Colors and shapes play a large part in patterns. Color (or the lack
of) differentiates and defines lines, shapes, forms, and space.
However, the use of color in design is very subjective, and color
theory is a science on its own. Colorists study how colors affect
different people, individually or in a group, and how these affects
can change across genders, cultures, those with color blindness,
and so on. There are also many color nuances, including overuse,
misuse, simultaneous and successive color contrast, distinctions
between how to use different color hues versus levels of
saturations, and so on. For this discussion, let’s focus simply on
when and how to use color in visualization to achieve unity.
In The Functional Art, Cairo writes, “The best way to disorient
your readers is to fill your graphic with objects colored in pure
accent tones.” This is because pure colors – those vibrant “hues
of summer” that have no white, black, or gray to distort their
vibrancy – are uncommon in nature, so they should be limited
to highlight important elements of graphics. Subdued hues
– like gray, light blue, and green – are the best candidates for
everything else. Most colorists recommend limiting the number
of colors (and fonts and other typography) to no more than two
or three to create a sense of unity in a visual composition. Unity is
created when patterns, colors, and shapes are in balance.
When thinking about color use in your data visualization, focus on
how you are applying your color efforts as visual targets:
Perceptual Pop-Out is the use of color as a visual beacon or target
to pre-attentively detect items of importance within visualization.
The shape, size, or color of the item here is less important than its
ability to “pop out” of a display. Consider a visit to the eye doctor,
when your vision is tested by the ability to spot a flash of color in
a sea of darkness.
Conjunction Target is the inefficient combination of color and
shape. Rather than giving target feature one visual property,
P2/3
© 2015 Radiant Advisors. All Rights Reserved.
conjunction targets mix color and shape. That distracts and
causes visual interference, making visual analysis and other
cognitive processes slower and more difficult. Thus, it is prudent
to maximize cues like perceptual pop-out, while not inadvertently
interrupting the pre-attentive process with conjunction targets.
As illustration, the figure
below is inspired by a picture
made by visualization guru
Stephen Few. It is much
harder to see the number
6 in sequences without the
benefit of shading. Likewise,
we can take advantage
of perceptual pop-out by
adding an additional color
element into the picture. This example shows how elements of
pattern recognition, color, and shape can be used together while
avoiding the clutter of conjunction targets.
Counting and Numerosity
Color and counting work in tandem, as do counting and patterns.
Relatively new research shows that the brain has an ordered
mapping, or topographical map, for number sense, similar to
what we have for visual sense and other pre-attentive features.
There are two relevant counting conversations within the scope
of data visualization. First is how data visualizations try to
reduce counting by clustering or similar approaches designed
to replace similar data objects with an alternative, smaller data
representation. Histograms take this approach. Visual spacing on
linear scales versus logarithmic scales is another example.
Second is numerosity – an almost instantaneous numerical
intuition pattern that allows us to “see” an amount (number)
without actually counting it. This varies among individuals (people
with extreme numerosity abilities are known as “savants”).
Numerosity itself is not an indicator of mathematical ability. For
most of us, numerosity gives us the ability to visually “count”
somewhere between two and ten items. We can further enhance
numerosity with visual elements like color. As an exercise, glance
quickly at Figure 1 above. How many orange sixes do you see? If
you “see” 7, you are correct. Most types of data visualization will
include a numerosity effect, however those that take advantage
of data reduction processes will be most beneficial for numerosity.
Consider scatter plots, histograms, and other clustering
visualizations.
CONCLUSION
The downside to being able to visually create meaning so quickly
and efficiently is that our brains can betray us and leave us with a
wrong idea – visual bias. Data visualizations are intended to clearly
and effectively communicate the correct information and insight.
Thus, we should pay close attention to recognizing key cognitive
elements in visualization, and how these should be used together
to craft a meaningful representation of data. This Research Brief
has highlighted some key cognitive elements affected by data
visualization. The next Brief in this series will review the visual
elements of formal analysis that are building blocks to visual
discovery.
This brief was originally published by the International Institute of
Analytics (IIA) in July 2015
P3/3
Radiant Advisors is a leading strategic research and advisory firm that delivers innovative, cutting-edge research and thought-leadership to transform today’s organizations into tomorrow’s data-driven industry leaders.To learn more, visit www.radiantadvisors.com
© 2015 Radiant Advisors. All Rights Reserved.
When you first look at a data visualization, you may not realize just
how much careful thought and effort has gone into crafting it. ata
visualizations are multi-faceted, not just because of their ability
to represent multiple types and layers of data in a meaningful
way, but because to craft the visualization itself requires attention
to every detail from color choice, to the methods in which the
important pieces of information are connected into patterns, to
the layout and typefaces used throughout the graphic.
In the previous brief, The cience of ata isualization, we
focused on features color, patterns, and counting that exploit
our brains’ intrinsic pre-attentive horsepower and form the basis
of a visual dialogue with data. In this brief, we’ll take a deeper
look into how to establish visual meaning of data through the
intentional selection of specific elements. These elements lines,
textures, shapes, and typography are some of the visual cues
that influence how our eyes move around a visual to separate
important areas from unimportant areas. These visual cues guide
us to organize the information to facilitate meaning through visual
discovery.
LI
The most basic building block of visual analysis, lines have several
purposes in data visualization. They are used to create complex
shapes discussed in the following section), to lead us visually
through or to) different areas of the visualization, or as a way to
layer texture on a visual surface.
Lines are especially potent tools to reinforce patterns or order.
This is because they offer powerful cues for our brains to perceive
whether objects are intended to grouped or linked together.
onsider a diverse group of different colored shapes. With its
pre-attentive capabilities, the brain will automatically group
them by shape and color. Adding a line to connect a subset of
common shapes will add more connectedness, and produce a
more powerful pattern see Figure 1).
Figure 1 – This figure shows the added patterning influence of a
line as a visual cue.
Lines can be used as labels, directional cues, or they can also
be used as a way to create texture in data visualization. Texture
is one of the more subtle design elements to include in data
visualization, but worth a mention due to its relation to lines and
shapes and our previous discussions on color).
Texture is defined as the surface characteristics or, the feel) of
a material that can be experienced through the sense of touch
or the illusion of touch). It can be used to accent an area so that
it becomes more dominant than another, or for the selective
perception of different categories. olors, shapes, and textures
can be combined to have further levels of selection. Finally,
textures of increasing size can represent an order relation. Because
it is usually accompanied with real adjectives like rough, smooth,
or hard, texture may seem intrinsically three-dimensional real).
It can also be two-dimensional. onsider how some graphs take
advantage of angles like points or circles).
In data visualization, lines are one of the best ways to represent
texture. For example, when used in tandem with color, lines can
create texture through what is called color weaving to produce a
tapestry of woven colors to simultaneously represent information
about multiple co-located color encoded distributions. For
example, color weaving is similar to the texturing algorithms
and techniques used to visualize multiple layers in topographical
maps, weather maps, and or other climatology visualizations.
I N S I G H T S E R I E S
VISUAL DESIGN BUILDING BLOCKS
1 Pre-attentive features are defined as a limited set of visual properties that are detected very rapidly and accurately by our visual system, and are not constrained by display size
© 2015 Radiant Advisors. All Rights Reserved.
HAP
In his article The Tall ffice Building Artistically onsidered,
Louis ullivan 1 6) made a statement that has forever affected
how we approach the premise of shapes and forms. He said All
things in nature have a shape, that is to say, a form, an outward
semblance, that tells us what they are, that distinguishes themselves
from another. ver the years, this form follows function refrain
has been taken both literally and completely misinterpreted it has
also been described as coarse essentialism by data visualization
design gurus like Alberto airo). Today, in visual design, it is a
powerful mantra primarily applied to the relationship between
the forms of design elements, i.e., shapes and lines) to the
informational function it is intended to serve.
As forms, shapes are one of the ways that our brains understand
patterns. This is a time saving technique: we will immediately
group similar objects and separate them from those that look
different. o, shapes and forms should be used to achieve what
the visual is attempting to communicate about the data. Again,
refer to Figure 1 above. This visual shows, at a glance, the impact
that shapes can have when looking at information.
hapes are formed with lines that are combined to form squares,
triangles, circles, and so on. They can be organic irregular shapes
found in nature circles, etc.) or geometric shapes with strong
lines and angles like those used in mathematics). Likewise, shapes
can be two-dimensional ) or three-dimensional ). These
shapes expand typical two-dimensional shapes to include length,
width, and depth they are things like balls, cylinders, boxes,
and pyramids. In data visualizations like pictograms, infographic
forms and shapes can be expanded significantly through the issue
of icons and other symbolic elements as extensions of traditional
shapes. onsider, for example, the use of shapes of people in lieu
of dots or other shapes.
Further, like so many other elements of design, color has a hand
in shape selection. This is particularly relevant in two ways. First,
make sure you are using color priority in choosing a shape if you
want to use circles to emphasize areas of opportunity for sales
agents on a map, use green circles instead of red or orange).
econd, be aware of color contrast and luminance between
shapes. The higher the luminance contrast, the easier is it to see
the edge between one shape and other. If the contrast is too low,
it can be difficult to distinguish between similar shapes or to
even distinguish them at all.
Like many visual cues, there is often no one right way of
encoding visualization properly through the use of shapes and
forms. any times it becomes less a question of correct and
more a consideration for what is easier for the viewer. As an
example, both scatterplots and bar charts can be used to represent
absolute variables. Both of these also support the description of
the use of shapes, forms, and colors to aid visual meaning of data.
T P G APH
Generally speaking, when we think of typography within the
context of a data visualization, we think in terms of two choices
of type categories: serif versus sans serif. While the origin of the
word serif is unknown, a common definition for it has come
to be feet. Thus, serif typefaces like Times ew oman or
Baskerville, are those with feet. Building on the previous,
the word sans comes from the French without. Thus, sans serif typefaces are those without feet. erif fonts are usually
considered to be more traditional, formal typefaces, while sans
serif typefaces tend to have a more contemporary, modern feel.
While these rules of thumb exist, there are no absolutes of when
or when not to use a serif versus a sans serif typeface.
In fact, in his book Data Points: Visualization That Means
Something, athan au pointedly notes that while there has been
much discourse on the best typeface, there has yet to be any true
consensus. This goes to further emphasize that typeface selection
is highly variable and depends much on personal preference. As
key takeaways, consider these three points:
P2/3
© 2015 Radiant Advisors. All Rights Reserved.
P3/3
First, be conscious of the amount and types of
typefaces and fonts used in a single visualization
Second, different typefaces and fonts have
different connotations and perceptions
And, while there is no absolute rule on typeface
selection, general rules of thumb apply
That said, there are a couple of important points to keep in mind
when making a typeface or font selection. First, typography, like
another other visual element, is no stranger to bias. There are some
typefaces for example, omic ans that have been reduced to
a sort-of comic strip application and are not taken seriously. thers
like Baskerville and Palatino conjure up nostalgia imagery due
to their historical use in vintage graphics. ome typefaces have
been custom-created for use in advertising like those fonts used
in tar Wars or Back to the Future, for example and are pigeon-
holed into their use in genre-related opportunities which is not
necessarily a bad thing, but one to be aware of). any typefaces
are also said to have personality. Like omic ans, others may come
with a more light-hearted or conservative personality. atching
type personality with the tone of the message in the visualization
is certainly not an exact science.
A good technique to see if you’re choosing appropriate fonts is to
use a font that seems completely opposite of what you’re trying
to convey. eeing how wrong a typeface can look will help
you make a more appropriate selection. hristoph Papenfuss’
blog, Performance Ideas, shares a visualization that is a perfect
example of how fonts should not be used in data visualization in
a 11 spending report from Papenfuss’ hometown in Germany.
The visualization has a long list of errors in design. pecific to
typography choices, it uses text that is so dense it is rendered
almost completely unreadable. ou can see graphic at http:
www.performance-ideas.com 1 6 poor-visualizations .)
The most pointed advice one can be given on typography is
this: use typeface and fonts with a purpose. It is easy to dismiss
the importance of these selections, possibly because we are so
conditioned to read text that we have become accustomed to
focusing on the content of the words and not what they look like
visually. However, the visual appearance of words can and does)
have just as much effect on how a document is received as the
content itself. Fonts can create mood and atmosphere they can
give visual clues about the order in which a document should be
read, and which sections are more important than others. Fonts
can even be used to control how long it takes someone to read a
document. Like colors, typefaces are typically chosen in corporate
style guides and other branding design decisions. It is valuable to
understand why typography is as important in data visualization as
any other design element.
P TTI G IT T G TH
ou might have noticed a theme in many of these discussions on
visual cues. They all seem to tie back to one or more pre-attentive
features discussed in the first brief. It is easy to see how visual
elements like lines, textures, shapes, colors, and typography
stimulate cognitive pre-attentive features in our brains that are
so critical in visual analysis. Hence, these are the building blocks
of visual discovery, intended by design to be layered upon each
other and used in mix-and-match fashion to make the most of the
visual capacity of data visualization.
This brief was originally published by the International Institute of
Analytics (IIA) in August 2015
adiant Advisors is a leading strategic research and advisory firm that delivers innovative, cutting-edge research and thought-leadership to transform today’s organizations into tomorrow’s data-driven industry leaders.To learn more, visit www.radiantadvisors.com
© 2015 Radiant Advisors. All Rights Reserved.
The first two briefs in this series The cience of ata isualization
and isual esign Building Blocks) have attempted to disentangle
the compounding factors that affect visual cognition, including
perceptual pop out, pattern recognition, numerosity, etc.
These previous briefs have effectively set the foundation
for understanding the science of data visualization and key
visual design building blocks that make data visualizations so
immediately meaningful by leveraging our brains’ pre-attentive
horsepower.
This brief, esigning for isual xperience, moves beyond the
premise of achieving efficacy in balancing art and science, to
understanding how to create a visual experience for complex
information through the lens of data visualization. We focus
on the value of data visualization as a form of communication
and storytelling, and we emphasize achieving balance within a
triangle of design constraints that amplify the viewer experience
of visualization and capitalize on visual memorability and learning.
WE REMEMBER VISUALLY
ecently, attributes-based visual recognition has received a lot of
attention with in-depth studies occurring in both academia and
industry. any of these studies have provided clear data and
learning opportunities. We’ve long known that we can remember
upwards of 1 , images at one time, and that we can recall
these images accurately at an percent recollection rate. This
speaks volumes to a person’s visual capacity and the power of
pictures as recollection engines in the brain. We also know now
that visualizations that blend information with influential features
such as color, density, and content themes such as recognizable
icons) significantly and reliably increase learning, memorability,
and recall. Further, seeing and interacting with an image in
combination with traditional written and verbal instruction has
been consistently associated with higher levels of retention and
understanding of salient ideas.
The Great Less is ore ebate
There is ongoing debate in the visualization community regarding
the role that a visualization type plays in data understanding and
memorability. And, like all great debates, there are convincing
arguments on both sides of the topic.
The conventional view argues that visualizations should be
devoid of chart junk and curated with as few design elements
as possible, and that simplifying a visualization increases
memorability and information saliency without the distractions
that lead to potential misinformation and conclusions. upported
by many data visualization gurus, this position is reinforced by
many psychology lab studies.
However, there is a body of research that argues against
lightweight design: chart junk might actually improve retention
and force a user to expend more cognitive effort to be able to
understand and learn from the visual, thereby increasing both
knowledge and understanding of the underlying data.
Figure 1 isual representations and examples of) the Less is ore approach
The original visualizations can be found at: ooblogs and The preadsheet Page
I N S I G H T S E R I E S
DESIGNING FOR VISUAL EXPERIENCE
© 2015 Radiant Advisors. All Rights Reserved.
Less is ore
A recent collaborative research project by computer scientists
from Harvard and cognitive scientists from IT1 explored
cognitive memorability of visualizations. The researchers began
with hopes of uncovering empirical evidence to support the
theory that while each of our sets of memories is unique, we have
the same algorithm embedded in our minds to convert visual
communication to memory, and thereby learn and retain learning.
sing a publicly available data set, the researchers augmented
an object and scene annotations to include spatial, content, and
aesthetic image proprieties). ome of their research findings were
surprising. For example, one highlight was that unusualness and
aesthetic beauty are not associated with high memorability, but
instead negatively correlated with memorability. This challenged
the popular assumption that beautiful images beauty being a
function of aesthetics) are more valuable in memory currency.
Another interesting find was that a visualization is instantly
and overwhelmingly more memorable if it includes a human-
recognizable element, such as a photograph, person, cartoon, or
logo. These types of elements essentially provide our memory
with a visual cue to build a story around, linking back to our
most primitive form of visual communication: symbolism, or the
practice of representing things by symbols. This finding creates
a compelling case for the use of icons in visualizations such as
infographics, which often rely on symbols to communicate large
or complex data in straightforward ways.
ore is ore
n the other side, a study2 on what makes a visualization
memorable began with researchers building a broad, static
visualization taxonomy of the large variety of data visualizations
in use today. The researchers collected nearly 6, visual
representations of data from various publications and used a
wide range of attributes to categorize these images. ext, the
researchers exposed the images to participants via Amazon
echanical Turk) and tested the influence of features such as
color, density, and content theme) on participants’ memorability.
The study confirmed previous research findings that faces
and human-centric scenes are more memorable than others.
pecifically, people and human-scale objects contributed most
positively to the memorability of images. The results of this study
also confirmed that certain design principles make visualizations
inherently more memorable than others, irrespective of a view’s
individual context and biases.
However, along with validations this study also revealed some
contrary findings. For example, visualizations with low data-to-
ink ratios and high visual densities more chart junk) were actually
more memorable than minimalist data visualizations. Likewise,
unique visualizations that left a lasting impression were more
memorable than traditional, common graphs bar charts or line
graphs which are considered part of the data visualization canon.
THE VISUAL EXPERIENCE BEGINS WITH A PICTURE
The above arguments aren’t to say that bad data visualizations
those that fail to take into account design considerations or
those that incorrectly graph information are okay. f course,
they’re not. But, that said, the research does reinforce the
power of pictures as a key part of how we communicate, learn,
and remember. oreover, it does support the idea that there
is more than one way to communicate effectively through data
visualization.
Through our intrinsic hard wiring to communicate, learn, and
remember visually, we develop the capacity for visual dialogue.
This visual dialogue is paramount to the experience of visualization.
onceived by athan nobler, former chair of the niversity of
onnecticut’s art department, visual dialogue is the exchange
that occurs between the artist, his work, and its consumer. It builds
upon the basis of visual memorability and literacy.
P2/4
1 Isola, P., iao, ., Torralba, A., liva, A. What makes an image memorable? I onference on omputer ision and Pattern ecognition P ), 11. Pages 1 -1 .
© 2015 Radiant Advisors. All Rights Reserved.
ognitive psychologists describe how the human mind, in its
attempt to understand and remember, assembles bits and
pieces of experiences and information into a story. To aid in this
story-building process, we automatically conjure images visual
cues associated with the story. This is the core of visual data
storytelling, but it speaks inherently to the power of pictures as
a visual experience. ata visualizations highlight the connections
between visualizations, design, and elements of science, and they
allow audiences to explore and develop a connection through
personal insight and experience.
THE PICTURE SUPERIORITY EFFECT
The concept of The Power of Pictures is not new it has been
described as an argument that’s both old and perennial. As an
example, consider this quote from a turn of the century text on
nglish composition: In some respects words cannot compare
in effectiveness with pictures. The mere outlines in a Greek vase
painting will give you a more immediate appreciation of the grace
and beautify of the human form than the pages of descriptive
writing. A silhouette in black paper will enable you to recognize a
stranger more quickly than the most elaborate description in the
world.
The Picture uperiority ffect P ) recognizes that concepts
learned by viewing pictures are more easily and more frequently
recalled than those learned purely by textual or other word-form
equivalents including audio or other information that is learned
by hearing). evelopmental molecular biologist ohn edina
quantified the P in the following way: when we read text, three
days later we only remember 1 percent of the information. et,
text combined with a relevant image is more likely to be recalled
at a much higher rate 6 percent in three days.
Whether art, memories, or illustrations for educational purposes,
people are drawn to pictures. This is because of the experience:
pictures conjure emotions, memories, and insights. They
stimulate new thinking. In this context, even more important than
our incredible ability to remember huge quantities of images with
a highly accurate recollection rate is the fact that we remember
them better than words. And we remember the images long after
we have forgotten the words that go with them. While words and
definitions vary from language to language, visualization as a
human communication mechanism is universal.
Figure isual representation and example of) the Picture uperiority ffect
TRIANGLE OF FORCES
In a previous brief in this series, we noted that data visualization
is a highly curated hybrid of forces and elements: when you first
look at a data visualization you may not realize just how much
careful thought and effort went into designing it. In this brief,
we’ve discussed two research views on when a data visualization
accurately conveys a message and when it reaches the tipping
point of too much.
Frequently, the intent of visual design is to clearly and effectively
communicate a single message. The best data visualizations are
those where nothing stands between the visual’s message and its
audience. While research is ongoing and the debate continues,
the simple truth is that when visual cues are used correctly, they
can bring data to life and give it more context, meaning, and
resonance. ata visualizations should be focused on the message
in the data, and visual enhancements like hue, saturation, size,
and color) should be used for emphasis rather than explanation.
P3/4
2 Borkin, ., o, A., Bylinski, ., Isola, P., unkavalli, ., liva, A., Pfister, H. 1 ). Proceedings from I I F I 1 : What akes a isualization emorable?. Atlanta, GA: I .
© 2015 Radiant Advisors. All Rights Reserved.
P4/4
very graphic is shaped by a triangle of constraints: the tools and
processes that make it, the materials from which it is made, and
its purpose or utility. This idea of design constraints as a triangle
of forces came from acob Bronowski, a mathematician, whose
discussions on context and visualization had a large impact. In
The hape of Things Bronowski wrote: The object to be made
is held in a triangle of forces If the designer has any freedom, it
is within this triangle of forces or constraints.
However, Bronoswki’s triangle of forces is not fixed. ach of its sides
can move, and as they do, the other sides in tandem so that they
move along with it. ach move of one side puts strain on the other
two. Thus, it is important to not only recognize the parameters of
the triangle of forces, but to strive for balance within it.
Figure isual representation of the Triangle of Forces Theory
This brief was originally published by the International Institute of Analytics (IIA) in
September 2015
L I
emorability is an intrinsic feature of visual information and
is reproducible across a diverse spectrum of visualization
mechanics. isuals that blend information with influential features
are significantly more memorable. We have high visual memory
capacity and can recall thousands of images for a long time. ore
importantly, concepts learned by viewing pictures are more easily
and more frequently recalled than those learned purely by textual
or other word-form equivalents.
ertain design principles make visualizations consistently more
memorable than others, irrespective of a viewer’s individual context
and biases. isuals that blend information with influential features
are able to create a visual experience for the viewer and, therefore,
are significantly more memorable. eeing and interacting with an
image also has been consistently associated with higher levels of
retention and understanding of salient ideas. Poorly designed data
visualizations that misrepresent data may very well be memorable
but for all the wrong reasons, making them possibly dangerous.
adiant Advisors is a leading strategic research and advisory firm that delivers innovative, cutting-edge research and thought-leadership to transform today’s organizations into tomorrow’s data-driven industry leaders.To learn more, visit www.radiantadvisors.com
© 2016 Radiant Advisors. All Rights Reserved.
Previous briefs in this visual design series have stepped through
the science, artistic curation, and various cognitive processes
enlisted in building a data visualization. In each of these briefs,
we’ve covered assorted design considerations for developing
data visualization for improved visual discovery and analytics.
However, so far we approached this primarily through a lens
of functionality and principles for use. In addition, there is a
substantial contribution that the individual consumer of the
visualization brings to the table in how data visualizations are used
and understood, whether it’s an analyst user, business leader,
customer, and so on. However, although this fact is acknowledged
and a wealth of long-standing literature covers the technical
aspects of both design and visualization, the collaborative, social,
and organizational aspects are less studied.
In this fourth and final brief, we tackle some of the key elements
of designing for influence by exploring the importance of viewer
perception and how this can be leveraged through certain
applications of techniques used to influence the user. This is the
human element of visualization, not the scientific, artistic, or
cognitive, and it plays a critical role in communicating through
data visualizations. In particular, interactivity and emotion
(storytelling) are two ways we can leverage data visualization to
influence our audience.
TH P W F I W P PTI
Perception is not a passive experience. ecent research in the
psychology of perception indicates that visual perception is
an exploratory and active process wherein we do not see a
sequence of static images, but a continuous flow of changing
scenes and imagery. Thus, we have to actively look around
to get a more comprehensive image of the information being
presented. To support this, the use of data visualization is often
aptly characterized as an active and exploratory process with the
goal to yield sometimes-complex insights. uring the course of
this process, data analysts and other users routinely generate
hypotheses, and then test them against the data that is visualized.
This visual analysis process is akin to the scientific method: we
guess what we think the information presented is intended to
mean (or could mean) and then look for additional data points
and visual cues to refine and confirm that.
This process of visual querying supports the notion that
perception is indeed active as opposed to passive. Figure 1,
which is borrowed from is aster’s ix, Pohl, and llis’ book,
Mastering the Information Age1, illustrates this process by
indicating ten steps or aspects of the active nature of processing
visual information.
Figure 1 From ix, Pohl, and llis, this visual illustrates a simplified view of the
broad visual analytics process
Though data is core to the analytical process, the human is
fundamentally at the heart of visual analytics. Therefore, the user
is in a position to affect multiple aspects of the visual process,
including visual perception, human interaction, and problem
solving analysis and activities driven by data visualization.
While there are multiple cognitive and perceptual processes
that will affect user biases and responses to data stimuli, we
must remember that visualized data is largely used by people.
Further, these people may or may not have received formal data
analysis education or possess detailed knowledge of visual design
principles. It is people who use a visual to make decisions and
take action.
I N S I G H T S E R I E S
DESIGNING FOR INFLUENCE
1 http: www.vismaster.eu news mastering-the-information-age
© 2016 Radiant Advisors. All Rights Reserved.
Also, while reactions will to some degree vary from person to
person, the important takeaway is that people are different, and
we have to be aware of these variances and accommodate them.
Instead of focusing this final brief on more of the wide context of
core elements for successful visual design, we narrow the focus
on the human aspects between a user and a data visualization
in terms of interaction and analytical processes and emphases.
With this in mind, we can better grasp the influence of actions
like interacting with visual stimuli or inviting emotion with visual
data storytelling.
I ITI G I T A TI I AL I
In data discovery, we often describe the discovery process as
iterative and agile. Further, we assume that analysts working
with data require the ability to move quickly and fluidly through
information and tools to extract, explore, and uncover new
meanings and analytical models generally referred to as
insights in previously unexplored information, or even in new
configurations of blended data. Another important element,
especially for the use of data visualizations as a vehicle to visual
discovery, is interactivity the ability for users to interact directly
with the data and the accompanying visualization to deepen
analytics insight.
Interactivity supports the type of visual thinking that drives visual
discovery. Without interactivity, visual discovery falls short of its
intended purpose and its analytical functionality is constrained
by the limitations of static imagery. However, with the right
interactivity, data visualization becomes a natural extension of
the users’ thought process. Interactivity, then, is the element that
allows users to play with data: to manipulate information into
patterns, theorize about meanings, project interpretations, explore
possibilities all while balancing the fixed content, context, and
relationships of the data with creativity and imagination.
Interactive capabilities in data visualization should support and
facilitate the users’ intentions when visually exploring data sets.
onsider the following list of interactive category types suggested
by researchers i et al2:
• elect: identify and select items of interests, possibly as a
precursor to another operation
• xplore: activities that include movements such as zooming,
panning, resampling, etc.
• econfigure: spatially rearrange the data e.g., sort, rotate,
change attributes assigned to axis)
• ncode: alter visual appearance e.g., change view or adjust
attributes like color, size, and or shape)
• Abstract laborate: show more or less detail drill down or up)
• Filter: select show data matching specific conditions or criteria
• onnect: highlight data related items e.g., brushing)
A word of caution here: the work of an analyst is influenced by a
host of cognitive biases, and many of these are set into motion
by the way information is “fed” into the perceptual cognitive
processes enlisted through data visualization. Interactive
visualization designs may influence biases and vice versa), and
thus we need to aware of how interactive visualizations affect
these and work to minimize them. For example, consider the
difference that a data visualization printed on paper may yield
insomuch as the users’ ability to visually move through the data by
drilling down or expanding certain elements, and how much more
valuable this process would natively be in a digital environment.
ITI G A I A WITH TI
Award-winning writer and director obert c ee said that the
most powerful way to motivate people to action is by “uniting
an idea with an emotion. The best way to do that is by telling
a compelling story that weaves in a lot of information and also
arouses the listener’s emotion and energy. c ee argues that a
good story fulfill s a profound human need to grasp the patterns
of living not merely as an intellectual exercise, but within a very
personal, emotional experience. A good story turns a one-sided
narrative into a conversation it influences an audience to action.
isual data stories should always put data at the forefront
of storytelling. However, data stories differ from traditional
P2/3
2 i, . ., ah ang, ., tasko, . T., acko, . A. ). Toward a deeper understanding of the role of interaction in information visualization. Visualization and Computer Graphics, IEEE Transactions on, 13 6), 1 -1 1.
© 2016 Radiant Advisors. All Rights Reserved.
P3/3
storytelling that typically chains together a series of causally related
events to progress through a beginning, middle, and end with the
goal of delivering a message. While data stories can be similarly
linearly visualized, they can diverge. ore important, data stories
can also be interactive to invite discovery, solicit new questions,
and or offer alternative explanations.
What makes visual data storytelling different from other forms of
storytelling is the complexity of the content being communicated
within the constraints of effective data visualization. It is essentially
information compression that condenses information into
manageable pieces by focusing on what’s most important, and
pretending it is bound entirely within the visualization(s) used
to illustrate the message. To that end, graphical techniques and
interactivity can enforce various levels of structure and narrative
flow. For example, consider an early reader picture book, which
illustrates salient points with a visual sometimes interactive to
emphasize key points of learning. And, though static visualizations
have long been used to support storytelling, today an emerging
class of visualizations that combine narratives with interactive
data and graphics is taking more of the spotlight in conveying
visual narratives.
For the purposes of storytelling, the basic plots of visual data
stories can be articulated as the following:
• hange over time see a visual history as told through a
simple metric or trend
• rill down start big, and get more and more granular to
find meaning
• oom out reverse the particular, from the individual to a
larger group
• ontrast the this or that
• pread help people see the light and the dark, or reach of
data (disbursement)
• Intersections things that cross over, or go from less than to
“more than” (progression)
• Factors things that work together to build up to a higher-
level effect
• utliers a powerful way to show something outside the
realm of normal
As a best practice, visual data stories are most effective when
they have constrained interaction at various checkpoints within
the narrative, allowing the user to explore the data without
veering too far away. isual cues for storytelling include things
like annotations to point out specific information using color to
associate items of importance without having to tell them or even
visual highlighting (e.g., color, size, boldness) to connect elements.
ome data visualization tools are being enhanced with storytelling
capabilities, too, and vendors in the space are supporting the use
of this engagement and influencing technique as part of their
tool’s platform. Tableau, for example, has its story points lik its
storytelling snapshots ellowfinBI its storyboard.
L I
Interactivity and storytelling both affect how the user experiences
a data visualization and are primary means of influencing audience
response to information presented.
With interactivity, a user is invited to get hands-on with the data,
and explore the information with visual thinking that drives
visual discovery. Thus, the data visualization becomes an extension
of the user’s thought process and enables them to “play” with
data and discover their own insights. Likewise, storytelling is an
appropriate way to unite an idea with emotion in data visualization.
The best data stories are influential: they are interactive to invite
discovery, engaging to solicit new questions, and thought-
provoking by offering alternative explanations to convey
compelling visual narratives.
adiant Advisors is a leading strategic research and advisory firm that delivers innovative, cutting-edge research and thought-leadership to transform today’s organizations into tomorrow’s data-driven industry leaders.To learn more, visit www.radiantadvisors.com