This document discusses advanced data visualization (ADV) and provides strategies for implementing effective ADV solutions. It outlines seven primary capabilities of ADV solutions, including dynamic data, visual interfaces, multidimensional analysis, and proactive alerts. It also describes methodologies like storyboarding and prototyping to enable ADV. Key benefits of ADV include improved operational efficiency, faster insight from data, and enabling users to create their own visualizations.
2. Contents
1 Introduction
2 Defining the problem
3 Methodologies to enable ADV
6 Functional capabilities
8 ADV gallery – Top 10 visualizations
14 Technical platform capabilities
15 Benefits realized
16 Risks and lessons learned
3. Solving the data visualization dilemma
1
Our brains are wired to love information, but when it comes
to handling data, we quickly develop headaches. Advanced
data visualization (ADV) is a rapidly emerging concept that is
becoming pervasive in business and society. ADV has a lofty
goal of transforming data into information. Merely noting how
annual reports have changed over the past 10 years — with data
displayed prominently in graphical formats — shows the impact
of ADV. Three converging trends have brought data visualization
to the forefront as a value driver. First is the prominence of big
data as table stakes in any organization. The second has been
the democratization of visualization tools, which allows access
to users who do not have advanced technical skills to build
visualizations. Finally, the pervasiveness of infographics in our
daily lives has increased expectations for visual representations.
Introduction
4. Solving the data visualization dilemma
2
“[Data] scientists will need visualization
experts the way writers need editors.”
— Harvard Business Review, Visualizing Data, April 2013
Defining the problem
Despite solving for the fundamental capabilities of big data and
providing easy-to-use tools for visualization, organizations are
still struggling with the basics: graduating from static reporting
to interactive, online presentation tools. The data visualization
discipline needs to be seen as an analytic process, not a reporting
outcome. This is the first barrier to overcome on the business
intelligence (BI) maturity model.
The overarching pain points in achieving data visualization that
are impediments to the goal are threefold.
1. Consumers want to easily recognize patterns in complex
data sets.
2. Companies need to synthesize large amounts into a single
palette. This is the “one-page thinking” principle.
3. The struggle of balancing breadth and depth of a complex data
model turns most users away.
Another primary obstacle to achieving value in ADV is
addressing the convergent skill sets needed: It is rare to find an
expert in programming, design and statistics that can readily
generate ADVs. Combining the right skills in a team seeking to
build data visualizations starts with the ability to ask the right
questions about data and toolsets. As we find answers, it becomes
possible to align and deploy ADV solutions and capabilities.
81% of executives state they highly value
data visualization, yet only 14% say they
interact directly with data visualization
tools and technology.
— Based on research of more than 500 Grant Thornton
Technology Solutions engagements
5. Breadth of solution
Speedtosolution
POC
Analytic Assessment
Approach (A3) methodology
Reporting
mockups
storyboards
Proof of concept (POC) –
Developed and visualized directly in
OBI based on known requirements
and developed data models
Models rapidly develop conceptual
designs and visuals using storyboards
and wire-framing design tools. Used
to validate key analytic paths
Comprehensive subject area
design, including data source
strategies and, measurement
strategy, reporting requirements
3
ADV solutions should contain seven primary capabilities that
address these obstacles (see Figure 1). It is important that both
functional and technical platform capabilities include each of these
components, just as classic BI/data warehousing solutions strive
to address reporting.
Methodologies to enable ADV
We hear two frequent questions across our BI and
analytics projects.
1. How do I know what is possible when it comes to data
visualization? (This deals with the classic conundrum
of knowing what to ask for, and also seeking the “silver
bullet” answer.)
2. How do I get started?
To answer these questions, Grant Thornton has developed
tiered methodologies (see Figure 2) to comprehensively address
initiating data visualization that take into account breadth of
solution and speed to deliver.
7 primary capabilities
1. Dynamic and immediate data
2. Visual interfaces with interactivity
3. Multidimensional analysis
4. Animation/use of motion
5. Personalization to end users
6. Actions/action frameworks
7. Proactive alerts
Figure 1: Fundamentals of ADV solutions
Figure 2: Tiered methodologies
6. Solving the data visualization dilemma
4
Storyboards and mock-ups
Using storyboards and mock-ups, we can rapidly develop
conceptual designs and visuals. Prior to the storyboard
process, we usually conduct a demonstration of BI application
functionality to set the stage. Creating a conceptual design, the
storyboard and mock-up process reduces development time and
rework (see Figure 4). Additional activities include prioritization
of content and reporting requirements, exploration of design
options, and rapid prototyping in the actual storyboard sessions.
Analytic Assessment Approach (A3)
The A3 methodology focuses on defining three key strategies or
inputs (see Figure 3).
1. The measurement strategy that defines key metrics, hierarchies
and calculations, which are important to the business. This is
the precursor to key performance indicators (KPIs).
2. The reporting strategy, which focuses on the current and
future-state delivery mechanisms for reporting.
3. Data strategy that assesses the target data sources, and how
data will be extracted and transformed for analysis.
Key outcomes of the A3 methodology include: analytic
roadmaps; detailed implementation and resource plans; business
case and return on investment calculations; and technology
selection and utilization plans.
Division heat map — portal into detail
using key metrics and indicators
Division financial reporting
(currently published monthly — period agnostic)
Project financial
summery
Project financial
detail
New business/
CRM
AR/cash
management
DFO/cash management summery (initiative/KPI-based)
I.
Measurement
Strategy
III.
Data
Strategy
Inputs Outcomes
II.
Reporting
Strategy
Implimentation Plan
Create Prototype
Tool Utilization
Figure 3: Analytic Assessment Approach
Figure 4: Financial reporting review
7. 5
Prototyping and proof of concepts
A proof of concept (POC) is a data visualization developed
and visualized directly in the BI technology, based on known
requirements and a sample data model. A POC relies on the
storyboard conceptual vision — focusing on a primary subject
area, detailed scenarios and aggregate presentation views.
Often, this process leverages the storyboard and uses it as an
interim landing or navigation page for new users in Oracle
Business Intelligence. POCs are typically detailed, visualized
analytic scenarios based on data models and reporting
requirements.
8. Solving the data visualization dilemma
6
Modes of delivery
The seven standard capabilities of ADV are delivered in three
primary modes.
1. ADV customers engage with the toolset via visual analysis and
discovery. Users interrogate the visualization — interact, drill,
pivot and zoom — to answer questions and pose new analysis.
2. Users engage via a familiar display of snapshot or point-in-
time reporting. The easiest way to relate to this mode is the
classic balanced scorecard report. At the end of the day/
month, the scorecard is the snapshot of reporting at that point
in time, with further information on KPIs, etc.
Functional capabilities
3. Proactive alerts to end users — regardless of device or
toolset, data visualizations can alert end users without the
need to interrogate data visualizations to find answers to a
predetermined question.
These modes of delivery combine with the ADV capabilities to
frame the functional capabilities.
Figure 5: Graphical relationships
gallery
modes of
delivery
capabilities
relationships
Data
visualization
gallery
Primary
modes of
delivery
Functional
capabilities
Standard
graphical
relationships
Data
visualization
gallery
Primary
modes of
delivery
Functional
capabilities
Standard
graphical
relationships
1. Visual analysis
2. Reactive snapshots
3. Proactive reporting
1. Nominal comparisons
2. Rankings
3. Time series
4. Part-to-whole
5. Deviations
6. Distributions
7. Correlations
1. Dynamic and
immediate data
2. Visual interfaces
with interactivity
3. Multidimensional analysis
4. Animation/use of motion
5. Personalization to end users
6. Actions/action frameworks
7. Proactive alerts
1. Classic waterfall
2. Strategy trees and wheels
3. Geo-spatial/geoprompting
4. Sparkline graphs
5. 80-20 relationships
6. Comparative distributions
7. Scatter cloud
8. Boxplot and whisker
9. Bubble chart
10. Master/detail views
1. Visual analysis
2. Reactive snapshots
3. Proactive reporting
1. Nominal comparisons
2. Rankings
3. Time series
4. Part-to-whole
5. Deviations
6. Distributions
7. Correlations
1. Dynamic and
immediate data
2. Visual interfaces
with interactivity
3. Multidimensional analysis
4. Animation/use of motion
5. Personalization to end users
6. Actions/action frameworks
7. Proactive alerts
1. Classic waterfall
2. Strategy trees and wheels
3. Geo-spatial/geoprompting
4. Sparkline graphs
5. 80-20 relationships
6. Comparative distributions
7. Scatter cloud
8. Boxplot and whisker
9. Bubble chart
10. Master/detail views
Data
visualization
gallery
Primary
modes of
delivery
Functional
capabilities
Standard
graphical
relationships
1b
9. 7
1
Harvard Business Review. Visualizing Data, April 2013.
2
Few, Stephen. “Selecting the Right Graph for Your Message,” Perceptual Edge, Sept. 18, 2004.
Understanding graphical relationships
With functional capabilities defined through general capabilities
of ADV and the modes of delivery, it is also necessary to have a
fundamental understanding of standard graphical relationships.
Data scientists need designers like writers need editors1
.
Understanding the basic tools of graphical relationships and
where they are used is a common cure for writer’s block when it
comes to ADV.
There are seven classic forms of graphical relationships. The
vast majority of quantitative depictions in business settings can
be described as one or a combination of these seven graphical
elements2
. Understanding these fundamentals can drive value in
selecting the right visualization concept.
1. Nominal comparisons are simple comparisons of the
categories and subcategories of one or more components in
any order.
2. Rankings simply list data points in a defined order by
a dimensional value selected — commonly shown in
descending or ascending order.
3. Time series relationships are a sequence of data points that
are ordered in common time buckets and typically plotted for
trending purposes.
4. Part-to-whole comparisons identify how subsets of a data
population relate to the total population value — displaying
ratios to the whole.
5. Deviations provide a comparative analysis of a standard
deviation on a data point for a selected set of dimensions
or values.
6. Distributions describe basic statistical discrete distribution
views of a selected population or data set.
7. Correlations refer to any of a broad grouping of statistical
relationships involving dependence between the
different groups.
10. Solving the data visualization dilemma
8
Below are the top 10 visualizations based on Grant Thornton’s
client projects and initiatives focusing on ADV and executive
analytics3
. Maintaining gallery visualizations are critical to
answering, “What is possible?”
1. Classic waterfall
Waterfall graphics show how an initial value is increased and
decreased by a series of intermediate values. They are favorites
of financial and accounting departments to show contributions
and profitability.
ADV gallery – Top 10 visualizations
3
All gallery screen shots are from Oracle Business Intelligence Enterprise Edition samples.
Figure 6: Classic waterfall
11. 9
2. Strategy trees and wheels
A strategy tree shows an objective and its supporting objectives
and KPIs hierarchically. The contribution wheel consists of a
center circle (or focus node) that represents the starting objective
of the diagram.
3. Geospatial/geoprompting
Geospatial reporting provides comparisons with a map backdrop
or comparison of distances between. Geoprompting provides
heat map alerts for users and prompts them to select areas and
drill to greater detail.
Figure 7: Strategy trees and wheels
Figure 8: Geospatial/geoprompting
12. Solving the data visualization dilemma
10
4. Sparkline graphs
A sparkline is a very small line chart, typically drawn without
axes or coordinates. It presents the general shape of the
variation — typically over time — in some measurement, such
as temperature or stock market price, in a simple and highly
condensed way.
Figure 9: Sparkline graphs
13. 11
5. 80-20 relationships
This report measures how the upper group of a specific
population set contributes in descending order of value. Filters
enable users to set a percentage limit of value for the top group,
and the report renders the corresponding percentage of the
population that makes up that value.
6. Comparative distributions
Comparative distributions are representations of statistical
distributions, by individuals, for a selected population. It allows
users to see how a metric is distributed among different categories.
Figure 10: 80-20 relationships
Figure 11: Comparative distributions
14. Solving the data visualization dilemma
12
7. Scatter cloud
This report provides a graphical summary of a set of data.
Individual values are represented by the position of the point in
the chart space. It displays measures of central median, dispersion
and skewness.
8. Boxplot and whisker
This report displays a boxplot and whisker diagram comparing
the spread of detailed data point values between individuals of a
dimension. It depicts a set of values for each dimension individual
through seven number summaries: smallest observation (bottom);
lower decile (10% mark); lower quartile and upper quartile
(IQR); median and average; upper decile (90% mark); and largest
observation (top).
Figure 12: Scatter cloud
Figure 13: Boxplot and whisker
15. 13
9. Bubble chart
Bubble charts are used in scatter plot scenarios where more
than two variables can be used. Data points are depicted by
the location and size of round data markers (bubbles). Bubble
graphs are used to show correlations among three types of values,
especially when you have a number of data items and you want to
see the general relationships. Bubble charts are useful to segment
populations of data, apply quadrant labels and prompt users for
further investigation.
10. Master/detail views
The master/detail linking allows you to establish a relationship
between two or more views; one view is called the master and will
drive changes in one or more views called detail views. You can
think of a master/detail relationship in a manner similar to what
you do when navigating from one report to another, but you do
not lose sight of the master view.
Figure 14: Bubble chart
Figure 15: Master/detail views
16. Solving the data visualization dilemma
14
Technical platforms need to address many advanced
requirements. We focus on three primary platform capabilities
of note.
Engineered systems
An engineered system simply refers to the “appliance concept”
to deliver the function of BI, analytics and visualizations. Apart
from the classic IT approach to technical platforms that often
considers hardware and software separately, analytic technical
platforms are increasingly thought of as an engineered system
possessing all critical components — software applications,
middleware, integration tools, hardware, etc. Perhaps the most
popular engineered system to date is the Apple iPad. This
solution-in-a-box thinking is a key requirement for ADV
technical platforms.
Technical platform capabilities
In-memory processing
In-memory processing is a fairly simple, yet very powerful,
innovation. Retrieving data from disk storage is the slowest
part of data processing: The more data you need to work with,
the slower the analytics process. The usual way of addressing
this performance issue has been to preprocess data in some way
(cubes, query sets, aggregate tables, etc.). In-memory processing
makes it possible to see the data more actively and at a deeper
level of detail, rather than in predefined high-level views. It allows
data visualizations to be more like natural thoughts.
Advanced interaction via write backs
Interactivity with data visualization is paramount, and often users
of a visualization tool need to provide additional input to alter
or enhance the analysis. From a BI standpoint, this is called a
“write back” and has special complexities and implications. This
goes beyond standard selection of parameters or prompting on
predetermined values or filters. Certain BI tools handle write
backs better than others; however, any ADV technical platform
must address this critical requirement. Our clients most often
use write backs to the underlying data model in what-if analyses,
predictive models and interactive commentaries with the data set.
17. 15
CEOs are demanding faster insight from data on hand, which
provides the platform for most business leaders and analysts. Data
visualization allows data discovery and visual analysis and reduces
time to insight.
As data visualization and BI tools drive interactivity with
underlying data, you can apply the global positioning system
(GPS) analogy. A strong ADV tells us where we are and where
we are going. ADV should enable end users to create their own
visualizations, providing a true democratization of analytics tools.
You can reap these benefits from data visualization efforts, as well
as the broader BI function:
Benefits realized
1. Improved operational efficiency
2. Alignment across organization and functional groups
3. Decreased time to insight
4. Faster response to changes
5. Ability to identify new business opportunities
6. Higher employee and partner productivity
7. Improved compliance with established standards
18. Solving the data visualization dilemma
16
The risks and lessons learned in executing data visualizations
relate back to our three main problem areas: recognizing patterns
in complex data, synthesizing data into a single point of view,
and balancing breadth and depth. The following risks and lessons
learned are common throughout ADV initiatives:
1. Data quality. Do not underestimate the importance of
data quality. Master data management tools cleanse data
at the integration level, and BI tools expose data issues to
be addressed. Data visualizations can mask data issues and
provide users with inaccuracies that will taint the analysis.
2. Content misrepresentation. Taking into account functional
capabilities, it is possible to select inappropriate graphical
representations and modes of delivery for data visualizations.
This can cause a misrepresentation of the data and the
information that the ADV is trying to convey.
3. Biases. Data visualizations can give power to the underlying
biases of the developer, designer or statistician and
contaminate the analysis of the end user.
Risks and lessons learned
4. Cluttered design. With all the functional capabilities for data
visualization, it is possible to take things too far — especially
in a single view. This can turn away the typical end user.
5. Data overload. Exposing too much data, without a logical
progression, or using data that is not absolutely necessary for
the intended purpose of the visualization, will overload the
end user and limit the effectiveness of the tool.
6. Delivery device agnostic. With dozens of potential interface
mechanisms, it is important to design the data visualization
with the intent of being flexible regardless of device — online
browser, laptop, tablet, smartphone, screen projection, etc.
7. Balance flash vs. function. Think simple and modern. Form
must always follow function with ADV, making the purpose
of the analysis the most important. Flashy graphics get “oohs”
and “ahs” initially, but are often abandoned quickly for
something else that works.
Conclusion
As organizations deal with exponentially increasing amounts of data, the patience of end users is decreasing. We see continued
struggles in addressing data visualization and turning data into information. Perhaps the greatest sign of a successful data
visualization or infographic is the degree to which it is used to solve problems. Data visuals must provide opportunities for
comprehension, conveying knowledge and clarity in understanding. Finally, success can be measured in retention, or how well
the visualization imparted meaningful knowledge. Using these fundamental factors for success, we can continuously improve
our data visualizations and techniques.
19. 17
About the author
John Stilwell is a senior manager in Grant Thornton’s Business
Advisory Services practice. He is currently a national lead
in Grant Thornton’s Business Technology Solutions group
with a focus on Oracle Business Intelligence. Stilwell has deep
experience in the area of analytics and business transformation
initiatives. He is a recognized national speaker and thought leader
on the topics of foundation analytics, mobile analytics, scorecard
and strategy management, and multidimensional reporting tools.
Stilwell has more than 15 years of consulting and technology
experience in a range of industries where he has provided clients
with solutions, including analytics, enterprise performance
management, strategic planning and strategic cost reduction.
John Stilwell
Senior Manager
Business Advisory Services
T 913.272.2721
E john.stilwell@us.gt.com