To succeed in any economic climate, businesses need to combine intelligent agents, 3-D visualization and predictive analytics into a framework that can detect underlying issues with revenues, expenditures and profitability and proactively provide solutions.
The Work Ahead in Intelligent Automation: Coping with Complexity in a Post-Pa...
Embracing Real-time Analytics for Proactive Business Management
1. • Cognizant 20-20 Insights
Embracing Real-time Analytics for
Proactive Business Management
Companies can achieve their strategic goals in any economic climate
by combining 3-D visualization, intelligent agents and predictive
analytics into a framework that recognizes underlying business
issues and proactively offers solutions.
Executive Summary • A healthcare payer is losing membership
due to layoffs among its client base, and its
In these times of economic uncertainty, com-
revenues are decreasing. The company needs
panies must adapt quickly to changes in the
to develop a comprehensive cost containment
revenue stream and find ways to stay competitive
methodology in pharmacy benefits, radiology
and profitable, even with substantially reduced
and other high-cost medical categories. To do
staff due to lower revenues. To gain competitive
this, it wants a real-time analytics system that
advantage, most companies are digging deeper
can identify and proactively provide solutions
into their volumes of data and turning to real-
to these problems and potentially replace some
time analytics.
of its labor-intensive operations in the claims
Let’s consider some real scenarios that companies processing area.
face today:
• A casualty and property insurance company
• A high-tech company wants to develop a micro- faces declining revenues due to the inability of
chip that would perform advanced analytics in policy holders to pay their premiums because
real time, to be used by the retail industry in of job loss. With pricing optimization a key
digital signage. This microchip would decrease product differentiator in the marketplace, the
the cost of advanced analytics by millions of company wants to use analytics to increase
dollars and allow retail companies to conduct profitability by identifying and providing
real-time marketing analytics campaigns. potential solutions in high-cost areas by
deploying an early warning system.
• A social media company wants to use real-time
advanced analytics to improve its click- through • Aninternational consumer packaged goods
rate by 20% by matching targeted advertise- company is experiencing decreased sales and
ments to users. wants to gain back market share by utilizing
cognizant 20-20 insights | april 2012
2. analytics in the areas of foresight, blind spots, and complex data sets. To truly appreciate IAVM,
portfolio optimization and consumer insights. we should first understand the components that
create its foundation. These include the following:
These challenges can be addressed through
the use of agile and resilient computer systems
that can detect underlying issues with revenues,
• Business Competency Model: One of the
issues that the IAVM strives to resolve has
expenditures and profitability and proactively existed since the beginning of the 21st century:
provide insights to help companies adapt quickly leaders have a specific vision, or mental image,
to continuously changing conditions. of where they want to take their companies,
but there is no visualization mechanism to
Such systems depend on the utilization of three
describe that vision to the rest of the company.
scientific and technology techniques that have
This became clear in 2000, in my work with
been in the making for years: intelligent agents,
a Fortune 100 company that was trying to
three-dimensional (3-D) visualization and data
determine how to change course in a fiscal
mining with predictive modeling techniques. These
quarter without negatively impacting profits.
three technologies form the core of the Intel-
We developed a business model, which later
ligent Agent Visualization Model (IAVM), a deci-
became the business competency model (BCM),
sion-support framework that allows companies
that could respond to recessionary times by
to efficiently design, build, deploy and update an
bridging the gap between corporate strategy
agile and robust enterprise analytics system that
and operational decision-making. It accom-
supports profitability under any economic climate
plishes this by making sure the entire company
by understanding business changes and providing
mirrors the executive management committee
proactive solutions to decision-makers.
and that the committee is organized according
An essential concept of IAVM is that business to the corporate goals and vision of its leaders.
data, like our universe, is three-dimensional. IAVM
• Workforce Turnover Efficiency ratio: The
seeks to improve users’ ability to see patterns WTE is an asset management ratio that allows
within business data by increasing their depth companies to design restructuring plans based
perception of traditional two-dimensional data on contributions to revenues. It was developed
analysis. This improved visualization is achieved in 2001, when a Fortune 100 company realized
through techniques such as self-populated maps it needed to lay off tens of thousands of
that enable executives to quickly compare per- employees in order to stay profitable but did
formance across different geographical regions. not have a metric or KPI to measure how each
Such a 3-D dashboard would also include individual and group contributed to revenues.
drill-down capabilities to allow further examina-
tion into issues detected by the IAVM. • Weighted Outlier Variable: The WOV is a way
to separate clusters of data and understand
To create agile systems using enterprise analytics, the driving factors for any changes. It was
companies must focus on three main areas using developed in 2003, when I was designing fraud
IAVM design and implementation: and abuse analytical detection models. To my
surprise, probability theory and statistics had
• Corporate goals answered the dispersion (standard deviation)
part of the equation,1 and Albert Einstein had
• The business model shown mathematically how to clarify driving
• Metrics factors using an algebraic concept (quadratic
equations) that had been around for over
Doing so will result in responsive and flexible
2,000 years.
systems that allow survival and prosperity even in
harsh economic conditions. Companies that use • Depth perception studies: Business analytics
proven science and technology in their decision- visualization borrows from analytics meth-
support systems will earn an advantage in the odologies and algorithms used in diagnostic
marketplace in good times and bad, since they imaging. I realized the link in 2004, when I was
will be able to quickly adapt to change without researching the area of neuroscience to better
negatively impacting their core business. understand diseases that both my parents
were diagnosed with. Moreover, I realized that
IAVM Foundational Components cognitive science and medicine had found
The IAVM framework is a culmination of 11 years of that depth perception (binocular summation)
research and design of business analytics in large involves the brain making predictions about
cognizant 20-20 insights 2
3. size, movement and distance. The result of • Intelligent agents: An intelligent agent is
adding depth to our vision capabilities had software that is autonomous; interacts with
been calculated to improve vision acuity by a other agents (is sociable); reacts to its environ-
minimum of 140%. ment; and proactively tries to reach its goals
by producing solutions. This technology allows
• Commoditization of the statistics meth-
for software to detect and suggest solutions
odology: In 2007, I learned that the Analysis
to business problems. Basically, advances in
Services team at Microsoft Research Lab-
technology (more data processed more quickly
oratories had optimized regression and
using smaller form factors) allow us to perform
partition algorithms. From this, I realized that
multiple calculations in a very short time. Intel-
the statistics methodology had become a
ligent agents are currently used in a number of
commodity and that the additional key ingre-
industries, such as in electrical grids to ensure
dients were variable creation, visualization and
a continuous flow of electricity to hundreds
domain knowledge.
of millions of consumers, as well as in large,
• Three-dimensional visualization: In 2008, distributed commercial systems to detect and
after seeing the work from the Visualization control intrusion.
Group at the Lawrence Berkeley National
Laboratory (see page 8), it became clear to Methodology
me that 3-D visualization could be adapted In business, it sometimes seems easier to live with
to business analytics to share strategic vision a familiar problem than implement an unfamiliar
across the enterprise. solution. This is particularly true in corporate
decision support systems; however, the 3-D visu-
• Optimized delivery model: In 2009, I found
alization of analytics clarifies underlying issues in
that Cognizant’s business model is an optimal
a way that anyone can understand.
delivery model for the IAVM. There are three
aspects of our model that are tailor-made for Current decision support systems are difficult
the IAVM: our on-site/offshore ratio for solution and expensive to manipulate and seldom proac-
delivery; our depth of analytics experience; and tively provide solutions to issues. On the contrary,
our domain expertise in multiple industries.
IAVM High-Level Framework
Goal Assessment Corporate Alignment
Clear definition of vision, goals BCM ensures that the organization
and stakeholder responsibilities. supports the corporate goals and vision.
1 t Org Stag
ge men Reaaniza e 2
Sta sess (BC lign tion
s M a
lA Pro ment l
G oa ces
s)
Metrics Definition
IAVM Implementation
The WTE ratio ensures accurate
Implementation
Definition (WTE
Design, build, test,
Stage 6
measurement of how
Stage 3
Metrics
IAVM
implementation,
individuals and sub-organizations
visualization and
contribute to revenues and
maintenance.
)
profitability.
Age s
nt D ate
Upd )
Sta
esi
gn KPI (WOV 4
ge ge
5 Sta
Agent Design KPI Updates
Interface, tasks and WOV separates clusters of data
information agents. and clarifies driving factors in
large and complex data sets.
Figure 1
cognizant 20-20 insights 3
4. IAVM Continuous Improvement Process: Kaizen Analytics
BCM
Goals GIS
Analysis Alignment
3-D
Dashboard IAVM Visualization
Ad Hoc Metrics
WOV Reports
WTE
Figure 2
IAVM is a decision-support framework that allows as costs have declined and ease-of-use has
companies to efficiently design, build, deploy and improved to the point that anyone in the organi-
update an agile and robust enterprise analytics zation can use these tools.
system by understanding business changes and
providing proactive solutions to decision-makers. Step 1: Assess corporate goals
and business rules
The high-level process of IAVM involves six
The first step in IAVM is to assess corporate
different steps that constitute a continuous
goals and business rules. Before designing any
improvement method, or kaizen analytics (see
decision support system, the company needs a
figures 1 and 2).
clear understanding of the corporate goals and
Instead of great technological breakthroughs, how those goals flow through the organization.
the kaizen approach aims to involve the entire General statements of increased profitability and
workforce in a continuous improvement process. decreased costs must be translated into specific
Hence, most of the improvements are small and metrics that can be reported, measured and
process oriented (like making shelves easier predicted. Discovering business rules is essential
to reach), but they involve the entire workforce during the assessment since these rules tend to
rather than a selected few, inspiring the enterprise mirror corporate compliance and workflow.
as a whole to be vibrant and innovative. A good
The conceptual design of the 3-D visualization
example of how this works is at Toyota, whose
begins within this phase because the visualization
employees provide management with 100 times
needs to mirror the corporate vision and goals.
more suggestions for improvement than other
For example, a soda manufacturer and distributor
auto manufacturers.
may want to see the aggregate visualization as
Businesses that want to improve their analytics a series of 3-D soda cans, or a retailer may want
capabilities should follow the kaizen approach and to see the aggregate visualization as a category
make business analytics available throughout the of consumer goods. These visualizations can be
entire organization. In some companies, analytics self-populated maps like the ones used by the
is limited to the purview of the few — statisticians, Lawrence Berkeley National Laboratory Visual-
physicians, molecular engineers and actuaries — ization Group, with underlying geographical infor-
often because it is seen as expensive and difficult mation system (GIS) and dashboard technologies.
to interpret. This premise is no longer applicable, A 3-D visualization of the enterprise’s analytic
cognizant 20-20 insights 4
5. and predictive capabilities will allow executives can be used in M&A, due diligence and financial
and field staff to use the power of the human analytics.
brain to its fullest potential.
In today’s economy, companies like to say that
Step 2: Realign corporate structure human capital is their most important asset.
The second step in IAVM is to evaluate the Indeed, the last 10 years have seen the develop-
company’s organizational structure and make ment of a service economy and increased reliance
recommendations for how to better align the on the knowledge worker. As a result, the mea-
company with its corporate goals. This is where surement of management efficiency in utilizing
the BCM comes in.2 The BCM is a three-pronged human capital has moved to the forefront of
structure that aligns the company’s financial this benchmarking exercise; hence, it is essential
goals and organizational model with strategic to develop a financial performance tool that
planning, assessment tools and knowledge determines how an organization is managing its
management (see Figure 3). Its leading feature is workforce.3
its efficiency, allowing a company to turn around
Asset management ratios measure the ability of
in a short time period, even one financial quarter.
assets to generate revenues or earnings. As such,
This type of agility is a necessary characteristic
they complement liquidity ratios when analyzing
for any decision support system that involves
financial performance. There are six other asset
human-computer interaction (HCI).
management ratios: accounts receivable turnover,
days in receivables, inventory turnover,4 days in
Step 3: Define metrics
inventory,5 operating cycle6 and capital turnover.
The third step is to define metrics and determine
how they aggregate through the company in WTE is calculated by multiplying average daily
order to predict and meet corporate goals. An salary (ADS) with the actual number of days to
organization must measure what it expects to fill an open position (TTF), dividing that sum by
manage and accomplish; otherwise, it has no the average number of days to fill a position (ATF)
reference with which to work. The IAVM uses a and then dividing again by 10 (see Figure 4).7
company’s current metrics and enhances them
by using the WTE ratio, which measures the rela- WTE is useful for companies with a large number
tionship between the cost per employee and the of employees (over 10,000). These companies
timely management of project staffing. This ratio can be in different industries such as healthcare,
manufacturing, financial services, telecommuni-
cations and other services. Also, it can be used
to measure performance efficiencies within any
BCM Framework organization, including but not limited to IT and
business processes.
Corporate Step 4: KPI updates using
Executive the weighted outlier
Committee
The IAVM also uses the weighted outlier meth-
odology8 to improve visibility into data patterns.
An outlier is an observation that lies outside
Knowledge
Tools Strategy Management the overall pattern of a distribution in the data.
Usually, the presence of an outlier indicates
Workforce Turnover
Efficiency™ Ratio
IAVM
Figure 3 Figure 4
cognizant 20-20 insights 5
6. some sort of problem. The weighted outlier The visualization architecture consists of three
variable (WOV) separates clusters of data while main layers: 3-D interactive visualization, a geo-
simultaneously clarifying the driving factors in graphical information system and a dashboard
large and complex data sets (see Figure 5). A with drill-down capabilities. The 3-D interac-
weighted outlier creates variables that maximize tive visualization uses the medical concept
the differences in the data, while simultaneously of binocular vision, which adds an additional
minimizing the similarities in the data to detect predictive variable to two-dimensional data.
potential fraud. This effect could be described as
“squeezing and pulling out” the potential fraud Business data is three-dimensional; however,
from the data set. A significant WOV should also business analytics tend to be flat, or two-dimen-
substantially increase the efficiency of a data sional, like an Excel table or chart. The difference
model for fraud detection. between a 2-D analysis and a 3-D analysis is
depth. Depth perception allows an individual to
Step 5: Designing intelligent agents accurately determine the distance to an object.
The design of the IAVM framework takes into
In analytics, depth is referred to as dimensional
consideration three different types of intelligent
analysis. Dimensional analysis is used in engi-
agents:9
neering, physics and chemistry to understand
• Interface agent: Collects information from the characteristics of multi-dimensional data and
users and delivers requested information. formulate hypotheses about the data that are later
tested in more detail. In business analytics, we
• Task agent: Performs most of the autonomous can create a 3-D variable that allows the end-user
functions. For example, task agents calculate in
to “see the depth” of the data. This variable is
real-time the mean and standard deviation of
called a 3-D vector analysis. This variable, when
a specified value and then decide the outlier
combined with cluster analysis and a visualization
limits for an alarm script. Also, these agents
tool, answers the recurring business question:
may decide whether the solution is a potential
How deep can I go into my data and see patterns
data error, fraud issue, new pattern or risk
in which sound business decisions can be made?
management issue.
• Information agent: Used for one-time retrieval The main goal is to increase the user’s under-
of information that has reusable capabilities. standing of the data by adding depth perception
(i.e., predictive modeling) to traditional 2-D data
Step 6: IAVM implementation/ analysis. This method, binocular summation,
visualization increases visual perception by a minimum of
The conceptual framework of the IAVM is depicted 140% in clinical studies.11
in Figure 6 (next page).10
Weighted Outlier Effect
K=18
K= Kurtosis
K=3
S= Skewness
S= 2 S=27
X1 WOV
Figure 5
cognizant 20-20 insights 6
7. IAVM High-Level Architecture
User 1 User 2 User h
Goals and
Task
Results
Specifications
Interface Interface Interface
Agent 1 Agent 2 Agent k
k
Tas
d
ose
Tas
k Prop tion
Solu
Task Task Conflict Task
Agent 1 Agent 2 Agent j
Solution
Information
Information Integration
Reply
Request
Info Collaborative Info Info
Agent 1 Query Processing Agent 2 Agent n
Query Answer
Database Database Database
1 2 k
Figure 6
The IAVM uses this increased visual perception Other potential applications for real-time IAVM
to its advantage. An example is a self-populated include but are not limited to:
map that allows executives to determine potential
issues and solutions to achieve corporate goals • Retail: As a consumer browses through a
(see Figure 7, next page). The geographical infor- store (brick and mortar or Internet), intelli-
mation system gives the user a spatial dimension gent agents react to browsing and purchasing
among different geographical regions for com- patterns to recommend additional articles to
parative analysis. The dashboard view should purchase. This output then can integrate with
have drill-down capabilities that allow users to a marketing campaign to send coupons that
examine the root causes of the issues detected target the consumer’s preferences.
by the IAVM. • Financial services: Early warning systems
react to diverse credit card purchases, and
Real-Time IAVM Applications investment mechanisms proactively detect
To fully understand the potential for IAVM,12 fraud and abuse.
we must recognize how intelligent agents are
currently used in the following industries:
• Healthcare: Systems detect and proactively
recommend diagnoses and treatment based
on real-time clinical and claims data in a digital
• Healthcare: As patient care becomes more
hospital setting or in a claims processing clear-
data intensive, intelligent agents are used
in intensive care settings to administer inghouse.
medication by proactively reacting to constant • Internet gaming companies: An inflation
monitoring of vital signs. control tool acts as a central bank regulating
the supply of money to control inflation in
• Air traffic control: The volume and complexity virtual economies.
of managing air traffic control systems requires
the utilization of intelligent agents to avoid • Internet advertisement: Mobile agents detect
collisions and manage departures and landings. patterns in user behavior and proactively com-
municate with other agents to determine what
• Manufacturing: Robotics has become one of advertisements to display.
the main applications in the manufacturing
industry, and intelligent agents are used to • Communications: Intelligent-agent technology
react and proactively make decisions regarding efficiently transfers calls and detects potential
quality control processes. outages.
cognizant 20-20 insights 7
8. Self-Populated 3-D Visualizations
Source: Lawrence Berkeley National Laboratory
Figure 7
Conclusion adapt during difficult economic conditions and
flourish during strong economic times.
The IAVM has multiple applications in analytics
around big data for the high-tech, healthcare, As an added benefit, the IAVM proactively brings
retail, pharmaceutical, life sciences, CPG, banking potential solutions to issues based on sound and
and financial industries. It can be used for M&A, proven mathematical and scientific methods
risk management, financial analysis, corporate like standard deviation, risk detection, outlier
asset management, restructuring, fraud detection analysis and visualization. It allows decision-mak-
and best practices identification. This framework ers to gain confidence in their understanding of
incorporates proven business, scientific and why a goal-related issue has surfaced (or been
technological methods and processes to provide detected), and why a specific solution has been
companies with a flexible and robust decision recommended.
support system that will allow them to rapidly
Footnotes
1
Karl Pearson, “Contributions to the Mathematical Theory of Evolution — On the Dissection of Asymmetri-
cal Frequency Curves,” Philosophical Transactions of the Royal Society of London, Vol. 185), 1894, pp.
71–85, The Royal Society.
2
Alberto Roldan, “The Business Competency Model: Turning Around in a Quarter,” Business Analytics blog,
April 10, 2008, http://atomai.blogspot.com/2008/04/enterprise-business-analytics-turning.html.
3
Noel Capon, John U. Farley and Scott Heonig, “Determinants of Financial Performance: A Meta-Analysis,”
Management Science, Vol. 36, No. 10, 1990, pp. 1,143–1,159; Bronwyn Hall, “The Relationship Between Firm
Size and Firm Growth in the U.S. Manufacturing Sector,” Journal of Industrial Economics, Vol. 35, Issue
4, 1987, pp. 583–605; Edwin Mansfield, “Entry, Gibrat’s Law, Innovation, and the Growth of Firms,” The
American Economic Review, Vol. 52, No. 5, December 1962, pp. 1,031–1,051, American Economic Associa-
tion; and Robert Gibrat, Les Inégalités Economiques, Paris: Sirey, 1931.
4
Inventory turnover is similar to accounts receivable turnover. It measures how many times a company
turned its inventory over during the year. Higher turnover rates are desirable, as they imply that
management does not hold onto excess inventories and that its inventories are highly marketable.
Inventory turnover is calculated as follows: Cost of sales/average inventory.
5
Days in inventory is the average number of days a company holds its inventory before a sale. A low
number of inventory days is desirable. A high number of days implies that management is unable to sell
existing inventory stocks. Days in inventory is calculated as follows: 365 or 360 or 300/inventory turnover.
6
Operating cycle = number of days in receivables + number of days in inventory.
cognizant 20-20 insights 8