Predictive Analytics beyond BI

1. Sponsored by
Expanding BI’s role
by including Predictive Analytics
In today’s economic downturn, organizations are looking for
ways to improve the way they do business to keep ahead of
the competition and grow revenue.
In a 2009 CIO Insight survey of senior managers and
IT executives, respondents listed their top priorities as
improving business processes, delivering better customer
service, generating more business from new and current
customers, and differentiating the company from
competitors via IT. But faced with the challenging economic
environment and reduced funding for new initiatives, how do
organizations focus on meeting these prioritized objectives?
The path to success in all of these areas, traditionally, has
been to use business intelligence (BI) information to make
decisions. Increasingly, organizations are finding that the
benefits of BI can be enhanced when complemented by
predictive analysis. Specifically, more insight can be gained,
and even better decisions made, by coupling business-
relevant information with an easy-to-use predictive analytics
solution.
A Natural Extension to BI
Business intelligence provides valuable insight into the
state of affairs within an organization. The information
is critical to decision-making. But when combined with
predictive analysis, synergies can be leveraged to improve
business and operations.
Many industry analysts like to make an analogy between BI
and predictive analytics by citing a quote from the famous
hockey player Wayne Gretzky, who said: “A good hockey
player plays where the puck is. A great hockey player plays
where the puck is going to be.”
Comparably, BI tools help users know what has happened
and what is happening, while predictive analytics tools
help to elicit more from this information by providing
an understanding of why these things happened and in
predicting what will happen.
For example, BI tools can report which sales region had
the highest sales, how many widgets were sold in stores
in different ZIP codes, the average spending per online
customer vs. in-store customer, and how many customers
stopped doing business with your company last year. All
of this information is essential for developing new product
and services, allocating resources, investing in marketing
campaigns, and so on.
Predictive analytics tools, though, can give deeper insight
into why these things happened. For example, knowing
the average customer spends $100 per visit to a store
is one thing. Knowing that a certain 20 percent of the
customers are responsible for 80 percent of all revenues
and that they are more likely to buy particular products
bundled together is much more valuable. Also, identifying
which products influenced the purchase of others or the
strength of the relationship between products purchased
together would give more insight into specific buying
patterns. This added level of analysis can yield valuable
results. It helps you understand how that prized segment
of your customer base would respond to very targeted
promotions.
Similarly, knowing that the average response rate to a
direct-mail marketing campaign is, say, 4 percent, an
organization can decide how often to run these campaigns
factoring in mailing costs and the revenue generated by
a campaign’s sales. Knowing the types of customers and

2. Sponsored by
being able to correlate that with what they purchased and
when they are likely to purchase again would allow an
organization to target those customers at the right time
with the right offerings. This would allow the company to
while ensuring customers are offered a product or service
they would actually be interested in.
That’s the difference between BI and the power of BI
combined with predictive analytics.
MANy ApplicAtioNs
Predictive analytics helps organizations look forward
and make educated decisions that anticipate the future
needs of customers. It combines known information
about customers, sales, operations, or finances, with
critical insight that helps solve problems, achieve business
objectives, and uncover hidden patterns not easily
identifiable through reports or dashboards. The combined
knowledge is used to take actions that can improve
business.
A traditional example of predictive analysis’use would be
to identify trends like poor customer service or customer
dissatisfaction and correlate complaints to customer churn.
Having insight into why customers are leaving or why they
stay, an organization can take action to retain them. For
instance, by surveying customers, an organization might
find that 30 percent of their customers consider the price
of the service to be the most important factor in choosing
a company. Another 30 percent might love to receive perks
and consider such offerings a distinguishing factor that
keeps them coming back. And the rest might simply feel
that timely and courteous service is essential.
Having this level of insight into customer likes and dislikes
can help an organization make predictions about the future
actions of these customers. Correlating this information
with actual customer actions allows an organization to
take action. For example, having identified a segment of
the customer base that attaches importance to pricing, an
organization might offer discounts or reduced rates if the
customer signs a multi-year contract. Those who love perks
might be offered free shipping, a free music download, or
an extra day at a hotel.
In another area, an organization might use predictive
analytics to cross-analyze sales data and marketing
spending, perhaps finding that 80 percent of the sales in
response to direct mail or e-mail campaign come from only
group of 20 percent in future campaigns, the organization
can significantly increase the ROI of these campaigns.
Additionally, an organization might use purchasing
information tied to a customer loyalty program to
understand which products are purchased together, by
whom, and when. Having this information, the organization
can try to increase revenues by cross-selling distinct
bundles to select customers. For instance, a retailer might
find that customers who bought the highest-priced suits
also bought shoes to match and a protective trench coat.
Having that information, a store might selectively place
trench coats next to the high-end suits. Or it might develop
a marketing campaign that offers discounts on shoes and
trench coats when a customer buys a suit valued over a
certain price.
With such success from traditional predictive analytics
usage, organizations are looking to expand its influence to
more areas of operations and to more users. In particular,
predictive analytics is increasingly being used to help
identify key influencers in customer satisfaction, employee
retention rates, customer churn, and other areas.
For example, Human Resources (HR) might use predictive
analytics to help select job applicants. Specifically,
employers want to predict which job applicants are going
to make a commitment to their job. Predictive analytics
can be used to show which personality traits are better
predictors for worker productivity and turnover.
Predictive analytics might also be used to retain talented
employees by helping predict if an employee is likely to
leave based on the types of services they consume from
the company such as training, taking advantage of 401k
plans, or the number of vacation days taken. Armed with
this information HR managers can target top performers
with programs designed to increase their investment in the
company and hence their likelihood of staying.
Predictive analytics can also bring value to other areas
of operations, such as manufacturing. For example,
organizations could use predictive analytics to help identify
and predict equipment maintenance for it products,
ultimately increasing customer satisfaction. By analyzing
Predictive analytics tools can give deeper
insight into why these things happened
increase the effectiveness of their marketing promotions 20 percent of its customer base. By selectively targeting this

3. Sponsored by
data collected by systems such as an aircraft’s health
and usage monitoring system and flight maintenance
log records, an aircraft manufacturer can determine the
relationships between how the aircraft is being operated
and maintained and the consumption of parts. The
deeper understanding of these correlations will allow
the manufacturer to take proactive action to reduce
direct maintenance costs or even improve manufacturing
processes.
Limitations Stymie Usage
Such potential benefits derived from predictive analytic
solutions are getting the attention of many organizations. In
fact, a 2008 IDC BI and Analytics Survey found that predictive
analysis tools were the number-two priority for purchasing
within the next 12 months (second only to business activity
monitoring tools).
Why the growing interest? According to IDC, the median ROI
for BI projects using predictive technologies was 145 percent,
compared with a median ROI of 89 percent for projects
without them.
Forrester Research recently forecasted predictive analytics
and data mining will also grow at a rapid pace, more than
doubling in growth to nearly $2.2 billion within the next five
years.
While many have noticed the synergies that might be gleaned
by combining BI and predictive analysis, predictive analytics
tools have not been as widely embraced as BI tools. There are
several reasons for this.
Predictive analytics tools are often designed for analysts. They
assume a high-level knowledge of statistical analysis methods.
In particular, an analyst would be needed to determine which
mathematical tools to apply to a problem: linear regression, a
chi-squared distribution, something else? As such, many tools
are difficult for business managers and others to use.
Additionally, many predictive analytics tools require special
programming skills. Users must not only know which formulas
are appropriate for doing a specific analysis, they must then
know how to create and enter the formula to analyze the
relevant data. Some tools require the use of programming
languages like C or C++; others might rely on the statistical
analysis programming language R. This puts these tools out
of the reach of the majority of business users. And in some
cases, IT must get involved, making many predictive analytics
efforts rigid and not flexible enough to meet rapidly changing
market conditions.
Another shortcoming with many predictive analysis tools is
that they are standalone tools. This complicates matters in two
ways.
First, getting access to information can be difficult. BI solutions
often provide a means to access relevant information for
decision-making. If the predictive analytics tool does not
integrate well with a complete BI solution or does not
accommodate data access and data mining, the user will have
to obtain the information for analysis in a brute-force manner.
Second, if the tool is standalone, it might not provide an easy
way for the results of the analysis to be shared, viewed, or
made part of a decision-making workflow. But being part of a
solid BI infrastructure makes it easier to share the results with
the right users who need the information to transform the
business.
Bringing Predictive Analytics into the
Fold
SAP offers a way to overcome these limitations and make
predictive analytics a part of the normal business decision
making process.
Its SAP BusinessObjects Predictive Workbench helps uncover
trends and patterns to solve business problems, anticipate
business changes, and make forecasts.
SAP BusinessObjects Predictive Workbench integrates
with your existing data environment – as well as SAP
BusinessObjects Enterprise environments – and it allows for
efficient discovery of important and predictive findings.
At the heart of the offering is an easy-to-use visual workflow
interface. Using Predictive Workbench, business users
can quickly create analysis routines that draw on specific
Organizations are looking to expand predic-
tive analytics influence to more areas of
operation and to more users

4. Sponsored by
datasets. The interface is point-and-click; no coding is
required. The tool can help you determine the best model
to use for a particular project. Additionally, the Workbench
supports the entire data-mining process, making it easier to
put this solution in the hands of more data analysts.
The results of a predictive analysis routine can be easily
shared with users who need the information to make
business decisions. And in turn, these models created by
an organization’s analysts can be run by users themselves
so they might apply the analysis routinely to new data as it
is acquired. As an example, the results can be visualized in
dashboards, reports or mobile devices making it easier to
share these insights across an organization.
SAP BusinessObjects Predictive Workbench can easily be
added to a company’s repertoire for business decision-
making. In particular, it does not necessitate the rip-and-
replace approach some other predictive analytic solutions
require. SAP BusinessObjects Predictive Workbench works
with existing BI solutions and can output findings in a format
that is easily used and shared.
When combined with the SAP BusinessObject XI platform,
SAP BusinessObjects Predictive Workbench gives
organizations the predictive analytic muscle needed to stay
competitive in today’s economy.
In particular, SAP BusinessObjects Predictive Workbench
allows organizations to:
• Overcome the limitations of traditional solutions and
make predictive analytics a part of the normal business
decision making process
• Uncover trends and patterns to solve business problems,
anticipate business changes, and make forecasts
• Efficiently discovery important and predictive findings by
way of an easy-to-use visual workflow interface
• Enable easy sharing of information with those who must
make timely decisions. n
For more information, go to:
http://www.sap.com/
The results of a predictive analysis rou-
tine can be easily shared with users who
need the information to make business
decisions