Fundamentals of Predictive Analytics | 3

Chapter 1: Introduction

Overview of Business Analytics

What makes decision making difficult and challenging? Uncertainty is probably the
number one challenge. If we know how much the demand will be for our product, we
could do a much better job of planning and scheduling production. If we knew exactly
how long each step in a project will take to be completed, we could better predict the
project’s cost and completion date. If we knew how stocks will perform, investing would
be a lot easier.

Another factor that makes decision making difficult is that we often face such an
enormous number of alternatives that we cannot evaluate them all. What is the best
combination of stocks to help me meet my financial objectives? What is the best product
line for a company that wants to maximize its market share? How should an airline price
it's tickets so as to maximize revenue?

Business analytics is the scientific process of transforming data into insight for
making better decisions. Business analytics is used for data-driven or fact-based decision
making which is often seen as more objective than other alternatives for decision making.

As we shall see, the tools of business analytics can aid decision making by creating
insights from data, by improving our ability to more accurately forecast for planning, by
helping us quantify risk, and by yielding better alternatives through analysis and
optimization (Camm, et. al., 2015).

The Analytics Life Cycle

The analytics life cycle is a series of phases that every business intelligence and
analytic project tends to go through. The lifecycle is essentially one of iteration in that
once data and analytic products have been deployed to end users the discovery that end
users make and the additional questions they ask creates the need for new content and
so the process starts again.
 Fundamentals of Predictive Analytics | 4

Analytics Lifecycle

1. Connecting to Data. Every analytic project will have one or more sources of data.
Whether these are spreadsheets, transactional systems, or data warehouses.
These sources need to be made available to the Business Intelligence (BI) tools for
analysis

2. Preparing Data for Analysis. Having established connections to data, the best step
is to prepare it for self-service analysis. In some cases your data will not be ready
for analysis - you may have data quality issues or the shape of your data is not
ideal for analysis tools to consume. In this case you will need to transform your
data by applying one or more data conversion steps to it. Through this process you
can merge multiple data sources and incorporate data science models as well.
When this is complete your data will be stored in a new set of tables and ready for
the next step.

Once your data is stored for reporting, and if you want to enable self-service
reporting or generate a lot of end user content from that data you need to define
a meta-data layer. This layer acts like a business translation layer and makes it
 Fundamentals of Predictive Analytics | 5

easy to create new content from it in a governed and secure way. Almost all BI
tools have some form of meta-data. Some of it by project or workbook or even
report-level - others have a dedicated layer to ensure reuse of logic across all your
content layers.

The meta-data layer needs to be designed to ensure it is flexible enough to

support many end-user queries but not too complex so that it becomes unusable
by typical business.

3. Build Analytics Content. Once data is prepared it is now possible to begin creating
reports, visualizations, and dashboards. The content should support the
operational and strategic needs of your end users by providing them with the data
to make decisions and to better understand the business.

Operational reports typically service functional day-to-day operational needs.

They help people to know what to react to and what to do next in their working
day. Dashboards tend to be higher level, supporting longer term decision-making
and tend to incorporate many reports - from big numbers, times series analysis
and categorical analysis. Users want to be able to filter and interact with these to
understand more completely the trends and issues in their business.

Augmenting dashboards with actions allows you to create closed-loop analytics

whereby end users can take action directly from within their dashboards. This
supports operational workflows and decision making, and ensures that business
outcomes are data-driven.

4. Analyse Manually and Automatically. Once reports and dashboards have been
deployed to end users, those people will use it to analyse their business on a day
to day basis. Content has to be easily accessible, usable and have their interactions
needed for end users to be able to slice and dice their data if they see issues they
want to explore further.

Manual analysis is one process which end users use to uncover insights through
data discovery. This includes using filters, drill down and drill through
functionality. This method’s effectiveness is, however, limited based on a user’s
data literacy, the volume of data they need to analyse, and the time and
inclination they have to analyse their data. For this reason automated analysis
solutions help to accelerate the analysis process.

5. Collaborate and share insights. Having discovered insights it is natural for the
users to want to share them with others. This is one of the most critical aspects of
the analytics life cycle. The ability to share is what drives user adoption and
increases the return on investment of a BI and analytics project.
 Fundamentals of Predictive Analytics | 6

Collaboration can occur in a number of ways:

- Through sharing a piece of content - like a report, so that others can see it
too.
- Commenting on a report or dashboard, so that you can draw other
peoples’ attention to issues or opportunities that may exist.
- Annotating a visualization - to draw a user’s attention to a spike or drop in
data with an associated explanation.
- Creating Data Stories or Presentations that combine multiple charts and
reports into a narrative that explains what happened and why, and what
actions need to be taken.

It is by collaborating with data that new questions get asked and that drives the
development of new content requirements which starts the analytic life-cycle again.

Introduction to Predictive Analytics

Predictive analytics is a branch of science and statistics understanding the patterns

from the historical data and applying these patterns to current data to understand the
future possible outcomes (Acadgild, 2018). Today organizations are sitting on tons of
data. For organizations, it makes perfect sense to analyze this existing historical data for
trends, build a model based on these patterns and apply the model to the current data to
predict the future events. Tools required for predictive analytics can include data mining
techniques, artificial intelligence, statistics, machine learning.

Predictive Analytics means

- understanding the future
- looks into the future to provide insight into what will happen and includes
what-if scenarios and risk assessments
- uses many techniques from data mining, modeling machine learning, and
artificial intelligence to analyze current data to make predictions about
future

Competitive advantage of Predictive analytics

- growing volumes and types of data, and more interest in using data to
produce valuable insights
- fater, cheaper computers
- easier-to-use software
- tougher economic conditions and a need for competitive differentiation

How Predictive Analytics Works?

 Fundamentals of Predictive Analytics | 7

The term “predictive analytics” describes the application of a statistical or

machine learning technique to create a quantitative prediction about the future.
Frequently, supervised machine learning techniques are used to predict a future
value (How long can this machine run before requiring maintenance?) or to
estimate a probability (How likely is this customer to default on a loan?).

★Predictive analytics starts with a business goal: to use data to reduce

waste, save time, or cut costs.

Predictive Analytics Workflow

Predictive modeling uses mathematical and computational methods to

predict an event or outcome. These models forecast an outcome at some future
state or time based upon changes to the model inputs. Using an iterative process,
you develop the model using a training data set and then test and validate it to
determine its accuracy for making predictions. You can try out different machine
learning approaches to find the most effective model. Predictive analytics
optimizes business processes and reduces operational costs. It deepens the
engagement of customers and enhances the customer experience in order to
identify new product and market opportunities, as well as reduces risk by
anticipating and mitigating problems before they occur. The following are the
industries where predictive analytics can be applied:

a. Energy and Utilities. Energy consumption patterns and management

b. Financial Services. Fraud identification, loan defaults, and investment
analysis.
 Fundamentals of Predictive Analytics | 8

c. Food and Beverages. Supply chain demand prediction for creating,

packaging, and shipping time-sensitive products.
d. Healthcare. Re-hospitalization and risk patterns in health-related data.
e. Insurance. Fraud identification and individualized policies based on vehicle
telemetry.
f. Manufacturing. Quality assurance optimization and machine failure and
downtime predictions.
g. Transportation. Service and delivery route optimization
h. Marketing. Consumer behavior prediction, churn analysis, consumption,
and propensity to spend.
i. Travel. Buying experience optimization, upselling, and customized offers
and packages.
________________________________________________________________________

EVALUATION

________________________________________________________________________
 Fundamentals of Predictive Analytics | 9

Case Analysis

Company Background

The company utilizes the First Tennessee Bank name in the state of Tennessee, its
primary market and where it has dominant market share. Outside Tennessee (with the
exception of suburban Memphis within Mississippi and Arkansas, and suburban
Chattanooga in Georgia), it utilizes the First Horizon Bank nameplate, the company's
official name. First Tennessee is the only major bank headquartered in the state. Until
recently, it had expanded into Northern Virginia (2003), Maryland (2003).

The bank is a full-service provider of financial products and services for businesses
and consumers.

On the path to becoming the Chief Marketing Officer of First Tennessee Bank, Dan
Marks has shown all the earmarks of being a “numbers guy.” There’s the natural sense of
skepticism—a need to see the proof in the numbers—that comes along with being the
analytical sort. Do bank marketers have a unique way of thinking, of looking at the world?
Marks affirms his belief that they do, but with a caveat. “Within banks, marketing people
tend to be great at seeing opportunity and conceptualizing new ideas,” explains Marks.
“But in today’s banking market, it’s increasingly important to view them with a critical
eye, to discern where it makes most sense to focus the bank’s resources. That’s the
balance—between creativity and discipline, between art and science— that we need to
strike.”

What Marks had in mind went beyond budgets to the very heart of First
Tennessee’s marketing processes, with granular metrics and analytics providing the basis
for optimization. “Our aim was to shift from the ‘marketing-as-an expense’ mindset to
the idea that marketing is a true profit driver.”

The market that Marks is talking about is defined not only by increasing
competitive intensity, but also by the strategic challenges it presents to banks, not least
of which is how to optimally focus marketing resources. Banks offer a more diverse
portfolio of services than before, and they do so over a wider range of channels. While
this trend has given banks more latitude to compete, it has made formulating and
modulating marketing strategies, tactics and programs considerably more complex.
That’s because as the range of options has grown—a good thing by any measure—
 Fundamentals of Predictive Analytics | 10

marketing resources are as scarce as ever. To optimize how they invest them, banks need
a way to continually measure their effectiveness, learn what works and adapt over time.

For banks today, having more ways to communicate with customers is a good
thing. But it has also made it harder for banks to figure out where and how to most
profitably commit their marketing resources. Leveraging predictive analytics, First
Tennessee Bank is applying the ultimate acid test. It’s combining a granular understanding
of the needs of customer segments with real P&L data to optimize its marketing spend,
focusing on programs that deliver the highest ROI. First Tennessee’s ability to target its
campaigns more intelligently has increased its response rate by 3.1%, cut key marketing
costs by nearly 20%, and enables the bank to get the most from its resources.

Selling First Tennessee’s President of Banking on the idea wasn’t hard. Putting it
into place required action on a number of fronts. Conceptually, think of an upside-down
pyramid, with the technical capability to do predictive analytics at the bottom,
underpinning the effort. Though it’s an essential foundation, its enablement represented
a relatively modest share of the effort. More extensive (and the next level up in the
pyramid) was the need to secure the business intelligence (from the bank’s data
warehouse and finance organization) and from that develop the framework for ROI-based
modeling. Based on product parameters like fees, spread and account balances, the
model would create tiers of profitability for different kinds of products and accounts
and—by extension— different segments of customers.

At the top of the inverted pyramid was Mark's biggest challenge and, in many
ways, the most important ingredient to his success— driving change at the process level.
In this context, the key processes relate to choices made by marketing managers and
within the lines of business around program funding.

Should, for example, more funding be put into customer acquisition activities like
lead generation? Or should the funding emphasis be on retention and cross-selling
opportunities? Traditionally, such decisions were the epitome of “in-the-box thinking”—
guided by a mix of past experience, intuition and conventional assumptions about where
the opportunity is.

Mark's aim was to promote a new way of thinking about opportunity that would
permeate the entire organization. “The message we’re getting across to our people is a
new way of managing and optimizing our marketing resources around ROI—one that
looks ‘under the covers’ at the relative profitability of all these programs and uses that as
the basis for decision-making,” says Marks. “We understand that the most effective way
to drive this change is not ‘top-down,’ but through a dialogue that moves all of us—
including the LOBs—towards this new way of thinking.”

First Tennessee used to structure its marketing campaigns around product lines.
Over the last few years, the bank has developed a highly systematic and targeted
 Fundamentals of Predictive Analytics | 11

approach. Here’s how it works. Start with a granular understanding of each customer’s
banking needs drawn directly from lots of customer data points. Using predictive analytics
models developed by Marks’s staff, each customer is “scored” on their likelihood to
purchase each product in First Tennessee’s portfolio. The corollary benefit of this
approach is that it helps the bank’s marketers to pinpoint product clusters that represent
“sweet spots” for cross-selling opportunities. That’s just the beginning. What sets the First
Tennessee approach apart is how it applies a rigorous, systematic approach to prioritizing
which opportunities make it to the campaign stage. By combining product revenue and
cost information from its data warehouse with the segment data discussed above, First
Tennessee’s model generates a quantitative measure of the expected profitability of a
given product offered to a specific subset of its customers, such that each
product/segment offer under consideration is assigned an expected ROI value. With this
hierarchy established, Marks and his team now have an evidenced-based framework
through which to prioritize programs and allocate resources accordingly.

The bank’s performance numbers also underscore the effectiveness of ROI-based

optimization. For instance, the rate of response to its marketing campaigns has risen 3.1
percent, a reflection of its ability to more accurately target offerings to specific customer
segments based on their needs. First Tennessee’s recent growth in market share,
producing gains across most of its geographic footprint in and around Tennessee,
provides yet another window on its success. Overall, the bank has tallied a 600 percent
return on its investment in predictive analytics through more efficiently deployed
resources. By gaining the ability to target the most attractive segment for specific offers—
a “quality over- quantity” approach— First Tennessee has been able to optimize its
campaign expenditures, evidenced by a 20 percent reduction in mailing costs and a 17
percent reduction in printing costs. Looking down the road, Marks expects tight resources
and intense competition to be par for the course in the banking market. But with
predictive analytical capabilities in place, he sees these conditions as only heightening the
bank’s hunger for opportunity, since it has the ability to pursue it efficiently. “We’re
committed to profitability and we're committed to strengthening our customer
relationships through every aspect of the way we do business,” says Marks. “Predictive
analytics gives us the intelligence and insights we need to follow through on this
commitment.”

Business Objectives
1. Identify areas where marketing resources should be allocated
2. Identify which services to offer to clients

Questions

1. Identify how predictive analytics was used to solve the business problem. Explain
how the predictive analytics solution works.
2. What were some of the predictors used for the predictive analytics model?
3. How did Dan Marks build an analytics culture?
 Fundamentals of Predictive Analytics | 12

Copyright IBM Corporation 2013 IBM Philippines

IBM Plaza, Libis, Quezon City May 2013

References:

YellowFin Guide (2020). The analytics lifecycle. Retrieved from

https://www.yellowfinbi.com/evaluation-guide/developing-content/the-analytics-
lifecycle

Acadgild.
(2018). An introduction to predictive analytics: using data science to predict the
future. What is predictive analytics? Retrieved from https://acadgild.com/blog/an-
introduction-to-predictive-analytics

Camm, J., Cochran, J., Fry, M., Ohlmann, J., Anderson, D., Sweeney, D., and Williams, T.
(2015). Essentials of business analytics. Cengage Learning Asia Pte Ltd. 1st edition.