Predictive analytics

EXECUTIVE SUMMARY
The calculus of marketing has become complex as an increasing number of variables enter into
the equation of consumer acquisition. Fragmentation of product markets, media and channels
and their inter-dependence has added several imponderables in the planning and execution of
marketing strategies. If companies continue with their old practices of mass marketing, they are
more likely to invest large sums of money that will not yield a rate of return.
The rate of return from marketing investments can be raised with an intelligent use of information.
A growing number of companies use customer data for segmentation and learn to target groups
of customers with personalized service. Their ability to achieve their goals improves as they learn
from the data they receive from past marketing campaigns.
The costs of serving customers have increased and companies need to increasingly cater to
micro-segments to gain a competitive advantage. They have to take into account not only
demographic and socio-economic variables but also behavioral traits that help to segment at a
more granular level. Competitive differentiation is achieved by more than product differentiation;
an array of transactional conveniences as well as nuances of relationships provides the edge
companies need.
Business intelligence goes further and uncovers opportunities and alerts companies to
competitive threats. Data points to needs that are often overlooked in the course of mass
marketing. Similarly, companies can discover in the patterns of data a lurking threat from a
competitor and take prompt action to pre-empt it.
In an environment of uncertainty that fragmentation has caused, predictive analytics helps to
anticipate with some degree of accuracy the outcomes that can be achieved from the products,
services, process benefits and relationships that companies offer. The virtuous circle of data
based forecasts and feedback from marketing campaigns lowers the error rate in decision
making.
Increasingly, companies have to choose the moments when the customer is most receptive to
their messages rather than overwhelm them. These opportunities come when customers decide
to contact call centers or are receptive to relevant information that contributes to their participation
in events rather than those jarring calls from telemarketers and frustrating mass mailings. Call
center agents and sales people have to learn to make impromptu offers by reading the mood of
the customers. They are more likely to be effective if they are armed with information about the
profile of the customers.

In sum, predictive analytics is about achieving better results with lower costs. The key to
achieving this goal is to choose the most relevant product, service, channel and media to reach a
specific segment of the population.
INTRODUCTION
In the early years of this decade, the euphoria over Customer Relationship Management (CRM)
software, widely shared in the 1990s, evaporated and consumers publicized their despair over
the poor financial results realized from their investments akin to the disillusionment with other
information technologies. With the benefit of hindsight, customers can now see that the early
CRM technologies had a modest objective of accumulating transaction data. The truth is that the
“irrational optimism” about CRM clouded judgments in the 1990s. The “irrational pessimism” that
ensued missed the promise of CRM, i.e., the ground had been prepared for decision support
solutions including predictive analytics.
On the rebound from the bubble pop, an increasing number of customers realize that the payoff
from the investments in CRM will come from the efficiencies that can be realized from the
surefooted implementation of business strategies. Gut feeling and intuition is giving way to
statistical forecasts as large data sets and their analysis with a new generation of analytical and
forecasting software helps to make decisions with more predictable outcomes.
Business uncertainty has increased in recent decades with globalization, technological and
demographic change. Consumers have many more products to choose from and product
obsolescence is much more rapid. Companies cannot any longer expect to dominate mass
markets and must learn to select activities that will be most profitable given their competencies
and resources.
Unlike in the past, mass media is not any longer able to hold the attention of large numbers of
consumers. Network television faces intense competition from cable and satellite television.
Newspapers and magazines must cope with the Internet and Blogs. Companies have other
means of promoting products such as public relations, events and viral marketing. These
channels interact in a variety of ways that were not foreseen in the past.
Similarly, companies must choose from a host of channels for distribution of their products.
Besides the traditional channels like direct sales force and convenience and department stores,
e-commerce offers several different choices.

Fortunately, business intelligence and analytics software enables companies to navigate their
way in this environment of pervasive clutter. Unlike in the past, companies cannot any longer
afford the trial-and-error methods since they would have to conduct several experiments before
they find one that works for them. Instead, they need to be able to predict, with acceptable levels
of accuracy, the customer segments, the products, media and distribution channels that will be
most profitable.
Armed with the data and analysis they need, companies can now conduct targeted marketing
campaigns with better results. The data gathered from marketing campaigns helps them to
improve the quality of their databases, make better forecasts in the future and improve their
returns from customer acquisition, retention and extension.
PREDICTIVE ANALYTICS AND CRM
Customer Relationship Management software was a rage in the late 1990s but enthusiasm for it
began to wane in the 2001-03 period. An estimated 50% of the CRM projects did not yield a
payoff. While CRM was admittedly a fine tool to keep track of account and transaction
information, it did not yield a profit due to its inability to contribute to actionable conclusions. The
picture began to change when predictive analytics was added to the menu of functions available
with CRM. Increasingly, companies recognize that predictive analytics is an indispensable tool for
decision-making.
Predictive analytics has proved to be valuable in an environment where uncertainty has increased
because of a wider array of means available for companies to promote their products and
services. Cars, for example, can be promoted by showcasing them in malls, advertising in a
glossy magazine or on network or cable television. Increasingly, companies are realizing that they
need data on customer responses to each of these means of promotion.
They also have more options for channels, media, services and products that they can offer.
These options are often inter-dependent and companies need to plan how they can work
cohesively. For example, new products, where customers need to learn new features, are better
promoted by call centers supported by technically adept staff. Countless communities on the web
present opportunities for companies to advertise to their target audience. Companies must find
the means to evaluate each of the options they have to sell their products and find the data to
measure their effectiveness.

Customers also can articulate their individual needs are not satisfied with the staples that were
common in the past. Companies need information on behavioral traits of customers that underlie
their distinct needs so that the most relevant products and services are offered to each segment
of the population.
Mass marketing is a costly means to promote products and overwhelms customers who are
exposed to the din of growing numbers of marketing messages. People are so tired of
advertisements that they are not paying attention. A recent study by Yankelovich Partners, a
marketing-services consultancy, found 65% of people now feel they are swamped by ad
messages and 59% feel that ads have very little relevance to them. Almost 70% said they would
be interested in products or services that would help them avoid marketing pitches.
Unsurprisingly, a recent Deutsche Bank study of effectiveness of TV advertising on 23 new and
mature brands of packaged goods and concluded that there was a positive cash flow in 18% of
the cases. Over a longer term the picture improved, with 45% of cases showing a return on
investment. Much of the positive cash flow was accounted for by new products suggesting that
innovation was more important than advertising.
Predictive analytics helps companies to evaluate the most cost effective channels and media as
well as the communication messages for specific segments of the population. Companies that
excel in marketing, such as Gillette and Pepsi, who in the past relied greatly on television
advertising, have recognized that tech-savvy 12-24-year-olds do not respond to television as well
as the ageing “baby boomers.” Recent product launches of Code Red Soda (Pepsi) and Venus
Razors (Gillette), meant for young women, changed their tactics and reallocated at least half of
the marketing dollars from television to interactive games, viral marketing programs and media
that this younger generation enjoyed.
Gillette placed Web applets on teen sites to draw the elusive teenage girls (as they begin to
shave) to learn about and interact with this new brand. Pepsi realized that technically
sophisticated young men are drawn to interactive game contests and it could attention by offering
cases of the Code Red soda as rewards to the winners. Code Red was able to achieve the sixth
highest soda sales (2.2% share) in convenience stores with relatively little television advertising
compared to other Pepsi product launches.
Predictive analytics goes beyond the traditional CRM methods to find the patterns in the data that
is collected. It identifies segments or affinity groups among customers, it seeks to determine the
causes of observed patterns of purchasing behavior and evaluates the results of marketing
campaigns to target customers more efficiently in the future. The analysis helps to identify

specific channels and media most relevant for individual segments of the population which lowers
the costs of acquiring new customers and to retain them.
CenterParcs, a European travel management company, believes it can forecast when a customer
will book a holiday, its location and the duration of the stay. It is able to foretell travel behavior by
using its predictive analytics software which has helped it to reduce its direct mailings to a quarter
of the earlier level even as it has increased revenues at the same time. It also claims an average
occupancy rate of around 90% in its holiday homes around Europe.
None of these techniques would be useful without large volumes of data for numerous series.
CenterParcs, for example, draws on more than 100 million customer records dating back as far
as 1982 and combines these with external data sources such as demographic or geographical
information. It uses information on 60 to 80 variables to estimate the probability of them booking a
holiday with them for a destination at a certain time of the year.
Companies can use a broad range of techniques to predict outcomes with increasing accuracy.
While numerical data was common in the past, large quantities of textual information can also be
used now to predict consumer behavior. This is particularly useful to identify the pain points that
help to find new product and services. A great deal of the data received by call centers, for
example, is in the form of conversations with customers. In terms of analytical techniques,
companies have a choice between artificial intelligence techniques such as neural networks
which find patterns in the raw data. On the other hand, linear regression models are useful for
finding causation and prediction. Techniques such as logistic regression can estimate the odds of
a customer buying a product or the risk of default. Finally, methods such as time series analysis
find patterns in data over a period of time.
For more information
http://www.economist.com/business/displaystory.cfm?story_id=2787854
http://www.infoconomy.com/pages/information-age/group65470.adp
http://www.marketingprofs.com/4/diorio2.asp
http://www.dbta.com/frontpage_archives/7-03.html
HOW DOES PREDICTIVE ANALYTICS HELP?
Marketing has increasingly become information driven function as companies have to take
decisions on when, how and where to serve their customers. They need to make decisions over a
longer time frame since the costs and benefits of servicing customers are not necessarily

matched at a given point of time. Also, they have to plan for the combinations of products,
services as well as channel and media they want to use to reach their customers. Besides looking
at the demand side, companies have to find the means to design their supply side especially the
logistics of meeting the demand. Predictive analytics contributes by estimating the unknowns in
all these calculations.
Customer Acquisition
Customer acquisition has become an increasingly complex task as companies have to consider
the value of their customers over a life time. A great deal of investment in acquisition of
customers is lost when they defect later. On the other hand, some customers may not be
profitable in the short-term, such as young professionals in the financial services industry, but that
their loyalty, gained early, can reap a bonanza later. PriceWaterhouseCoopers, for example,
estimated that promotional costs of as much as eight hundred pounds on young professionals
were worth the gain in their net present value over a life time in the United Kingdom.
Companies have to learn to carefully choose the customers they want to serve taking into
account the costs of servicing them including the gains and losses in the present and the future.
They are more likely to achieve this purpose with much granular levels of segmentation than has
been the case. Hitherto, companies have typically segmented the population based on income,
age, location, ethnic group and relative preferences for price, quality, convenience and
functionality of products.
For the future, individual companies are more likely to gain an edge if they identify behavioral
traits that create unique opportunities for positioning. Till recently, for example, overweight
women were not considered as targets for fashionable clothing which most strategists assumed
were reserved for slim women. The mindset changed and apparel manufacturers were able to
design sexy clothes for them and gain market share.
Customer Retention
Companies become aware of customer dissatisfaction usually after they defect when they can do
little to reverse the damage. Retention of customers costs less than acquiring them, $280 to
acquire compared to $57 to retain an existing customer according to estimates of Gartner Inc., so
it helps to take preemptive action to avoid their loss. Predictive analytics provides insight about
the motivation for leaving a vendor. What is more critical, predictive analytics can forecast

defection by tracking the behavior that eventually leads to loss of a customer. Companies can
take advantage of this information to make offers to dissuade a customer from leaving.
Fifth Third Bancorp of Cincinnati was a typical case which lost nine customers for every ten of
them it acquired. It installed a CRM and a predictive analytics system which had ninety variables
to compare daily transactions with customer’s history to uncover telling signs of a displeased
customer about to leave. The same system had also in-built triggers to make new offers to win
back the customer. By the end of the six-month pilot, Third Bancorp had achieved a 400% return
on investment, had cut new-account attrition by half and reduced overall household attrition by
nearly a third.
Contact Management
Customer touch and respect is critical for companies to retain their customers. Companies have
to be able to make their sales pitches when they are most effective to avoid the irritation
customers experience when they are inundated with messages. Call centers, in the past, were
considered a cost center and the focus was on lowering the time that was spent with customers.
In an environment of clutter and rock-bottom patience for sales pitches, companies now welcome
the opportunity they get when customers voluntarily call them. They like to extend the time spent
in order to explore needs and to make offers that are most appropriate for the current needs of
the callers. According to one recent survey of 6000 marketing and IT executives, 41% of them
believe that customer facing employees have access to data they need to service their
customers.
HBOS, formed by the merger of Halifax Bank and Bank of Scotland, found that inbound calls
proved to be invaluable especially when it decided to use predictive analytics to make impromptu
offers to callers. It decided that it needed to provide its telephone customer service
representatives the tools to cross-sell the company's diverse set of financial products whenever
an in-bound call was received. The customer service agents receive on-screen cross-sell prompts
based on the analysis of each customer's individual profile. Halifax then expanded the
deployment to a total of 1,000 contact center agents and added a second channel, which proved
to be even more effective. Halifax realized a 10-15 percent increase in new leads with close to
half of those leads resulting in sales.
Observations of customer behavior across channels provide a hint of specific interests of
customers. Many customers browse through information on web-sites before they decide to call a
company to discuss specifics of products and make a decision to buy. Today, call centers have

access to data such as that displayed by the dashboard of Insight RT which summarizes
graphically information about the browsing behavior of specific customers.
Channel Management
With proliferating channels for marketing, companies have to carefully assess the precise impact
each of them have in informing, motivating customers and acting as a conduit for sales often in
unsuspected inter-dependent ways.
One example of use of analytical methods is the case of home furnishings retailer Restoration
Hardware Inc. which reinvented itself from an entirely store-based operation until 1998 when it
added catalog sales and a web site. Sales quadrupled between 1998 and 2002 but this was
offset by higher costs of especially catalogs with a loss of $34 million in 2002. An in-depth
analysis arrived at a surprising conclusion that 40 percent of online purchases were linked to the
catalog, and these customers spent 30 percent more than other web shoppers. Also, consumers
who received the company's catalog spent 25 percent more in its stores than those who didn't.
The catalog was repositioned for higher value items and their focus was increasingly on existing
customers and those who had the potential to be repeat customers.
Advertising Efficiencies
Advertising will remain ineffective as long as messages are lost in the clutter. Increasingly,
companies have to struggle to attract the attention of customers. The key is to target customers
who have a demonstrated interest in the message. American airlines worked with Wall Street
Journal to appeal to business travelers who are less likely to be influenced by lower prices that
competitors like Jet Blue generally offer.
The subscribers of Wall Street Journal Online, who read its travel columns were considered as
potential customers of American Airlines and were served with its advertisements. Based on the
extent of interest they showed in the columns, the readers of these columns were segmented as
infrequent or frequent travelers. The number of infrequent travelers who viewed the ads
increased by 115% and by 145% for the frequent travelers. In addition, the rate of recall by those
who viewed the ads increased by 314%.
Events trigger sales campaigns
Consumers are often motivated by events when they decide on purchases. Special occasions
create the mood and the motivation for exceptional buying. Many buying decisions are made at
the time of marriage, birthday celebrations, relocation, and move to college or at the time of

traveling. Companies need to be able to keep track of the life history of their clients, gift giving
behavior as well as the age of their children to be able to predict spurts in buying. According to a
study conducted by Gartner Inc., offers triggered by events, on a monthly basis, generate
response rates ranging from 4% to 5%, which rises to between 16% and 50% when the system
can respond to customer activity triggers every day. This compares with modest response rates
of between 2.3% and 3.3% for traditional telemarketing campaigns
One case of astute use of events information for sales campaigns is Fidelity Investments which
identified over 100 such triggers that signal when customers would need to enter into trades, shift
to different investments, or move their assets to another investment company. Fidelity used
technology to sift through millions of daily customer transactions to identify when life and market
events were happening, and target occasions when customers would be receptive to advice or
new products or to a particular offer. These events triggered email, Web site, agent and call-
center programs that placed the right offer in front of the right person at the right time. Fidelity
could create twice as many qualified leads, doubled the chance that a particular offer was
relevant to a customer’s need and improved campaign response rate 200%.
At this point of time, a minority of companies use event or trigger based marketing campaigns
which leaves considerable scope for gaining competitive advantage. A recent survey of 6000
marketing and IT executives found that only 19% of companies take advantage of events to sell
their products.
Locating Customers
Demand patterns of customers are determined, to a large extent, by their location. Ethnic groups,
for example, tend to congregate in some regions. People of different age groups are concentrated
in some locations; the elderly increasingly prefer college towns or warmer regions. Similarly,
people with similar psychographics tend to prefer particular locations depending on their taste for
culture, outdoor activity, pace of life and social networking. Vendors can correlate geographical
information with consumption data to make decisions on stocking, ad placement and events to
target their customers.
Meineke Muffler, an auto repair chain, correlates psychographic and demographic data with
information on motor vehicles and average spending figures for exhaust, shocks and struts based
on records from state departments. This data helps in determining how many people and cars are
within a three-mile radius of a prospective site. Meineke uses the same information to design ad

campaigns and determine an optimal mix of inventory. The parts stocked in the shops are of
makes of cars that are found within the three mile perimeter.
http://www.cioinsight.com/article2/0,1397,1458008,00.asp
http://www.marketingprofs.com/4/diorio2.asp
http://www.economist.com/business/displaystory.cfm?story_id=2787854
http://www.bai.org/bankingstrategies/2004-jan-feb/real/index.asp
http://www.kmmag.com/articles/default.asp?ArticleID=68
http://www.callcentermagazine.com/GLOBAL/stg/commweb_shared/shared/article/showArticle.jht
ml?articleId=17600785&pgno=3
http://www.csc.com/solutions/customerrelationshipmanagement/knowledgelibrary/uploads/1530_
1.pdf
http://www.pwcglobal.com/uk/eng/about/svcs/cvc/pwc_Cust_Value_Report02.pdf
COMPETITIVE ADVANTAGE AND PREDICTIVE ANALYTICS
Predictive Analytics has increasingly become a tool for companies to gain competitive advantage.
Access to information and its analysis lowers the uncertainty of business by anticipating emerging
business opportunities, reduces search costs and helps to zero down on prospects and
eliminates costs of routine services to customers. Increasingly, companies are able to forecast
demand for products and services based on attitudes, events and behavior of customers. They
are also able to uncover needs of customers and reach out to them without waiting for them to
arrive at their stores. The accuracy of their forecasts improves as marketing campaigns throw up
more data and help to validate or invalidate their assumptions. Finally, companies are able to
lower operational costs by offering them the most relevant products and services.
One example of the changing fortunes of companies can be seen in the competition between
community banks and the larger banks such as Wells Fargo. Large companies, in the past, often
lost the initiative to smaller companies who had a better understanding of customers and an
ability to tailor products and services for their needs. In the banking sector, the community banks
had an edge in lending to smaller enterprises. A personal touch, less cumbersome procedures for
granting credit, willingness to provide small size loans and advice available from community
banks enabled them to corner most of the market for lending to small business. In more recent
times, the tables have been turned. The larger banks huge larger data warehouses and the
information to estimate credit risk; they find their prospects based on their credit scores thus

saving the transaction costs that were incurred when bank officers sought referrals, collateral and
analyzed financial statements from owners of small business.
Now it is possible to pre-approve a loan without meeting the small business owner so that the
large banks can do business in regions where they don’t have branches. In l993, Wells Fargo
pioneered credit scoring in small business lending. It’s lending to small business leapt to $l08
million in l995, a 6l percent increase from l994. The bank’s officials claim they can profitably make
small-business loans of less than $5,000 and even adjust credit lines and interest rates based on
the computer’s assessment of the risk. In addition, they are able to extend loans in virtually all 50
states, even though the bank had no branches outside California until its purchase of First
Interstate in l996.
Competitive Intelligence
Predictive analysis also enables companies to detect strategies of their competitors before
irreversible damage is done. Companies can study consumer preferences and any significant
trend that shows a shift in favor of their competitor can enable them to take pre-emptive action.
One example of this is the case of a company in the financial services industry which found its
competitor expanding its business in college campuses. To begin with, the bank found that 10
percent of the bank's customers were incurring 90 percent of the ATM costs. In addition, data
mining indicated that this 10 percent of the customer base was also composed of college
students; a good 30% of them were students. Further investigation revealed that the competitor
has an exclusive presence on campuses and was targeting college campuses for new
installations of ATMs in order to gain their loyalty that would yield benefits when they became
adults.
Customer Touch
The human interaction between a customer service agent and the client can facilitate
personalized selling that would otherwise not be possible on the Internet or by direct marketing.
Customer service agents have to be able to make impromptu offers after sizing up the mood of
the client and his or hers profile. Increasingly, companies are empowering the customer service
agents to predict purchasing propensity based on the behavior of customers in the past. The
results of predictive intelligence are displayed on dashboards that customer service agents can
access and read metrics, such as scorecards showing the customers propensity to buy, which
can help them to determine the time they want to spend with them.

The benefits of using predictive analytics in outbound call campaigns have been encouraging
thus far. Companies have been able to reduce customer churn by an average of 4-percent while
they have been able to increase sales by 10-20%. A typical example of the use of predictive
intelligence, in a telecommunications industry, would be to offer new services such as unified
messaging to customers who already have broadband access.
Affinity Marketing
The term, product differentiation, has lost its descriptive value as marketing strategists realize that
there is little to distinguish products, per se, from one another. New products, or their variations,
are quickly reinvented or reproduced by competitors blunting the edge that any one of them might
have had. In the automobile industry, for example, data gathered by McKinsey, a management
consulting firm, found that 65% of customers believe that the options available to them are alike.
Increasingly, companies bond with customers as people; they relate to their attitudes and
behavioral traits. Also, they understand the pain of the customers and provide them with
conveniences to encourage buying. In short, companies offer value propositions to their
customers instead of products.
The traditional forms of marketing, such as advertising on national networks or similar forms of
mass marketing are less and less effective in winning over customers. Increasingly, companies
segment their customers into affinity groups who have similar needs and tailor their messages,
channels and services to suit their profile. They redesign their loyalty programs to provide a
bundle of inter-related products and services that meet their needs. For example, frequent
travelers hanker for not only discounts on their tickets but also look for deals on related products
like hotels, restaurants, rental cars and tour packages.
Predictive analytics helps companies to identify the segments that help marketing strategists to
design products and services for individual groups of customers. All too often, the data available
with companies for analysis is not accurate or complete. They can, however, use the information
available to begin testing their hypothesis about consumer behavior and plan their marketing
campaigns. As an example, an apparel company can begin with the assumption that stylish
clothes are more likely to be in demand in New York compared to West Coast where casual
clothing is preferred. If a marketing campaign is targeted in only that region, companies will learn
more about the actual clothing habits in that region. Information of this nature affords a
knowledge advantage to companies which are more durable than simply a distinctive product.

Fingerhut, now a part of Federated stores, is a direct marketer who has catalogs printed format
and on the Web. It produces over 130 different catalogues that are based on the segmentation
models and propensity to buy findings derived form its six terabyte plus data warehouse that
monitors more than 65 million customers especially the most active 12 million. The company
studies about 3,500 variables over the lifetime of a consumer's relationship with it. Fingerhut
categorizes customers into affinity groups large enough to warrant a catalog with information and
list products meant for their specific needs. They found, for example, that people who change
residence triple their buying in the 12 weeks after the move, with most of that in the first four
weeks. Furthermore, they buy furniture and telecommunications but not jewelry and home
electronics. The result is a "mover's catalog" sent to these people during the 12-week window. To
save marketing cost, no other catalog is sent to them.
Process benefits or conveniences associated with selling are another means to differentiate a
product or service. Intrawest, one of the largest ski-resort companies used its CRM database to
address the complaints, of a large number of its guests, about the long time it took to be fitted for
skis and boots which could have been better spent on the slopes. To speed the process,
Intrawest collects the height, weight, and preferred ski length and boot size of all guests in
advance. When guests arrive to pick up their equipment, they find skis and boots in the correct
sizes all laid out for them. Customers can book accommodations, sign up for lessons, and
procure lift tickets online. All the key information is captured on a database. By understanding
what customers are looking for, Intrawest can develop the most appropriate packages and make
the overall experience smoother and more relaxing.
Operational Efficiencies
Predictive analytics enable companies to see the future with a relatively higher level of certainty
and lower the attendant costs. Without the benefit of the estimates that predictive analytics
provides, companies overestimate their risk which is reflected in relatively higher costs for their
clients.
An example of how a company could offer better terms to its clients is the case of CustomerLinx,
a company that specializes in marketing campaigns. Traditionally, marketing campaign
companies have billed their clients by the hour regardless of the outcomes achieved or the
benefits that the client obtained. CustomerLinx took a bold initiative to bill their clients based on
actual sales booked by their clients. This was possible because CustomerLinx could forecast,
with a known level of accuracy, how many prospects would actually buy products.
http://www.marketingpower.com/content24421.php

http://www.crito.uci.edu/itr/publications/pdf/survey_db_mktg.pdf
http://www.optimizemag.com/article/showArticle.jhtml?printableArticle=true&articleId=17700882
http://faculty.cs.byu.edu/~cgc/Teaching/CS_601R_W05/Success%20Stories%20in%20Data%20
Mining.pdf
BEST PRACTICES
Text Mining
Customer relationship data is available not only in quantitative form such as sales data, returns,
pricing, inventory, and regional dispersion of demand. A great deal of more valuable data is
available in the conversations with customer service representatives, the notes sales
representative make, chat sessions and contracts or patent documents. This kind of qualitative
data can help assess pain points expressed in the complaints of customers, feedback or
suggestions, preferences expressed as well as news articles. Till recently, it was hard to mine
textual data but increasingly vendors are able to include this capability along with statistical
analysis functions. The correlation of textual information with quantitative data helps to extract
insights which would otherwise have been elusive.
Hewlett Packard has been one of the pioneers in the use of textual and quantitative information
for understanding product needs of customers. When HP combined its information on customer
segments with the textual information it was receiving, it realized that the feedback from individual
segments was not the same. The hot button issues concerned product configuration and pricing
issues. Subsequently, HP was able to tailor solutions for each of these segments based on the
analysis. In addition, this information was used to construct predictive models which were applied
to prospect databases to target new customers. HP decided to extend the scope of its text mining
by including the notes taken by its sales staff on Siebel note pads and later included articles from
newspapers and magazines.
Predictive modeling can help companies take pre-emptive action to avoid harm to their brand
equity. J.D. Power and Associates, a California based customer research firm, is testing models
that will sift through comments from surveys to predict warranty problems, for automobile
manufacturers, before large number of vehicles have been shipped. Nextel Communications
faces the problem of customer churn as acutely as other telecom companies. It ferrets out the key
phrases in customer interactions to predict customer churn and make offers to avoid such a
situation.
Geographic location of On-line Customers
Growing e-commerce poses a challenge to companies who need to identify their customers
before they can place ads, make offers and provide service. Shoppers on web-sites are known
only by their IP address which could be located anywhere in the world. In the absence of

information on the country the customer belongs to, companies don’t know even the language
they should be using.
Digital Island of San Francisco has developed an Internet atlas application called TraceWare,
which correlates IP addresses at the country level. As an international Internet backbone
provider, Digital Island is in a position to gather the geographic information to determine where
the IP addresses originate.
HighWire Press, a publisher of more than 150 life science journals at Stanford University, finds
TraceWare useful in placing pharmaceutical ads. Pharmaceutical industry regulations vary in
individual countries; some countries prohibit advertisement in the industry and others don’t.
TraceWare helps in deciding where to place ads.
Customer Life Time Value
Economic Value Added was an esoteric concept, long ignored by managements, till Coco Cola
used it with spectacular success. The intuition underlying the concept was simple; a company
adds value only when its profits exceed the cost of capital. Coco cola refocused its business by
disinvesting units that did not meet this criterion or their economic value was lower than in other
departments. Customer Life Time Value has similar implications; acquisition of a customer does
not add value to the company unless the costs of servicing a customer are less than the value
that is added in terms of profits from sales. In practice, determination of value addition by each
customer is hard. Over a life time, a customer buys several products and services and the initial
sales create a beachhead for further promotions. A customer for broadband services, for
example, can become a customer for internet, unified messaging and a home network. On the
other hand, a customer could be acquired at a high cost by selling a DSL for no cost or at a
discounted price without yielding a benefit in the future.
Companies need to find a way to segment their customers so that they are offered a value which
is in line with the costs incurred to service them. One company which has adopted this approach
is Best Buy which has concluded that 20 percent of its customers are unprofitable. The profitable
customers are groups of customers like the suburban mothers and the upper-income men. Sales
people in fifteen percent of its stores are trained to better tailor to their needs. The pilot stores are
gaining sales at twice the rate of same-store sales and higher close rates as conventional stores.
Best Buy expects to roll out the customization program to the rest of their stores over the next
three years. Predictive Analytics helps to identify the characteristics of customers who are more
likely to be profitable.

Progress in Techniques
Traditionally, marketing managers have been content to use the RFM (Recency, Frequency and
Monetary Value) of purchases as the method to segment their prospects. This provides a
rudimentary and intuitively appealing way of classifying customers for promotional purposes. The
advantage of this method is that it requires only sales data to segment customers. It does not,
however, correlate sales data information with geo-demographics, psychographics or economic
conditions to estimate the likely purchases by a particular group which can improve the accuracy
of the predictive models.
Classical statistics is proven method of data analysis especially when testing hypothesis about
causation between variables. The most common methods that are use in marketing are linear
regression and logistic regression models, the former estimates integer numbers while the latter
calculates the odds of an event happening. These methods presume a probability distribution in
the data, a normal distribution or a bell curve is the typical assumption, when tests of significance
or validity are conducted. However, this assumption is hard to sustain when the data sets are
extremely large and countless variables interact with each other in complex ways.
Data mining methods are an alternative means to parse the data without making any assumption
about the probability distribution of the data. A common denominator of these methods is that
they look for patterns in large data sets without making an attempt to find the causal interactions
in the variables. These methods use artificial intelligence algorithms; a common method is neural
networks, to cluster the data into segments where data points with common characteristics are
separated from others.
One simple data mining method is the nearest neighbor method of clustering which is akin to
predicting a person’s buying behavior from their place of residence. People living neighborhoods
that are highly priced will tend to have higher level of education and are more likely to spend in
stores such as Williams-Sonoma. Similarly, people living in retirement communities are more
likely to travel. These methods require no mathematical calculations as is the case with classical
statistical techniques and are intelligible to most people. The downside with this method is that
the inferences are not as rigorously tested as is the case with classical statistical methods.
Similarly, segmentation of the customer database can be done with a simple method like decision
trees without using complex mathematical techniques. For example, a company may divide its
customer base into two groups; the first it is able to retain for the first year only and the rest who
defect after that. It could further divide those who are prone to defect into two groups; the first

responds to lower price and the other to better products. This could continue depending on the
granularity that is desired.
More sophisticated data mining techniques are methods like neural networks which mimic the
human brain’s tendency to learn and improve estimates based on the data received. This
involves making tentative estimates which increase in precision as more data is received. As an
example, a company might want to estimate the probability of a person clicking on an
advertisement for a sports product on the web. The neural network is required to estimate this
based on data on the age of a person, location and ethnicity. The estimates of the output are
compared to the actual data and if there is a variance more variables are included till a
satisfactory result is achieved.
For further reading
www.computerworld.com/printthis/2005/0,4814,102375,00.html
http://www.kmmag.com/articles/default.asp?ArticleID=68
http://www.leggmasoncapmgmt.com/pdf/TheEconomicsofCustomerBusinesses.pdf
http://www.research.ibm.com/journal/rd/471/apte.html
TECHNOLOGY THE ENABLER
Technological requirements for business intelligence and predictive analytics applications involve
a generational leap. The change comes as a result of the multi-dimensional nature of the beast.
Operational CRM was content to manage single dimensional information such as a sales
person’s leads, pipeline and performance at a point in time. Business Analytics add complexity by
requiring historical data to be able to compare performance over a period of time. This could go
further if the company requires performance comparisons across regions. In addition, the data
could become cross-functional as companies look at the financial impact of sales performance.
The volume of the data that has to be managed grows exponentially as the complexity of the
queries grows. Finally, the very nature of predictive analytics is to conduct a variety of different
queries which conflict with the more static design of CRM applications. The need to integrate
information from a variety of departments and regions stretches the capability of CRM
applications further. Most times, companies have no choice but to switch to data warehouses.
Companies can gain an edge from predictive analytics by completing the cycle beginning with
collection of data followed by its analysis and concluding with decisions as quickly as possible.
Time delays can occur as data is transferred from the operational sources or external sources to
a central point such as a data warehouse. Again, analysis of data contributes to latency. Finally,
time is lost when decisions are taken based on the analysis conducted. Technology can help to

lower the costs and time delays in the first two stages while decision latency is hard to reduce
except when companies take action based on rules. Marketing campaigns thrive on making offers
as quickly as possible often responding to events, such as the release of the latest Harry Porter
novel, as they unfold.
The timeliness and quality of the data are critical for wider adoption of business intelligence and
analytical tools in business. According the recent survey of 6000 marketing and IT executives,
data (quality, access, timeliness and usage) were considered to be the most important barrier to
the implementation of customer relationship management software.
At one end, companies choose data warehouses which are batch systems and use ETL (extract-
transform and load tools) for transferring data from operational sources to the central warehouse.
The extraction of data involves conversion of the data structure in the source files to a flat file, the
transformation turns the flat files into the data structures of the target database. Finally, FTP (File
Transfer Protocol) helps to transfer data from the source to the target database. ETL accounts for
60-80% of the time spent on BI projects. While a data warehouse accumulates multi-dimensional
data which is clean, the time delays are far too long for companies to be able to respond to
events as they happen.
The time delays in the transfer of data from transactional data sources to the data warehouse
take place as a result of the custom coding that has to often take place when data is converted
from especially formats that are not widely used such as legacy systems, data stored in
applications and proprietary software. As the variety of data incorporated in data warehouses
increases, so does the need to take recourse to custom coding. In addition, data warehouses can
either load data or be used for analytical processing at any given point of time. The conflict
between analytical processing and real time data is aggravated as larger volumes of data have to
be transferred, on the one hand, and the demand for increasingly complex queries grows on the
other.
The time delays can be shortened by tools that automate the process of conversion of data.
These tools employ proprietary scripting languages running within an ETL or DBMS server which
use language interpreters, stored in a meta-data repository, to process the incoming data. These
engines, however, cannot process unique data structures and need code written by humans to
process the data.
Another recent trend is the use of Enterprise Application Integration software which uses web
services to interlink heterogeneous platforms and help to transfer transaction data from them in

near real time. These systems are not as efficient as ETL tools in transformation of the data so
some companies have taken the initiative to integrate these two types of tools.
Companies which are interested in shortening the time delays have to consider other means. One
approach is to deploy data warehouse appliances, such as Netezza, which bundle the storage,
database server and campaign management server in a single system, which substantially
increases query performance without raising the investment costs.
http://www.intelligententerprise.com/showArticle.jhtml?articleID=59301112&pgno=1
www.intelligententerprise.com/showArticle.jhtml?articleID=59301169&pgno=1
http://download.101com.com/tdwi/research_report/2003ETLReport.pdf
http://www.thearling.com/text/dmtechniques/dmtechniques.htm

Predictive analytics

More Related Content

What's hot (19)

Similar to Predictive analytics (20)

More from Kishore Jethanandani, MBA, MA, MPhil, (20)

Recently uploaded (20)

Predictive analytics