SAS_Marketing Guide to Analytics

8/3/2019 SAS_Marketing Guide to Analytics

1/14


2/14

A MArketers Guide to AnAlytics

Table of Contents

Executive Summary ............................................................................ 1The World of Marketing Is Changing. Are You Being Left Behind? ............. 1

The Data Explosion .......................................................................... 2

The Power Shift to Consumers .......................................................... 2

Implications for Marketers ................................................................ 2

Building an Analytical Framework for Marketing.................................... 3

Analytically Driven Segmentation ....................................................... 3

Predictive Modeling ......................................................................... 6

Applying Predictive Models to Your Marketing Strategy ...................................7

Treatment Strategy ..........................................................................................8

Stepping Up to Advanced Predictive Models ..................................................9

Marketing Optimization Technologies ................................................. 9

Getting Started ................................................................................. 10

About SAS........................................................................................ 10


3/14


Executive Summary

Everybodys talking about customer analytics and how they can help companiesmarket more eectively. But or many marketing proessionals today, theres a gap

between theory and execution and its getting wider every day.

This paper is designed to give managers and other marketing proessionals an

introduction to applying analytics to marketing so you can signicantly improve

outcomes. It explains not only whyyou need to make this shit to analytically driven

marketing strategies and plans, but also how you get started and what kinds

o tools you need to develop and execute plans. Youll learn about building an

analytical ramework or marketing that will help you:

Increaseresponserates,customerloyalty,andultimatelyROI,bycontacting

the right customers with highly relevant oers and messages.

Reducecampaigncostsbytargetingcustomersmostlikelytorespond.

Decreaseattritionbyaccuratelypredictingcustomersmostlikelytoleaveand

developing the right proactive campaigns to retain them.

Delivertherightmessagebysegmentingcustomersmoreeffectivelyand

better understanding target populations.

The World of Marketing Is Changing.

Are You Being Left Behind?

Two dominant trends are transorming the business o marketing today:

Theincredibleexplosionofdatabeingcreatedbygovernments,businesses

and consumers.

Theshiftinpowerfromcompaniestoconsumersthatsbeendrivenby

advancements in technology.

Lets take a closer look at how and why and what these trends mean to your

organization.

1


4/14


The Data Explosion

The world contains an unimaginably massive amount o digital data today, and its

increasing tenold every 10 years. Not only are companies jumping on the databandwagon collecting, storing and linking massive amounts o data but customers

are generating massive amounts o new digital inormation through postings on

Facebook and Twitter, data created as they click through Web pages, data collected

by cookies, product reviews written by users and more. This kind o data is growing

exponentially in both size and strategic value to marketers who need to engage in a

1-to-1 manner with these consumers, as this data can be turned into a gold mine o

unique customer insight.

The Power Shift to Consumers

At the same time, theres been a continuous power shit to consumers. For example,

technology empowers them to easily nd the lowest-cost vendors or goods, and by

usingDVRsande-mailspamlters,theycanavoidmarketingcommunicationsfrom

businesses that they dont want to receive. Todays customers can also infuence

tens o millions o people to buy rom you or not by writing online reviews,

tweeting and blogging.

Another maniestation o this power shit to consumers is the act that they expect

product and service inormation thats personally relevant, timely and delivered via

their preerred channels. Failing to do this will ultimately rustrate customers and turn

them away, as your business will be perceived as out o touch or instance, by

blindly pushing products on customers rather than giving them timely access to helpulinormation at a time when they are open to oers and making purchasing decisions.

Implications for Marketers

To be eective in this new environment and maximize return on marketing investments,

organizations need a marketing process and strategy that is customer-centered and

powered by deep customer insight. Achieving marketing objectives and strategies will

require a much more granular analysis o customers and prospects than ever beore.

Consider the ollowing marketing objectives and todays real-world implications on

achieving them:

Revenue growth: Untargeted, mass marketing just wont cut it anymore.

To attain marketing-driven growth, companies have to gure out how to have

relevant conversations with their customers and prospects when and how the

customer preers.

Customer retention. Competition is erce, so you need ways to predict and

get in ront o attrition risks identiying customers most at risk o attrition and

the actors that infuence their decision, so you can generate proactive customer

retention campaigns.

1 Nielsen Consumer Research. Nielsen Global Online Consumer Survey, April 2009.http://blog.nielsen.com/nielsenwire/wp-content/uploads/2009/07/pr_global-study_07709.pdf

2

According to a Nielsen Global Online

Consumer Survey of more than

25,000 Internet consumers, people

now trust recommendations and

opinions from real friends and virtual

strangers more so than traditional

information sources, such as

corporate websites and ads.1
http://blog.nielsen.com/nielsenwire/wp-content/uploads/2009/07/pr_global-study_07709.pdfhttp://blog.nielsen.com/nielsenwire/wp-content/uploads/2009/07/pr_global-study_07709.pdf


5/14


Customer proftability. The Pareto principle suggests that you receive 80

percent o your prot rom 20 percent o your customers. Whether this is

precisely true or your company doesnt really matter. The point is that there are

protable customers, and there are customers who reduce prots. It is critical to

understand which customers all into each category.

Successully achieving these objectives is a key actor in competitive advantage

and long-term protable growth in todays data-driven, consumer-driven business

environment.

Building an Analytical Framework for Marketing

To achieve the core marketing objectives listed above, you need to build out an

analytical ramework that will enable you to deliver superior results rom your marketing

strategy. Analytics enhance the decisions you make as you execute on your strategies

and plans so you can be more eective and achieve better results. Key enablers o this

analytical ramework include:

Analytically driven, granular segmentation that enables you to identiy how

dierent customer segments are most likely to respond to specic campaigns or

marketing actions.

Predictive modeling capabilities that enable you to identiy the specic target

population likely to respond positively to a specic campaign or other marketing

activity. You can also use it to understand and predict the behavior o targeted

groups.

Optimization capabilities that help you to maximize economic outcomes by

making the most o each individual customer communication while considering

your companys resource and budget constraints, contact policies, the likelihood

that customers will respond and more.

The oundation or this analytical ramework is access to comprehensive, clean

customer data that can be analyzed to create unique customer insight and eective

segmentation. This data source should be continually updated as you interact with

customers and prospects (or example, purchase transaction data, online data rom

your website users and direct marketing response data.)

Analytically-Driven Segmentation

Customer segmentation is the process o dividing a customer base into groups o

individuals who are similar in specic ways relevant to marketing. It enables companies

to target groups eectively and allocate marketing resources appropriately. Two types

o segmentation to consider are oundation segmentation and targeting segmentation.

3


6/14


Foundation segmentation creates core segments that enable the company to

deliver a consistent treatment and ocus on long-term strategy. All customers must

be included, and each customer can all into only one segment. Segments can also

be subdivided into natural clusters, such as geography or level o protability. Key

attributes o oundation segments include value, prot, attrition, risk, demographics,

etc. With the introduction o a new product, where no targeting segmentation exists,

oundation segmentation would be the primary method o segmenting or the initial

marketing campaigns.

Targeting segmentation identies customers with specic needs and preerences.

Not all customers may be included in targeting segments, and each customer may

all into many dierent segments. This segmentation is useul or specic marketing

programs and campaigns; or example, identiying a targeting segment that is most

likely to respond positively to a specic campaign or customers most likely to leave or

a competitor. Attributes include behavior, time periods, account status, usage, etc. It

is ocused on short-term marketing activities, not on long-term relationship building.

Analytics enables you to go beyond oundation segmentation to targeting segmentation,

allowing you to execute more eective, sophisticated campaigns with messages and

oers that are highly relevant to recipients.

To better understand the importance o analytically driven, targeting segmentation, lets

look at how a telecommunications company can use it to improve outcomes. Marketing

would rst dene the segments with high churn (attrition) and high customer value. For

this case, we will dene the primary target group as the top 20 percent o the most

protable customers with a high churn rate. Assume that this group has the additional

characteristic o being high-usage customers. Next, marketing would subdivide the

group into three distinctive clusters based on two dierent revenue dimensions: usagerevenue and access revenue (see Figure 1). Usage revenue is the revenue gained rom

per minute charges, and access revenue is revenue gained rom rate plans.

UsageR

evenue

Share

Access Revenue Share

10% 20% 30% 40% 50% 60% 70% 80% 90% 100%

70%

60%

50%

40%

30%

20%

10%

0%

-100%

High-Profit Segment Clusters - three distinct clusters across two revenue dimensions

3. High-Usage Revenue Share

1.5% of Overall Base Average RPPU: $195.32 # Subs/Acct: 1.83 % Churned% Acquired: 0.94

2. High Balances

2.9% of Overall Base Average RPPU: $195.07 # Subs/Acct: 3.05 % Churned% Acquired: 0.97

Dimension of the circle defines the share of subscribers per cluster.

1. High-Access Revenue Share

5.5% of Overall Base

Average RPPU: $192.11 # Subs/Acct: 3.74 % Churned% Acquired: 0.85

Figure 1: A targeting segmentation example that provides for more analysis and better

targeting of offers.

4

A Large Bank Reduces Attrition

by Using Segmentation to Increase

Campaign Response Rates

In the banking world, getting to know

customers has grown increasingly

dicult as ATMs and online services

replace traditional ace-to-acetransactions in branch oces. For

larger banks, such as Portugals

Banco Santander Totta, the volume

o customers and transactions

compounds the problem.

But with SAS Customer

Intelligence, Santander Totta is

able to understand and predict

the needs and wants o its 1.7

million customers. By using SAS

to segment our customers in

credit card campaigns, weve seen

conversion rates increase rom 4 or

5 percent to 20 or 30 percent, says

SergioVieira,SantanderTottas

DirectorofCustomerRelationship

Management. Were increasing the

success o our campaigns by ve or

six times.


7/14


By developing these clusters, marketing would now have three distinct customer

groups that they could urther analyze and then target with unique oers. In this

example o the three clusters, the high-usage revenue cluster is shown to have the

highest prot per user, the lowest number o subscribers per account and the middle

churn rate. This inormation can be used to develop an oer in which this cluster o

customers would be most interested. For example:

Thehigh-usagerevenueclustercouldreceiveanofferforanewphoneifthey

renew their contract.

Thehigh-accessrevenueclustercouldreceiveanofferfora10percentdiscount

on their rst months ee i they renew their contract.

Theclusterinthemiddlecouldreceiveanofferfor100freeminutesiftheyrenew

their contract.

Because these oers target the specic usage pattern or each cluster, they should be

more attractive and generate higher response and revenue rates than those based on

non-analytical segmentation strategies.

When done correctly, the benets o analytically driven targeting segmentation are

signicant. For example, you can realize:

Moreprotablemarketingcampaignsbyfocusingmarketingeffortsonthe

customers who will be most likely to buy your products or services, as well as

identiying your most and least protable customers. You can also use it to avoid

markets that wont be protable and those markets where you can charge a

higher price or your products and services.

Moreloyalcustomers,whichleadstohighercustomervalueandincreasedprots.For example, you can build loyal relationships with customers by developing and

oering the products and services they want through their preerred channels. You

can also use this insight to nd ways to improve customer service and enhance

products to better meet customer needs.

Ahigherlevelofcompetitivenessintodayshighlycompetitivemarketplacefor

example, by nding ways to get an edge in specic parts o the market and

innovating new products to meet the needs o important customer segments.

5


8/14


Predictive Modeling

Predictive modeling is essential to the success o marketing strategies and plans in

todays environment. The goal is to use one or more predictive modeling techniques toidentiy the target population likely to respond positively to a specic campaign or other

marketing activity, as well as to understand the behavior o targeted groups.

Consider what happens when a telecommunications company does a normal random

marketing mailing versus a mailing based on predictive modeling, which enables

more strategic targeting. In this example, the company uses predictive modeling to

generate the graph in Figure 2, which analyzes the rst decile o customers (the top

10 percent by revenue) and shows that 30 percent o these customers have a high

likelihood o attrition a key group o customers or any company to ocus its retention

eorts on. The graph also helps the marketing department ocus its retention activities

on key target segments (and save the money to spend elsewhere). Failing to target

unds in this way leads to diminishing marginal returns but getting it right means the

telecommunications company benets rom:

Increasedresponseratebycontactingtherightcustomers.

Reducedcampaigncostbyselectingthecustomersmostlikelytorespond.

Strongercustomerrelationshipsbyunderstandingthetargetpopulationand

conveying messages that are highly relevant to them.

Percentlikely

to

attrite

0 1 2 3 4 5 6 7 8 9 10

100

90

8070

60

50

40

30

20

10

0

Customer base (deciles)

30%

48%

62%72%

Targeting the top 10% of customer basecaptures 30% of those likely to attrite

Figure 2: Predictive modeling can show how many customers have a high likelihood ofattrition to enable more strategic targeting.

6

Cabelas Drives Its Growth Strategy

Using Predictive Models

Using SAS, we create predictive

models to optimize customer

selection or all customer contacts.

Cabelas will use these prediction

scores to maximize marketing

spend across channels and within

each customers personal contact

strategy. These eorts have allowed

Cabelas to continue its growth in

a protable manner, says Corey

Bergstrom,DirectorofMarketing

Research and Analytics or

Cabelas. Were not talking single-digitgrowth.Overseveralyears,its

double-digit growth.


9/14


7

Key Predictive Modeling Techniques

You can take advantage o a wide variety o predictive modeling techniques to

meet your needs, including: Decision tree modeling: With a decision tree, a population is split into

subgroups that tend to be more homogeneous than the original sample. Each

o the subgroups continues to be split into even smaller subgroups until the

modelcannotbeimproved.Decisiontreesallowfornonlinearrelationships,

but they also clump probabilities and allow or less distribution.

Clustering: With clustering, groups o individuals are identied based on their

proximity to each other. The cluster procedure and discriminate analysis utilizes

an eective method or nding initial clusters with a standard iterative algorithm

that minimizes the sum o squared distances rom the cluster means.

Logistic regression modeling: A logistic regression is a generalized linear

model or predicting probabilities. It calculates the probability o a particular

record being a member o a target group based on the values o the predictor

elds.

Neural networks: With neural networks, data can be processed in parallel to

quickly nd complex relationships. Nodes in neural networks sum inormation

rom other nodes connected to it and pass inormation to the other nodes.

This allows or more complex, nonlinear relationships, but they can make

interpretation dicult.

Survival modeling: A survival model is a method o statistical analysis used or

determining time-to-event or one-time events. The model includes both the

actual probability o events and the eects o covariates. It can be used to studysurvival trends by demographic area, channel, credit class, rate plan and type

o churn, as well as to estimate remaining lietimes or present customers.

Applying Predictive Models to Your Marketing Strategy

Predictive modeling helps you execute your marketing strategy and ultimately

achieve your broader marketing objectives. As you think about the objectives you

are trying to achieve, consider the ollowing questions, which will guide you toward

selecting the predictive modeling techniques used to drive those marketing programs

and ultimately to the treatment strategy used to execute it:

Whywillmycustomerattrite?

Whenwillmycustomerattrite?

Whoissaveable?

Whowillbuy?Whatwilltheybuy?

Whichproductwillthecustomerbuynext?

Whenwillthecustomerbuy?


10/14

8


Forexample,whenmarketingasksthequestion,Whenwillcustomersattrite?you

could answer this question using a survival model that identies customers who will

likely leave to go to a competitor within three months, six months or even one year.

Withamplenoticelikethis,youcantakeproactivestepstopreventattrition.Other

types o predictive models can help you identiy which customers can be convinced to

stay, as well as:

Determinehowtomaximizerevenueforexample,byunderstandingwhich

customers will buy which products and designing campaigns accordingly.

Calculateacustomerspropensitytobuyspecicproductsinsightneededto

develop highly targeted campaigns and oers.

Identifythesequentialorderofpurchasesbyperformingamarketbasketanalysis.

Identifywhenapurchasewilllikelybemadebycustomersegmentsorindividual

customers by using a survival model.

Treatment Strategy

Ater determining the appropriate modeling approach to meet your needs, the next

step is to develop a treatment strategy. This involves using analytical models to

determine customer value and dene customer segments. When done correctly,

you can create an individual view o the customer that when combined with

segmentation and customer value analytics enables you to develop a specic

treatment strategy that will optimize outcomes.

To understand how this works in practice, lets consider an example. Taking another

look at Figure 2, we see that within the top 30 percent o revenue-producingcustomers, 62 percent have a high likelihood o attrition. To reduce this expected

attrition rate, marketers can use predictive modeling to create an oer specic to each

group o customers an oer with a high likelihood o compelling them to stay. Since

the needs o the highest-revenue-producing customers are likely to be dierent than

other groups, marketing must create separate oers or each customer segment. To

select the best treatment strategy or each customer group and ultimately achieve the

best results, the marketing department should use several dierent components to

describe and understand their customers, including:

Customerprole.

Customervalue.

Attritionpropensity.

Othercustomermodels.

Marketbasketanalysis.

When all o these components are combined, your marketing department has a

multichannel, integrated view o the customer that provides the inormation and insight

needed to make the best treatment decisions.


11/14


9

Stepping Up to Advanced Predictive Models

Onceyourcompanybecomesprocientwithabasicpredictivemodelingstrategy,

you can use more advanced models to realize even more perormance improvements.

Advanced models answer the same questions mentioned above, but with more

precision or sophistication. For instance, you can determine the specic time horizon

or the predicted attrition. And using an advanced survival model enables you to identiy

customers who will attrite or buy within certain windows o time (or instance, within three

months). Using the analytic insight enabled by advanced predictive modeling, you gain

an additional level o inormation to urther improve returns on your marketing eorts.

Marketing Optimization Technologies

Segmentation, predictive modeling and testing are great or improving the

eectiveness o individual customer interactions, but when used alone, they cant helpyou deal with the ull scope o realities that marketers ace every day. Issues such as

competing business goals across divisions, managing multiple marketing programs

against constraints such as channel capacity, controlling budgets, and managing

customer contact policies must be addressed. Internal political and tur battles can

make the decision process dicult. For example, deciding which campaigns are sent

to which customers can be a very volatile issue within a multi-product organization.

TomaximizeprotorROI,enforcecontactpolicies,andstaywithinbudgetwhen

your company has multiple products oered in hundreds o campaigns to millions o

customers, you need ar more than just experience and human intuition. You need a

technology-based solution optimization.

Optimizationtechnologiesapplymathematicaltechniquesthatenableyoutomaximize

economic outcomes by making the most o each individual customer communication

while considering business variables such as your companys resource and budget

constraints, contact policies, the likelihood that customers will respond and more. For

example, using what-i analysis delivered through marketing optimization sotware,

youcanincreasetheROIofcampaignsbyanalyticallydeterminingthebestoffersfor

individual customers and including analytical insight in the implications o business

constraints. You can also target customers to maximize protability, response rates,

asset levels or any other parameter you choose all while taking into account

customer preerences, propensities, protability, costs, contact policies and other

business goals and objectives relevant to campaigns and communications.

The Importance of Measurement

Measurement is a critical part o the

marketing environment its all about

accountability. So when you build a

marketing campaign, you also need

to identiy specic measurements

so you can determine which model

outperorms another, or understand

how outcomes may have been

dierent i a campaign had been

based on a dierent segmentation.

The good news is that analytics

lends itsel to measurement so

you can use it to see how you aredoing, identiy problems early and

continuouslyimprove.Differenttypes

o reporting can give you insight

needed to create better campaigns

and develop better modeling when

developing uture campaigns.


12/14

10


Scotiabank Gains Wallet Share Through Effective Marketingto Its Existing Client Base

Scotiabank mines through data rom about 7 million customers in any givenmonth to generate leads and create targeted marketing campaigns. In the past,

the bank ran one-o customer campaigns or a total o ve or six campaigns

ayear.Now,withSASMarketingOptimization,thebankhasincreased

exponentially the number o campaigns per year, which is helping generate

more sales. Were now running between 30 and 60 campaigns per year; these

go out to hal a million customers every month and help us generate about 6

millionleadsoverthecourseofayear,saidVicMoschitto,DirectorandHead

ofDecisionSupportandManagementforCanadianBankingatScotiabank.

Finding those customers would have been a major hurdle under our old

methodology, where every campaign was its own little island.

Using SAS Marketing Optimization, Scotiabank is able to:

Minethroughthebanks7millioncustomersandupto18monthsofdata

to derive insights or more targeted customer campaigns.

Maximizecampaignoutcomesbyhelpingreneindividualcustomer

communications.

IncreasemarketingROIbydeterminingthebestoffersforindividual

customers, while delivering analytic insight into the value o business

constraints (such as channel capacity and contact policies).

ByusingSAS,weregettinganROIinexcessof100percentasignicant

return or the cost o the solution, says Moschitto. The way were able tointegrate the sotware within our overall campaign and customer contact strategy

is giving us a signicant edge over our competitors.

Optimizationsolutionscanalsoenhanceyourcontactstrategysothatyoudont

oversaturate customers or violate corporate governance requirements. For example,

you can eliminate uncoordinated and conficting communications. Also, relevant

relationship actors such as customer risk, advertising exposure and householding are

incorporated into the optimization to ensure that valuable customers are receiving the

best possible set o communications across every channel.

And nally, with optimization solutions, you can increase organizational eciency.

For example, you can use what-i analysis to quantiy where changes in stang and

budget will really pay o, where youre leaving money on the table or where you have

unused capacity.


13/14


11

Getting Started

Soareyoureadytorealizethebenetsofananalyticalframeworkformarketing?

Given the rapid pace o change occurring in the world o marketing today, you likely

cant aord not to make strides toward employing analytically driven marketing

strategies, plans and tactics.

SAS can help you get started quickly by providing an integrated ramework or

enterprise marketing management. SAS Customer Intelligence provides the most

comprehensive suite o analytic marketing solutions, enabling marketing executives

tomakesmarterdecisionsandsolvemorebusinesschallenges.OnlySASprovides

the complete set o capabilities needed or a customer-ocused marketing process.

And all SAS applications whether they support segmentation, predictive modeling or

optimization run on a single, integrated ramework that lets you get more insight rom

your customer data and drive better marketing perormance.

To learn more about how SAS can help you achieve your marketing goals, visit

www.sas.com/customerintelligence.

About SAS

SAS is the leader in business analytics sotware and services, and the largest

independent vendor in the business intelligence market. Through innovative solutions

delivered within an integrated ramework, SAS helps customers at more than 45,000

sites improve perormance and deliver value by making better decisions aster. Since1976SAShasbeengivingcustomersaroundtheworldTHEPOWERTOKNOW .

Why SAS

Smarter marketing decisions.

SAS data integration and

analytics provide the knowledge

and insight needed to

understand your customers and

make smarter decisions.

More marketing challenges

solved. SAS Customer

Intelligence provides the

most comprehensive suite o

enterprise marketing solutions.

Leading business solutions

combined with integrated

data management, analytics

and reporting provide the ull

breadth o capabilities needed

to solve the most challenging

marketing problems.

A path to grow and evolve.

SAS can help you address

your needs at all stages o

your marketing organizations

development.

Proven success. SAS is astrong and stable company with

a loyal customer base around

the globe.

Theres garden-variety analytics, and then theres the stu that matters. And what

we have chosen to do is to put our ocus on the predictive analytics, because we

think thats where the value is. And the de acto standard, best guys on the planet

have been, are today, and always will be SAS. And thats why our alliance with

them is so distinctive and important.

Bill Green, CEO, Accenture
http://www.sas.com/solutions/crm/http://www.sas.com/customerintelligencehttp://www.sas.com/businessanalytics/index.htmlhttp://www.sas.com/businessanalytics/index.htmlhttp://www.sas.com/customerintelligencehttp://www.sas.com/solutions/crm/


14/14

SAS Institute Inc. World Headquarters +1 919 677 8000

To contact your local SAS ofce, please visit:www.a.m/f

SAS and all other SAS Institute Inc. product or service names are registered trademarks or trademarks of SAS Institute Inc. in the USA

and other countries. indicates USA registration. Other brand and product names are trademarks of their respective companies.

Copyright 2010, SAS Institute Inc. All rights reserved. 104697_S64720.1110

SAS_Marketing Guide to Analytics

Documents