tum analytics tools for retention ytl03101usen

7/31/2019 Tum Analytics Tools for Retention YTL03101USEN

1/6


2/6

2011, Frreser Research, Ic. Reprduci PrhibiedDecember 14, 2011

2Mus-Hae Aalics Fr yur Reei tl Ki

Fr Cusmer Ielligece Pressials

new customers exceeded rst-year revenues by almost 50%, orcing it to rethink its acquisition-

only strategy. On the other hand, augmenting acquisition with well-craed retention eorts

promises balanced growth by boosting the bottom line.

tackle retention at tHe customer level

Far too many rms ocused on winning new customers suer rom a leaky bucket. Teir

acquisition eorts eed the customer pipeline, but, at the same time, they lose customers at a

steady rate or a variety o reasons. Customer Intelligence (CI) proessionals can play a vital role in

plugging this leaky bucket by eeding insights about customers into the retention process. Unlike

in acquisition-ocused analytics, where CI teams try to acquire look-alikes o best customers,

retention-ocused analytics taps into an existing gold mine o customer data. o design retention-

ocused analytical solutions, CI pros must ocus on our main areas:

Identiy who to retain. o determine where to ocus retention eorts, rms need to understandwhich customers are worth retaining. As a starting point, tease out the 20% o customers with

the greatest potential to generate 150% o the prots.4 Existing value-based segmentation

models oer clues to identiy prot-based retention targets, but other actors, such as rate o

repeat purchases and service history, should also be considered.

Calculate the cost to retain. Once a retention target is identied, a rm must determinewhether to allocate resources in attempts to retain that customer or to let the customer ollow

their existing path and, possibly, churn. Customer lietime value is a key metric to understand

the cost o retention and the tradeo involved in resource allocation.

Estimate when to retain. An understanding o the customer lie cycle and the drivers o churnhelp determine when to introduce retention stimuli. A senior marketer at a leading retailer o

household goods and electronics told us, Its all about retention, not win-back, not reactivation;

its about intervening beore someone alls way out o pattern.

Design the appropriate retention oer. Based on an understanding o who to retain, at whatcost, and when to intervene, designing the appropriate next best oer through predictive

analytics helps ensure that the retention loop closes. Next-best-action strategies boost customer

lietime value by appropriately presenting the most acceptable oer to a customer.5

use two analytical approacHes to boost retention

As it relates to customer retention, analytics can serve two unctions: 1) to identiy and manage

customer attrition, and 2) to identiy opportunities to increase customer revenue and protability.

We recommend distinct analytical approaches toward each objective churn management and

cross-sell/upsell analysis (see Figure 1).


3/6




ch m e r p

Subscription-based models such as telecom, V, Internet, or any other bundled services widely use

churn and attrition analysis to understand customer deection. While churn is easier to spot when

the customer is in a time-bound contract or billing cycle, in cases where the tenure is not dened,

other proxies o churn are used. For instance, nancial services rms may look at churn as: 1) a

reduction in activity; 2) a lowering o average balances; or 3) a reduction in number o accounts.6 CI

proessionals should use churn analysis to:

Signal when to worry about customer attrition. Most rms deal with customer attrition inone orm or another. For instance, at least one in seven telecom and cable consumers is likely

to churn over a six-month period.7 Churn analysis uncovers predictors o customer attrition

or deection, whether voluntary or involuntary, and assigns likelihood to churn scores to

individual customers or segments.

Pinpoint timing o customer deection. Survival and hazardmodeling are time-to-eventanalytical techniques commonly used in churn analysis. Both are eectively used to estimate the

customer tenure. Paul Laughlin, head o customer and market insight at Lloyds Banking Group,

told us in Te Forrester Community For Customer Intelligence Proessionals: An analytical

technique I have seen bear ruit in this area [customer retention] is the use o survival analysis

[. . .] It [survival analysis] really comes into its own in helping with the timeliness question.

Identiying the likely tenure o a cohort o customers can help ocus eorts on those most at risk

o leaving.8

Refne retention targets urther. o extend the value o churn analysis, apply the learningsto eectively design an outbound retention campaign. Upli modeling renes churn analysisand ocuses oers and communications on those customers most likely to respond positively

to a retention campaign, avoiding those customers likely to respond negatively or not react at

all.9 A senior executive at Rogers Communications shared the ollowing scenario with us: As

heavy users o churn modeling churn rates decline over time, the challenge then is to build the

right economic model o outbound oers instead o blindly oering the richest oer to a small

number o customers at risk o churning.

c-s a u a ex f o p/s

Cross-sell and upsell tactics lure customers to buy more rom the portolio o products and services

oered by a rm. Firms use cross-selling to oer a product or service that goes with an existingproduct that a customer has and use upselling when a higher priced but similar item is oered to

the customer. CI proessionals should use cross-sell/upsell analysis to:

Target low-hanging ruit. On average, in the nancial services industry, consumers own ninenancial products but maintain just two and a hal o them with a single rm, opening up the


4/6




opportunity to cross-sell and upsell to boost share o wallet.10 Tis particularly makes sense or

multibrand, multiproduct scenarios.

Build product clusters and bundles. Association models and market-basket analysistechniques are typically used in cross-sell/upsell analysis to detect associations between discrete

events, products, or attributes and then assemble clusters o products that are purchased

together. Customers who have purchased some, but not all, products in a particular cluster are

good candidates to target or cross-selling or upselling. A senior analytics executive at AAA

National told us that, while its current cross-sell eorts are product-centric, it wants to move

toward a more member-ocused, next-best-oer approach to cross-sell.

Present relevant and contextual oers. Product associations as an output o the cross-sell/upsell analytical exercise should be placed within the context o the customers current lie cycle.

For example, just because diapers and baby ood appear in the same shopping cart does not

mean the customer is likely to buy all other types o baby products. ake into account the age o

the baby, or example, as a variable that should be considered to recommend an age-appropriate

baby product.

f 1 Reei-Fcused Aalical techiques

Source: Forrester Research, Inc.60482

Business applicationCommonly usedanalytical techniques

Cross-sell/upsell Association/affinity models(market-basket analysis)

Retention campaignplanning

Uplift modeling

Churn prediction Logistic regression

Description

These are used to identify purchase patterns and groupsof products purchased together by analyzing past co-

occurrences of events, purchases, or attributes. They donot involve a single output field to be predicted.

This models the dierence in behavior between targetand control groups, not just the behavior of a targetgroup in a retention campaign.

This estimates the probabilities of categorical outcomes,i.e., the probability of an event occurring or not occurring.It also gives the eect of the predictors on the categorical

outcome or target field.Survival analysis This gives the probability of a customer surviving up to a

certain time.

Hazard analysis This gives the probability that a customer might stop arelationship at a particular point in time.

Decision trees This is a set of rules with associated probabilitiesrepresented in a tree structure, which can be used forbinary or multiple outcomes.

Neural networks These are machine learning algorithms that use complex,nonlinear functions for estimation and classification andresemble biological neural networks in structure.


5/6




R E C o M M E n D A t I o n S

dont neglect tHe value of analytics in retention

Firms blikered b acquisii will ace a challegig ask ahead i iesig disprpriall

mre reai all he cusmers ha were aggressiel acquired. Bere geig he pi

makig a ugh rade a he risk cusmer dissaisaci CI pressials mus

ieree b shwig he alue ha aalics brigs reei ers:

a - h . Hw des chur aalsis appl a -subscripi-based mdel? Ee i chur maageme des direcl appl he

rms business model, experiment with the underlying concepts of determining customer

relaiship durai, chur predicrs, prbabili repurchase, ec., ad he ies i a

ull-fedged chur aalsis.

l h . there is a delicae balace beweereei ad acquisii. Ca he aalics applied drie reei pride a

information to make acquisition processes more eective? For example, uncovering churn

predicrs durig chur aalsis ca help i markeig plaig ad iesig i lead

surces ha are leas likel arac cusmers wh will chur i he uure. Cusmers

acquired hrugh he web chael ma hae a lwer prpesi chur ha hse

acquired hrugh a elemarkeig er.

a x h. Esure ha he isighs rmreei-based aalics are aailable he rlie emplees hadlig cusmer

interactions. For example, knowledge about a customers churn risk and churn indicators

ca help surprise ad deligh he cusmer durig a ace--ace ieraci a a bak adpeiall ifuece his r her decisi chur.

endnotes

1 Source: SAS helps 1-800-FLOWERS.COM grow deep roots with customers, SAS (http://www.sas.com/

success/1800owers.html).

2 Forrester asked respondents to our North American echnographics Afuent Online Survey, Q2 2010,

whether, given the state o the US economy, they were planning to change their shopping habits. wenty-

seven percent o all US online adults said that ree shipping oers will motivate them to buy more online

than they have in the past, and 26% o all US online adults said they will do more research online (beore

purchasing) to be sure they get the best price. Source: North American echnographics Afuent Online

Survey, Q2 2010 (US).

3 Source: Mark Colombo, senior vice president, digital access marketing, FedEx Services, speaking at

Forresters Marketing Forum 2008.


6/6



Forrester Research, Inc. (Nasdaq: FORR) is an independent research company that provides pragmatic and orward-thinking advice to global leaders in businessand technology. Forrester works with proessionals in 19 key roles at major companies providing proprietary research, customer insight, consulting, events, andpeer-to-peer executive programs. For more than 28 years, Forrester has been making IT, marketing, and te chnology industry le aders successul every day. Formore inormation, visit www.orrester.com.

2011 Forrester Research, Inc. All rights reserved. Forrester, Forrester Wave, RoleView, Technographics, TechRankings, and Total Economic Impact are trademarkso Forrester Research, Inc. All other trademarks are the property o their respective owners. Reproduction or sharing o this content in any orm without prior

written permission is strictly prohibited. To purchase reprints o this document, please email [email protected]. For additional reproduction and usageinormation, see Forresters Citation Policy located at www.orrester.com. Inormation is based on best available resources. Opinions refect judgment at the timeand are subject to change. 60482

4 Te top 20% o a companys customers generate 150% o the companys prot, while the bottom 20% o

customers drain 80% o prots. Source: Larry Selden and Georey Colvin,Angel Customers & Demon

Customers: Discover Which is Which and Turbo-Charge Your Stock, Portolio, 2003.

5 Te next best oer is whatever the customer nds most acceptable; it also maximizes the yield rom that

customer in sales, revenues, and prots in other words, customer lietime value (CLV). See the April 13,

2011, Boost Customer Lietime Value Trough Next Best Actions In Multichannel CRM report.

6 Customer attrition with its inherent message o ailure is a hard topic or marketers to tackle.

Marketers oen lack agreement on the timing o churn and the process o identiying customers with a high

likelihood o churn. See the October 20, 2008, Optimizing Customer Retention Programs report.

7 Over the next six months, telcos and cable companies can expect nearly 50 million o their subscribers

to switch to a new provider or cancel their service altogether. See the January 16, 2009, Te Science O

Churn: When And Why Consumers Switch Service Providers report.

8 Check out our discussion on analytical techniques used in retention. Source: Te Forrester Community For

Customer Intelligence Proessionals (http://community.orrester.com/message/15313#15313).

9 Upli modeling is an analytical technique that models the dierence in behavior between target and control

groups, not just the behavior o a target group. See the October 20, 2008, Optimizing Customer Retention

Programs report.

10 Forrester surveyed 4,630 US online adults to understand the breadth o their nancial relationships. Overall,

we ound that US online adults own an average o 8.9 nancial products. See the September 25, 2009,

Solving Te Cross-Sell Imperative In Financial Services report.
http://www.forrester.com/go?docid=58515&src=60482pdfhttp://www.forrester.com/go?docid=44400&src=60482pdfhttp://www.forrester.com/go?docid=46698&src=60482pdfhttp://www.forrester.com/go?docid=46698&src=60482pdfhttp://www.forrester.com/go?docid=44400&src=60482pdfhttp://www.forrester.com/go?docid=44400&src=60482pdfhttp://www.forrester.com/go?docid=55312&src=60482pdfhttp://www.forrester.com/go?docid=55312&src=60482pdfhttp://www.forrester.com/go?docid=44400&src=60482pdfhttp://www.forrester.com/go?docid=44400&src=60482pdfhttp://www.forrester.com/go?docid=46698&src=60482pdfhttp://www.forrester.com/go?docid=46698&src=60482pdfhttp://www.forrester.com/go?docid=44400&src=60482pdfhttp://www.forrester.com/go?docid=58515&src=60482pdf

tum analytics tools for retention ytl03101usen

Documents