Marketing Strategy - Customer Analytics Chapter03-2.4.pptx

© Robert Palmatier 1
Marketing Strategy
Chapter 3
Marketing Principle #2
All Customers Change  Managing
Customer Dynamics

Agenda
 Introduction
 Approaches for Managing Customer Dynamics
 Evolution of Approaches for Managing Customer Dynamics
 Lifecycle Approach
 Customer Dynamic Segmentation Approach
 Customer Lifetime Value Approach
 Choice Models
 Framework for Managing Customer Dynamics
 Inputs to Managing Customer Dynamics Framework
 Outputs of Managing Customer Dynamics Framework
 Process for Managing Customer Dynamics
 Managing Customer Dynamics Example
 Takeaways
 Case
2
© Palmatier

All Customers Change
 Another underlying issue facing managers that make marketing decisions difficult
is that all customers change
 Customer’s desires/needs for most products and services change over time or due
to specific events
 Individual consumer needs change (age, experience, and due to trigger events)
 Customers are embedded in industries/markets, which change overtime (PCs
20 years ago and now)
 Customer's needs vary not only due to inherent differences in people
(heterogeneity) but also as people and markets change (dynamics)
 Thus, we need to adapt our “static” segmentation of all customers based on
“generic” needs (MP#1) by focusing on our existing customers and accounting for
their time dependent needs (MP#2)
 Customer dynamics are changes in customer preferences that occur over time
© Palmatier 3

Exercise on Customer Dynamics
 Everyone take a few minutes to describe your or your parents
past car purchases
 Car at 20 _______________
 Top two reasons bought __________ & ____________
 Car at 30 _______________
 Trigger for change _______________
 Car now _______________
 Trigger for change _______________
© Palmatier 4

5 Sources of Customer Dynamics
 Discrete life events
 Typical lifecycle
 Learning effects
 Product lifecycle
 Constantly changing
environmental context
© Palmatier 5

Sources of Customer Dynamics
© Palmatier
Description Rate of
Change
Examples
Individual level
Discrete life events Immediate A first-time parent often changes their preference for cars,
vacations, and restaurants.
Typical lifecycle or
maturation as people age
Slow As people age, they become more focused on risk reduction, less
willing to change, and more focused on comfort and health.
Product learning effects Medium Customers might learn, after using a product for a time, that
there are certain specialized or high-tech features they would
like.
Product market level
Product lifecycle Medium During early stages, consumers may purchase more new
features, in later periods, they may get more price sensitive.
Environmental level
Changes in economy,
government, industry, or
culture
Slow to
immediate
As the culture around “health food” changes, consumer
preferences in response to dietary concerns (e.g., calories,
sodium, carbohydrates, gluten, fat) also change.
6

Customer Dynamics: A Fundamental
Assumption of Marketing Strategy
 Thus, customer dynamics is a fundamental “problem” that all firms must
address when developing an effective marketing strategy
 Customers change; failure to understand and address these dynamics will
lead to poor business performance
© Palmatier 7
Marketing principle #2: all customers change and an effective
marketing strategy must manage customer dynamics

Example: General Motors
 Buick: failure to manage customer dynamics
 Customers needs changed, customers moved to different suppliers
 Left Buick with a smaller portfolio of older customers
 Brand became associated with the elderly; sales dropped 50%
 Acura: success in managing customer dynamics
 Honda realized its customers were migrating to more expensive cars
 Launched Acura: a higher priced luxury car targeted to those customers
 Within a few years, Acura was one of the best-selling luxury brands in the US
© Palmatier 8

Agenda
 Introduction
 Choice Models
 Takeaways
 Case
© Palmatier 9

Evolution of Approaches for Managing
Customer Dynamics
 Lifecycle Approach uses generic stages of growth and their
position in the lifecycle to determine customer preferences
and associated strategies
 Dynamic Customer Segmentation segments a firm’s
existing customers on the basis of their similar, expected
migration patterns
 Customer Lifetime Value captures the contribution of each
customer according to his or her expected migration path
over the entire lifetime with the firm
© Palmatier 10

Evolution of Approaches for Managing
Customer Dynamics
© Palmatier
Speed of Response
Slow Fast
Size of Segment Managed
All Customers Niche Segment
Lifecycle Approach
Pros Cons
Simplicity Assumes all customers
follow one curve
Ease of use
Averages all customers
Uses generic customer stages of growth and their
position in the lifecycle to determine customer
preferences and associated strategies.
• Customer lifecycle
• Product lifecycle
• Industry lifecycle
Ignores causes of
customer dynamics
Dynamic Customer Segmentation
Pros Cons
Combines lifecycle and
segmentation methods
Segments are not
perfectly homogeneous
Matches strategic
marketing thinking
Puts continuous change
into discrete stages
Segments a firm’s existing customers on the basis
of their similar, expected migration patterns.
• May use hidden Markov models
• May use lost customer analysis
Identifies temporally
homogeneous groups
Customer Lifetime Value
Pros Cons
Provides insights for
AER decisions
Requires insight into
future migration
Supports a customer-
centric culture
Requires detailed
financial data
Captures the contribution of each customer
according to his or her expected migration path
over the entire lifetime with the firm.
• Uses discounted cash flows
• Accounts for expansion and retention
expectations
Captures dynamics and
heterogeneity
11

Lifecycle Perspective: “Ok” First
Approximation
 Assumes “all” customers, markets, products, industries move thru a
sequential set of “stages” as they mature
 Applies a standard set of marketing tactics at each stage
 Pros:
 Easy to apply
 Makes common sense
 Cons:
 Misses many unique trigger points and migration paths (ignores
temporal heterogeneity)
 Inaccurate and/or ineffective for many firms
© Palmatier 12

Different Lifecycle Approaches
 Customer lifecycle attempts to capture how individuals typically change
as they age and reach common age-related milestones
 Product lifecycle proposes that various products go through four typical
stages in relation to their acceptance by society: introduction, growth,
maturity, and decline
 Industry lifecycle comprises of five stages:
1. Early establishment of its range and boundaries
2. An innovation stage to set a “dominant design”
3. The shakeout stage, marked by economies of scale, such as that
smaller players get forced out
4. Maturity, when firms focus on market share and cash flows
5. The decline stage, when sales decay for the industry as a whole
© Palmatier 13

Typical Customer Product Lifecycle
© Palmatier
Introduction Growth Maturity Decline
• Product just launched on the
market, often perceived as
risky by customers.
• Most relevant features are still
unknown by customers and
sellers.
• High prices.
• Seller focuses on finding new
customers and promoting
product trials.
• Product gains acceptance in
the market, and sales
expand quickly.
• Users are more comfortable
with the product and know
which features they want.
• More focus on retaining
customers and generating
repeat sales.
• Customers become
increasingly price sensitive.
• Product is widely accepted,
and growth begins to slow,
because fewer new users
are available.
• Market becomes
competitive, and some
firms drop out.
• Intense price competition
reduces profits.
• Some sellers focus on niche
segments to avoid
competitive pressures.
• Destructive competition
and changing consumer
needs and desires lead to
product decline.
• Firms’ sales and profits
decline.
• Firms with higher cost and
those without a unique
advantage exit the market.
• Market often consolidates
with fewer suppliers.
Sales
Descriptions
of the Stages
14

Customer Dynamic Segmentation Approach
 Evaluates existing customer’s behaviors/needs in each AER stage to understand
temporal differences
 Customers are “temporally” similar in each stage (assumption)
 Matches marketing action domains (i.e., acquisition is often a self-contained
marketing domain)
 Dynamic-based segmentation is sometimes called the Acquisition–Expansion–
Retention (AER) model, because it captures customers entering the firm’s
portfolio and expanding over time, even as other customers slowly leave
 Acquisition stage begins with first contact, typically before the first purchase occurs,
when prospects and early customers have similar needs
 Expansion stage has firms that are trying to upsell or cross-sell to expand their sales
and engagement with existing customers
 Retention stage deals with customers who migrate not because of a mismatch in the
core offering or a life event but because they have a basic propensity to switch, in pursuit
of “greener pastures”
© Palmatier 15

Customer Dynamic Segmentation Approach
(AER Model)
© Palmatier
AcquisitionStage Expansion Stage Retention Stage
A1
A2
E1 R1
R2
Lost
Customers
Prospective
Customers
E2
16

Example: AER Strategies Emerge from a
Dynamic Segmentation in B2B Market
 Uses Hidden Markov Models (HMM)
 Simultaneously clusters and links clusters
 Give probabilities of moving among segments
 # clusters are data driven versus AER
 Can evaluate what strategies trigger or drive migration
 Example of 346 B2B customers over 6 years
 Migrations even occur in a mostly stable B2B customer
portfolio
© Palmatier (See Data Analytic Technique 3.1 for more details of HMM) 17

Hidden Markov Model (HMM) Analysis
Hidden Markov models (HMM) can uncover stages or
“states” of customer behaviors, as well as how those states
evolve. Because each state describes the common
behaviors exhibited by some group of customers at some
point in their relationship with a firm, HMM is a form of
dynamic segmentation.
DAT 3.1
Description
• To understand the dynamics (or stages) stages of a customer’s relationship with a
business.
• To dynamically segmentation the customer base.
• To predict when a customer might change stages, which may imply more or less
value for the firm.
• To determine when to proactively seek to build customer relationships.
When to Use It
How It Works
• A customer’s relationship with a firm exists in one of several possible unobserved (or hidden) stage, each with finite probability. Customer behavior
varies depending on the stage, such that a “stronger” stage customer likely buys more than one in a “weak” stage. In HMM, customers also have a finite
probability of transitioning from any one stage to another, partially as a function of marketing efforts. For example, advertising might cause customers
to shift from a weaker to a stronger stage.
• Customer behaviors (e.g., purchases) and firm actions (e.g., marketing) serve as inputs to the HMM, which estimates five outputs: (1) the number of
feasible states (or dynamic stages) in the data; (2) initial probability that a customer is in each state; (3) transition probabilities, or the probability that
customers move from one state to another; (4) the conditional probability of a behavior, given the customers’ hidden state; and (5) the effect of
marketing in moving customers across stages.
Example
To dynamically segment alumni donation behavior and investigate which of its marketing activities prompt donors to give money, ABC University
conducted a HMM analysis that identified three hidden states: dormant (corresponds with no donation), occasional (corresponds with infrequent
donation), and active (corresponds with frequent donation). With low marketing effort, dormant customers remain dormant in the next period with a
90% probability (red text); active customers have a 33% chance of becoming occasional in the next period (green text). With high marketing effort
though, the dormant customers become occasional donors with a 57% probability (red text), and active customers exhibit only a 25% chance of
becoming occasional (green text), both of which are good for ABC.
Dormant Occasional Active
Dormant 0.90 0.10 0.00
Occasional 0.08 0.55 0.37
Active 0.02 0.33 0.65
Dormant Occasional Active
Dormant 0.40 0.57 0.03
Occasional 0.03 0.50 0.47
Active 0.00 0.25 0.75
Transitions: Low Marketing Effort Transitions: High Marketing Effort
18
© Palmatier

Hidden Markov Model Example:
Relationship States and Migration Paths
© Palmatier
Damaged State
(57%)
Transactional
State (50%)
Transitional State
(29%)
Communal State
(61%)
62%
13%
Exploration
Migration
Recovery
Migration
Neglect
Migration
Betrayal
Migration
16% 21%
Strong
Relationships
Weak
Relationships
Poor
Relationships
Endowment
Migration
35%
38%
19
7%

Insights for B2B Hidden Markov Model
Example
 States in HMM describe different types of behavior that the same (or different)
consumers might exhibit at different points of time; transitions among states
capture the notion that consumers can switch from one state to another at any
point in time
 In the transactional state, the relationships indicate low levels of customer
trust, commitment, dependence, and relational norms
 If customers follow a positive migration path and move to the transitional
state, they do so only briefly
 In the communal state, the levels of trust, commitment, dependence, and
relational norms are higher than in any other state, and the relationship
produces good cooperation and profit
 The damaged state produces low levels of trust and commitment and very low
relational norms and cooperation, though customer dependence tends to stay
high
© Palmatier (See Data Analytic Technique 3.1 for more details) 20

Customer Lifetime Value (CLV) is a Key
Analysis Tool for Making AER Decisions
 How do you know best customers to acquire/expand/retain?
 In many banking initiatives only 1 in 3 “customers” remain after incentive
ends
 Are all customers worth acquiring or retaining?
 CLV approach: evaluates a firm’s profit as the sum of each customer’s
lifetime discounted cash flows
 Approach captures “true” contribution of each customer at any stage by
accounting for:
 Customer heterogeneity and dynamic effects (individual level, uses
transition expectations, and discounts future profits)
 Tradeoffs among AER strategies (e.g., how acquisition may affect
retention)
© Palmatier 21

CLV Approach (Australia and New Zealand
CMO Survey)
 Survey conducted on 255 CMOs and marketing directors in Australia and
New Zealand
 Those who “always measure the lifetime value of each customer”
achieved a 16% average increase in their annual marketing budget as
compared to 0% for those who do not measure it.
 75% of the marketers are engaged in some level of CLV effort within the
organization
© Palmatier 22

CLV Accounts for Varying Profits Across
Customers
© Palmatier
 Beyond 80/20 rule: firms earn 150% of their profits from 30% of their
customers
 CLV captures these difference in your existing customers so you can
acquire, expand, and retain the “best customers”
The Right Customers: Acquisition, Retention, and Development (HBR Press)
Breakeven
Range of Customer Probability
Number of
Customers
Modestly
Unprofitable
Modestly
Profitable
Highly
Profitable
Heavy
Losses
23

CLV Accounts for the Time Varying Profits of
Your Customers
© Palmatier
 On average, annual earnings typically increase over a customer's life due to
cross/up-selling
 But, some customers are more costly to acquire or retain (lowering prices, high
service levels)
The Right Customers: Acquisition, Retention, and Development (HBR Press) 24

Customer Lifetime Value (CLV) Analysis
 Need cash flow (sales - costs) for individual customers or group of
customers
 Calculate CLV of each customer:
 Where
 CFi,t = net cash flow generated by the customer i at time t
 T = time horizon for estimating the CLV
 d = discount rate
 CLV is the “value” added, by an individual customer, to the company
 CLV approach is a form of customer-centric accounting where firm’s
value is the sum of all its customers’ CLV
© Palmatier
 


T
t t
t
i
i
d
CF
CLV 0
,
)
1
(
25

Simplified Customer Lifetime Value Analysis
 Several simplifications make CLV calculations even more straightforward
 Assuming that T  infinity and that the contribution margin and marketing costs
(weak) do not vary over time
 Assuming that the contribution margin and marketing costs do not vary over
time, the CLV in dollars for the ith customer reduces to just five inputs:
1. Mi = margin for ith customer in $ (sales $ and margin as %)
2. Ci = annual marketing cost for ith customer in $
3. ri = retention rate for ith customer as a %
4. d = discount rate as a %
5. Ai = acquisition cost for ith customer in $
© Palmatier 26

Customer Lifetime Value (CLV) Analysis
This method quantifies the future discounted profitability
of a customer. It breaks down firm- or product-level
profitability to the customer level, enabling a customer-
centric approach.
DAT 3.3
Description
• To identify which customers are worth acquiring and retaining.
• To determine where to target marketing programs to maximize the firm’s return on
marketing investments.
• To understand the “true” value of a customer to a firm, including both revenues and
costs.
When to Use It
How it Works
Customer lifetime value is the dollar value of a customer relationship, according to both its present value and the projected future cash flows from the
relationship. The calculation process consists of three steps: (1) estimating the remaining customer lifetime, or number of years over which a customer is
likely to maintain a relationship with the firm, normally according to retention rates; (2) forecasting net profits from the customer over the predicted
lifetime, and (3) calculating the net present value of the future amounts. Because CLV ranks customers on the basis of profitability, it can target marketing
campaigns toward the most high value customers. The CLV formula is given as follows.
Both current and potential customers can be segmented according to expected long-term profits or CLV. The graph below plots the CLV distribution of firm,
which consists of inactive customers (low to negative CLV), active customers (positive CLV), and highly active customers (very high CLV) shows that the
right portion of the graph below highlight a firm’s most active customers). Firms can use such a graph to identify and target the most profitable customers
for marketing retention campaigns.
Example
where:
𝐶𝐿𝑉𝑖= Customer Lifetime Value
Mi = margin for ith customer in $
Ci = annual marketing cost for ith customer in $
ri = retention rate for ith customer as a %
d = discount rate as a %
Ai = acquisition cost for ith customer in $
A manager of a cable company wants to determine if it is strategic to acquire the Brett family, by estimating their household-level CLV. The manager
estimates that it will cost the company $65 (A) to get the Bretts’ to switch, and the Bretts’ will generate $100 profit each year (M), with a $10 annual
marketing cost to retain them (C). The estimated retention rate (r) is 65%, and the current discount rate is 5%.(d) . From the formula, the CLV for the
Brett’s is $235, which suggests the Brett’s on net, are profitable to the cable company.
CLV
#
of
Customers
Active Customers
Inactive Customers Highly Active
Customers
𝐶𝐿𝑉𝑖 =
𝑀𝑖−𝐶𝑖
1−𝑟𝑖+𝑑
− 𝐴i
27
© Palmatier

Example: RBC (Canada)
 Royal Bank of Canada (RBC)
 Identified medical students as high CLV customers
 Implemented a program to satisfy their needs early during the
progression of their careers: products such as credit cards, help with
student loans, and loans to set up new practices.
 In the first year, RBC’s market share in this segment increased from 2% to
18% and average sales were four times higher than average customers..
These customers were also very loyal.
© Palmatier 28

Some Generalizations on AER Strategies
from CLV Analysis
 AER strategy that maximizes CLV maximizes neither the acquisition rate
nor retention rate
 Investments in customer acquisition and retention have diminishing
marginal returns
 Under spending in acquisition and retention is more detrimental and
results in smaller CLV than overspending
 A suboptimal allocation of retention expenditures will have a larger
detrimental impact on long-term customer profitability than suboptimal
acquisition expenditures
(Kumar 2005)
© Palmatier 29

Customer Referral Value (CRV)
 What is the lifetime value of a customer bringing you
another customer?
 Highest CLV customers do not always generate the highest
CRV (weakness of CLV)
 “Advocates” real value is higher than a CLV would predict so
you need to protect these customers (3x more in one study)
 Identify advocates: protect, enable, expand
 Referral programs
 How to build advocates (Apple, BMW, not Wal-Mart)
(Kumar, Peterson, and Leone 2007)
© Palmatier 30

RFM Analysis is “Poor Man’s” CLV
 Direct marketers have been using a simplified version of the CLV for
decades, targeting customers to receive expensive catalog mailings. They
use three readily available customer behaviors:
 Recency or time elapsed since last purchase
 Frequency of purchases in last period
 Monetary purchases in last period
 These RFM (recency, frequency, and monetary) variables put customers
in rank-ordered groups, based on their value in the past year (not by
modeling but by rank-order sorting)
 Using the profits generated from a test mailing to a few customers from
each group, direct marketers then mail the catalog only to the groups
with an acceptable return on investment
© Palmatier 31

RFM Code Construction Where Higher
Numbers are Better
© Palmatier
F
M
One Sort
Five
Sorts
Twenty-five
sorts
Database
5
4
3
2
1
35
34
33
32
31
335
334
333
332
331
R
32

Test of a Expansion Strategy Using RFM
Analysis for Mailing
© Palmatier
RFM Cell
Breakeven
Analysis
Catalog mailing is
above breakeven
for customers with
these RFM codes
33

Lost Customer Analysis Informs AER
Strategies
 A firm contacts customers that have migrated away, to identify the cause for this change,
then works backward to fix the problem and ensure other customers don’t leave for the
same reason
 Takes a significant number of lost customers before a firm recognizes that it isn’t just
“normal” customer churn but rather an indication of an underlying problem
 Three-step process, which provides insights into both strengths and weaknesses:
1. Firms set regular intervals for contacting lost customers to identify the cause of their transition,
where they went, and potential recovery strategies
2. If the lost customer is not in the firm’s main target segment, firms could:
a. Change their acquisition criteria
b. Evaluate an expansion strategy to address a new subsegment of customers
3. If the lost customer is in the firm’s target market, firms should:
a. Fix the problem
b. Implement retention strategies to build brand and relational loyalty
 A choice model can benefit the lost customer analysis, and can inform analyses across all
AER stages, because it predicts the likelihood of observed customer choices/responses,
using data about that customer’s characteristics and past behaviors, as well as the firm’s
marketing interventions
© Palmatier 34

What are Choice Models?
 Analysis approach that attempts to determine the impact of
different factors (price, promotion) on consumer’s individual
choices (joining, cross buying, leaving)
 Most popular individual-level response model
 Uses past behavioral data
 No need to survey or get customer input (infers weights from past
customer's behaviors)
 Determines probability of choices and elasticities of marketing variables
 While regression analysis…
 Is not an individual model and uses continuous outcomes
 Gives weights of inputs, but not probabilities or elasticities
 Does not use “logit model,” which is better predictor of how people act by
capturing diminishing effects (inputs have less effect as you do it more)

Choice (Logit) Models Integrate Important
Characteristics of Consumer Behavior
 Assumes that consumers make rational choices based on their individual utility
with a diminishing sensitivity for marketing effort
 Thus, marketing has the highest effect on people who are “sitting on the fence”
(i.e., probability of .5 of choosing an alternative)
Marginal Impact
of a Marketing
Action
Probability of Choosing an Alternative
0.0 0.5 1.0
Low
High

Choice Models are Excellent for
Determining “Best” AER Strategies
 After determining AER positioning for each stage and ranking
personas on CLV, then you need to develop effective AER
marketing strategies
 Use database of past marketing actions and demographics linked
to actual customer responses (choices) in a stage as input to choice
model
 Choice model will provide:
 Elasticity for every input variable on outcome (e.g., how does age, kids,
credit, and direct mail impact choice)
 Elasticity = % change in output / % change in input
 Probabilities of customer’s choice (probability of upselling, retaining,
and can run on lists for acquisition targeting)
 Latent class choice models: “clusters” and runs choice model at
the same time (segments on elasticity to IVs, same things work for
segment)
© Palmatier 37

Observations / Choice
data
Choice
(0/1)
Income level
Intensity of
contact with
EMBA (internet,
mailing, call, visit)
Distance to
campus
Coworkers in
the program
1 1 3 7 1 7
2 0 6 4 7 3
3 0 6 4 7 7
4 1 7 7 1 7
5 1 2 7 1 7
6 0 5 6 6 3
7 0 3 1 6 3
Example of Choice Model on MBA Choice
 Inputs (no survey needed, infers importance of criteria
from behavior)
 Family income level
 Intensity of contact with EMBA office (internet, mail, call, visit)
 Distance to campus from home
 Coworkers in the program
© Palmatier 38

Customer Choice Determines Coefficient
Estimates, Probabilities, and Elasticities
© Palmatier
Coefficient Estimates
Variables / Coefficient
estimates
Coefficient
estimates
Standard
deviation
t-statistic
Income level 1.700397 1.172524 1.450202
Intensity of contact with
EMBA (internet,
mailing, call, visit)
23.18698 50.94818 0.455109
Distance to campus -1.0191 0.508133 -2.00548
Coworkers in the
program
2.092535 1.125706 1.858866
Const-1 -151.835 307.1951 -0.49426
Baseline n/a n/a
Elasticities of Income
level
Response Dummy
Response 0.245253 -0.38359
Dummy 0 0
Elasticities of Intensity
of contact with EMBA
(internet, mailing, call,
visit)
Response Dummy
Response 5.474283 -8.5622
Dummy 0 0
Elasticities of Distance
to campus
Response Dummy
Response -0.25866 0.404571
Dummy 0 0
Elasticities of
Coworkers in the
program
Response Dummy
Response 0.549165 -0.85894
Dummy 0 0
Elasticities
Probabilities
Respondents / Choice
probabilities
Response
probability
Dummy
probability
Predicted
Response
Predicted
Dummy
Observed
Response
Observed
Dummy
1 1.000 0.000 1 0 1 0
6 0.017 0.983 0 1 0 1
12 0.159 0.841 0 1 0 1
32 0.476 0.524 0 1 1 0
39

Choice Model Analysis
A choice model is a mathematical model that predicts how
the likelihood of an observed customer choice or response,
is influenced by a firm’s marketing interventions, and/or
customer characteristics.
DAT 3.2
Description
• To determine a customer’s most likely choice when faced with many product
alternatives.
• To determine the most important factors that influence customer choice likelihood.
• To segment and target customers according to the similarities in their choice drivers.
• To simulate the potential market share for various products on the basis of customer
choice.
When to Use It
How it Works
Example
A retailer is planning to introduce a store brand of bleach (Store Brand “C”) in their bleach category, which predominantly has two national brands
(Brand “A” and Brand “B”). The manager responsible for the store brand obtains transaction data from all the retail stores on the sales, price and
promotional efforts by Brand A and Brand B. With a choice model, the manager learns the weights and elasticities associated with price and
promotional efforts respectively. A 1% increase in the price of Brand A decreases the sales of Brand A by 2.7%, while it increases the sales of Brand B by
1.6%. A 1% increase in the price of Brand B decreases the sales of Brand B by 3.7%, while it increases the sales of Brand A by 1.2%. A 1% increase in
promotional intensity (a 10% price cut for 2 weeks) by Brands A and B increases their respective sales by 8% and 6% respectively. The model fits the
data very well, and hence the retail manager feel confident about the results. Also, the manager learns that frequent price promotions, and lower price
do help increase market share significantly. Hence, the retail manager decides to introduce the store brand (Brand C) as a low price, generic version of
bleach (price 20% below Brands A and B), with promotions held 20% more frequently than Brands A and B. The goal of the store brand introduction is
to steal market share from Brands A and B, especially in a category where price seems to have a large effect on sales (as learnt from a choice model).
• In a choice model setting, every individual is assumed to derive an unobserved product-specific utility from several product options. The individual is
assumed to pick the product option that provides the maximum utility. The dependent variable in a choice model is binary: every individual chooses (or
does not choose) a product option.
• Every product option’s attractiveness is assumed to stem from a finite set of attributes (e.g., brand name, price, advertising). The independent variables in a
choice model are the measure of the strength of attributes of each of the product option, e.g., product option 1 may have a low price, while product option 2
may have a higher price.
• The model uses the two inputs (dependent variable, independent variables) to estimate several outputs. First, it provides the weights (or coefficients) that
each attribute would have had to cause customers to pick a certain product. This provides the most important factors that influence customer choice
likelihood.
• Second, it is used in a predictive sense. For example, when we only observe product attributes, we can use the attributes and the weighs of the model to
predict the choices that are likely to be made by a new set of customers. This in turn can help a firm segment and target customers based on choice
likelihood.
• Third, the model can be used to simulate the market share of a product category, by adding up the product choices made by all customers faced with all
products. This can help manages plan their marketing efforts.
40
© Palmatier

Marketing Principle #2: All Customers
Change Managing Customer Dynamics
© Palmatier
Managing Customer Dynamics
Approaches & Processes
Lifecycle approach
Dynamic segmentation approach
Acquisition, expansion, retention (AER)
model
Lost customer approach
Analyses
Customer lifetime value (CLV)
Hidden Markov model (HMM)
Choice models
Factor, cluster, discriminant analyses
Inputs (CRM data) Outputs (AER)
Your Customers
• Individual customers’ sales,
margins, costs
• Behaviors/needs over
time/events
Past Marketing Programs
• Source of customers
• Past programs targeted at
specific customers
Lost Customers
• Cause of defection
• Characteristics of lost customers
Segmentation of Customers
• Customer personas
• Needs and CLV of personas
• Why and how they migrate
AER Positioning
Statements
• How best to position the firm in
each persona/AER stage
AER Strategies
• What marketing strategies work
best for each persona/AER stage
42

Inputs to the Customer Dynamics
Framework
1. The firm’s existing customer portfolio – ideally a firm’s CRM
system provides detailed customer-level data for the dynamic
segmentation analysis
2. Data linking past customer responses with specific marketing
programs and the programs’ cost
3. The qualitative and quantitative information gleaned from the lost
customer analysis, which reveals the causes of customer defection,
where they go, and potentially effective recovery strategies
© Palmatier 43

Outputs of the Customer Dynamics
Framework
1. A description of the firm’s customer personas and expected migrations
to understand how they change, including:
 Critical life event triggers
 The products and services they buy at different points in their
lifecycle migration
 When they stop buying and why
 How they feel at different stages in their lifecycle
 The CLV of customers in each persona
2. AER Positioning statements – how to best position the firm in each
persona/AER stage
3. AER strategies – what marketing strategies work best for each
persona/AER stage
© Palmatier 44

Process for Managing Customer Dynamics
To convert CRM, marketing program, and lost customer input
data into dynamic segmentation and AER positioning
statements and strategies, managers should follow a series of
steps:
1. Dynamic segmentation
2. Migration Paths and Triggers
3. Customer Lifetime Value of Segments and Migrations
4. AER positioning statements
5. AER strategies
© Palmatier 45

Dynamic Segmentation: Hotel Example
AcquisitionStage Expansion Stage Retention Stage
Learners
One Timers
Satisfied
Customers
Loyalists
Bored
Customers
Lost
Customers
Prospective
Customers
Upgraded
Customers
Path A
Path B Path C
Path D
1. Segment customers in each AER stage and name
the related personas
2. Find the migration paths, triggers, and CLVs of
each persona
3. Determine the AER positioning statement and
strategies (who, what, why, and how) for each
persona to maximize sales and profits, based on
the CLV of these personas
The main trigger point appears to be positive interactions with
the manager or key staff (20 percent of time).
The main trigger is being offered upgraded services on first
visit (20 percent of the time).
The main trigger is poor customer experience due to Internet
speed or insufficient work space (30 percent of the time).
The main trigger is either low perceived value or poor
interactions with hotel staff (30 percent of the time).
Migration Path Triggers
Steps to Dynamic Segmentation
Path A:
Path B:
Path C:
Path D:
© Palmatier 47

Markstrat Simulation: Making Decisions
When Dealing with Customer Dynamics
© Palmatier
Sales
Time
Managing Developed Markets (Sonites) as Customer Segments Change
Managing Developing Markets (Vodites) as Prototypical Customer Segments
Emerge
Innovators segment
Early
adopters segment
Followers
segment
Professionals
Savers
Growing
customer
segments
Changing
customer
needs
Pricing
pressures
Competitive
product
launches
High
Earners
Segment Descriptions
High Earners
• High income level
• Demand performance and convenience
• Purchase expensive products
Professionals
• Personal and professional usage
• Look for high quality, high-performance products
• Can afford expensive products
Savers
• Cautious in the way they spend their money
• Demand cheap, average convenience products
• Future growth rate could exceed forecasts
Shoppers
• Good product knowledge through comparison
• Demand high quality–price ratio
• Quite price-sensitive
Shoppers
48

Takeaways
 The second underlying challenge that firms face when making marketing
decisions is that all customers change. This principle can be either an
opportunity or a threat, depending on how well the firm understands and
manages it.
 Customer dynamics arise from five sources: Individual customers change
due to life events and move through typical lifecycles as they age.
Customer learning effects occur as customers gain knowledge about a
product category. Learning and experience effects also operate at a
societal level. Finally, each customer is situated in an environmental
context that is constantly changing, filled with outside entities trying to
change the customer’s perceptions and behaviors. Each source works
simultaneously and cumulatively to create customer dynamics.
© Palmatier 50

Takeaways
 Due to rapid technological and communication developments, the speed
at which customers change and their expectations about firms’ response
times have increased.
 There are three approaches to managing customer dynamics: lifecycle,
customer dynamic segmentation, and customer lifetime value
approaches.
 The lifecycle approach predicts that customers, products, and industries
go through similar lifecycles that can be used to inform marketing
decisions at different stages. This approach can be problematic though,
because it assumes an average rate of change.
© Palmatier 51

Takeaways
 The customer dynamic segmentation approach, with an AER model,
predicts that acquisition occurs when customers are just beginning to
interact with the firm (customer onboarding). Expansion is when the firm
tries to cross-sell or upsell customers and increase engagement, and
retention involves keeping customers who might otherwise tend to
migrate to competitors.
 Hidden Markov models (HMM) can uncover states, reflecting how a large set of
customer behaviors changes over time. A state is similar to a consumer
segment, describing common behaviors by a group of consumers at some point
in their relationship with the firm. Thus, HMM enables dynamic segmentation.
 As a powerful diagnostic tool, lost customer analysis can be integrated into a
customer dynamic segmentation approach. It often features a mathematical
choice model that predicts the likelihood of an observed customer choice or
response (e.g., joining, cross-buying, leaving), according to data gathered from
the firm’s marketing interventions and customer characteristics.
© Palmatier 52

Takeaways
 Customer lifetime value (CLV) seeks to capture the true contribution of
each customer, according to the migration path this customer is predicted
to follow throughout the relationship with the firm. This approach
beneficially provides guidance for making trade-offs and resource
allocation decisions among different AEP stages. It also can change a
firm’s culture, such that the focus is on profits as the sum of each
customer’s lifetime value, rather than the sum of a product line’s profits.
The firm then becomes more focused on customers, enabling firms to
detect and respond to market changes.
 The framework for managing customer dynamics uses three inputs: CRM,
marketing programs, and lost customer data. It produces three outputs:
dynamic segmentation and AER positioning statements and strategies.
© Palmatier 53

 TKL is a leading US-based distributor of electrical component products
 Three industry segments: construction, industrial, and original equipment
manufacturers (OEMs)
 Heavy fragmentation in industry, customers evolve as industry matures
(learning effects)
 Problem: market share and annual sales down, increasing customer churn
rates, drop in ratings of warranty, sales support, and delivery speed
 Need to manage customer dynamics (What product attributes are desired? How to
segment the market? Which customers to acquire and expand?)
 Data:
 Acquisition model: 1000 acquired and 1000 non-acquired customers
 Expansion model: 1000 retained and 1000 non-retained customers
 Drivers of acquisition/retention: price, warranty, delivery time, sales support,
industry group, firm size, centralized buying center
© Palmatier
Analytics Driven Case: Preempting and
Preventing Customer Churn at TKL
55

Analytics Driven Case: Preempting and
Preventing Customer Churn at TKL
 Results of acquisition model:
 Negative impact: increase in prices, sector: OEM’s
 Positive impact: increase in days of warranty, increase in sales support, sector:
construction and industrial
 No impact: decrease in days to deliver warranty
 Targeting and Positioning for Competitive Advantage
 Segmenting on Acquisition Probability: probability of acquisition higher for
larger sized firms in the construction and industrial sectors  market lowered
price, high sales support and high days of warranty to these customers
 Segmenting on Expansion Probability: probability of expansion was higher for
firms in the construction and OEMs sectors, larger firms, and firms with a
centralized buying center  market sales support and speed of delivery
© Palmatier 56

Readings
 Dealing with customer dynamics using AER modeling and strategies
(regression, CLV, and Bayesian)
 Getting the Most Out of All Your Customers (good overview of
tradeoffs between acquisition and retention uses regression)
 Customer Lifetime Value (explains CLV approach and how to use for
it for acquisition and retention of cohorts of customers)
 Know What to Sell, When, and to Whom (example of integrated
modeling techniques to simultaneously handle customer heterogeneity
and dynamics uses Bayesian models)
 Marketing Strategy: Chapter 3
57

Marketing Strategy - Customer Analytics Chapter03-2.4.pptx

Recommended

Recommended

More Related Content

Similar to Marketing Strategy - Customer Analytics Chapter03-2.4.pptx

Similar to Marketing Strategy - Customer Analytics Chapter03-2.4.pptx (20)

More from AparnaSachinKanchan

More from AparnaSachinKanchan (15)

Recently uploaded

Recently uploaded (20)

Marketing Strategy - Customer Analytics Chapter03-2.4.pptx