Customer Relationship Management: A Database Approach

1
Customer Relationship ManagementA Database
Approach
MARK 7397 Spring 2007
Class 2
James D. Hess C.T. Bauer Professor of Marketing
Science 375H Melcher Hall jhess_at_uh.edu 713
743-4175
2
Marketing Metrics
3
Traditional and Customer BasedMarketing Metrics

Traditional Marketing Metrics Market share Sales Growth Primary Customer Based metrics Acquisition rate Acquisition cost Retention rate Survival rate P (Active) Lifetime Duration Win-back rate
Popular Customer Based metrics Share of Category Requirement Size of Wallet Share of Wallet Expected Share of Wallet Strategic Customer Based metrics Past Customer Value RFM value Customer Lifetime Value Customer Equity
4
Primary Customer Based Metrics

• Customer Acquisition Measurements
• Acquisition rate
• Acquisition cost
• Customer Activity Measurements
• Average interpurchase time (AIT)
• Retention rate
• Defection rate
• Survival rate
• P (Active)
• Lifetime Duration
• Win-back rate

5
Acquisition Rate

• Acquisition defined as first purchase or
  purchasing in the first predefined period
• Acquisition rate () 100Number of prospects
  acquired / Number of prospects targeted
• Denotes average probability of acquiring a
  customer from a population
• Always calculated for a group of customers
• Typically computed on a campaign-by-campaign
  basis
• Information source
• Numerator From internal records
• Denominator Prospect database and/or market
  research data
• Evaluation
• Important metric, but cannot be considered in
  isolation

6
Acquisition Cost

• Measured in monetary terms
• Acquisition cost () Acquisition spending () /
  Number of prospects acquired
• Precise values for companies targeting prospects
  through direct mail
• Less precise for broadcasted communication
• Information source
• Numerator from internal records
• Denominator from internal records
• Evaluation
• Difficult to monitor on a customer by customer
  basis

7
Average Inter-purchase Time (AIT)

• Average Inter-purchase Time of a customer
• 1 / Number of purchase incidences from the
  first purchase till the current time period
  
• Measured in time periods
• Information from sales records
• Important for industries where customers buy on a
  frequent basis
• Information source
• Sales records
• Evaluation
• Easy to calculate, useful for industries where
  customers make frequent purchases
• Firm intervention might be warranted anytime
  customers fall considerably below their AIT

8
Retention and Defection

• Retention rate () 100 Number of customers in
  cohort buying in (t) buying in (t-1) / Number of
  customers in cohort buying in (t-1)
• Avg. retention rate () 1 (1/Avg. lifetime
  duration)
• Avg. Defection rate () 1 Avg. Retention rate
  

Plotting entire series of customers that defect
each period, shows variation (or heterogeneity)
around the average lifetime duration of 4 years.
9
Customer Lifetime Duration when the Information
is Incomplete

Buyer 1
Buyer 2
Buyer 3
Buyer 4
Observation window Buyer 1 complete
information Buyer 2 left-censored Buyer 3
right-censored Buyer 4 left-and-right-censored
10
(No Transcript)
11
Life Table with only right censoring

Buyer 1
Buyer 2
Buyer 3
Buyer 4
t
Buyer 1 Withdrew late (still active when last
observed) Buyer 2 Withdrew early (still active
when last observed) Buyer 3 Terminated late (did
not survive past observed date) Buyer 4
Terminated early (did not survive past observed
date)
12
(No Transcript)
13
(No Transcript)
14
(No Transcript)
15
Basic Survival Math
S(t) probability that customer will survive
until at least time t 1-F(t) where F(t)
is the traditional cumulative distribution f(t)
probability that survival ends at t
-S(t)F(t)
1.0
S(t)
f(t)
t
0
S(t0t)
---------- Conditional Survival probability
that customer lasts until at
least t0t given that they lasted until t0
S(t0)
16
Hazard Rate and Related Stuff
f(t)
h(t) -------- Hazard Rate prob that
survival ends at t given
that customer makes it to t
S(t)
H(t)cumulative hazard rate -lnS(t) S(t)exp-
H(t)
Constant Hazard Rate Model
h(t)h0, a constant in time ? H(t)h0
t ? S(t)exp(-h0 t) ? f(t)h0exp(-h0 t)
Et 1/h0
Et0t customer made it to t0 t0 1/h0
17
(No Transcript)
18
Proportional Hazard Rate Model
What if the event varies with customer/situational
factors X? h(t) hB(t)
? exp(bX), where hB(t) is the baseline hazard
rate.
The baseline hazard rate hB(t) is metaphorically
like an intercept because when X0, then
exp(bX)1.0 ? h(t) hB(t).
If bX gt 0, then exp(bX)gt1.0, so hazard rates
increase above baseline. If bX lt 0, then
exp(bX)lt1.0, so hazard rates decrease below
baseline.
The coefficients b are chosen in a
regression-like fashion, accounting for customer
factors and censored data. In SPSS this is done
in Survival/Cox Regression.
Why not have hB(t) ? bX? Hazard rates must be
positive!
19
Cox Regression Survival Analysis
20
(No Transcript)
21
(No Transcript)
22
(No Transcript)
23
Proportional Hazards Assuming Constant Baseline
Hazard
h(tAge) hB(t) ? exp(bX) 0.108 ? exp(-0.065
Age)
Et0t customer of Age made it to t0 t0
exp(-bX)/h0
t0 exp(0.065 Age)/0.108
E t Age made it to t0 exp(-bX)/h0
exp(0.065
Age)/0.108
24
(No Transcript)
25
Summary

• In the absence of individual customer data,
  companies used to rely on traditional marketing
  metrics like market share and sales growth
• Acquisition measurement metrics measure the
  customer level success of marketing efforts to
  acquire new customers
• Customer activity metrics track customer
  activities after the acquisition stage
• Lifetime duration is a very important metric in
  the calculation of the customer lifetime value
  and is different in contractual and
  non-contractual situations