Assignment 2
Customer Segmentation
The Case
The young and thriving credit card company FISA has just entered the Australian market
approximately 2 years ago and is rapidly growing there. You are a marketing analyst at
FISA. In a meeting with your Head Analyst Alec, you wonder why FISA is using an
undifferentiated approach in managing their customer relationships in Australia. Surprisingly,
it turns out that no one at FISA has a good understanding of your current customer
segments. Since FISA hast just entered the market 2 years ago, no one has undertaken a
customer segmentation analysis yet and marketing efforts have not really been adapted to
different segments. Since competition is fierce, Alec and you decide that you should perform
a Cluster Analysis in order to derive implications for adapting your offerings better to
customer needs.
The Research Goal
Your research goal is to segment your current customers according to their credit card
usage patterns (i.e., what do they use their credit cards for?) while also considering further
potential to extend the relationship and avoiding increased risk of default. Ideally you want to
understand which prototypical usage patterns can be observed among your customers in
order to tailor credit card plans to the different segments’ needs and to provide suitable
information material for each customer segment.
Data Sample and Variables
You talk to the Director of Data Management and he makes sure that you receive a suitable
sample. The sample you receive contains the annual data of a randomly drawn sample of
Australian customers, which accounts for approximately 5% percent of FISA’s current
customer base in Australia.
The data set contains the following variables:
PURCHASES : Total amount of purchases made from account.
ONEOFFPURCHASES : Maximum purchase amount done in one-go; the whole amount of
purchases that are paid in one go.
INSTALLMENTSPURCHASES : Amount of purchase done in an installment plan; the whole
amount of purchases that are made in installments (i.e., you buy a fridge for $ 1,000 but pay
for it in ten installments of $100).
CASHADVANCE : Cash in advance given by the user; cash in advance is if you withdraw
cash from your credit card account.
PURCHASESFREQUENCY : How frequently the purchases are being made, score between
0 and 1 (1 = frequently purchased, 0 = not frequently purchased)
ONEOFFPURCHASESFREQUENCY : How frequently Purchases are happening in one-go,
score between 0 and 1 (1 = frequently purchased, 0 = not frequently purchased)
PURCHASESINSTALLMENTSFREQUENCY : How frequently purchases in installments are
being done, score between 0 and 1 (1 = frequently purchased, 0 = not frequently
purchased)
CASHADVANCEFREQUENCY : How frequently the cash in advance being paid, score
between 0 and 1 (1 = frequently purchased, 0 = not frequently purchased)
CREDITLIMIT : Limit of Credit Card for user
PAYMENTS : Amount of Payment done by user; payments are the amounts the credit card
holder owed the credit card issuer that were actually paid back to the credit card issuer.
MINIMUM_PAYMENTS : Minimum amount of payments to be made by user; the cardholder
must pay a defined minimum portion of the amount owed by a due date, or may choose to
pay a higher amount. The credit card issuer charges interest on the unpaid balance if the
billed amount is not paid in full (typically at a much higher rate than most other forms of
debt). In addition, if the cardholder fails to make at least the minimum payment by the due
date, the issuer may impose a late fee or other penalties.
CREDIT_EXAUST: This variable was calculated by dividing PURCHASES by CREDITLIMIT.
It was suggested that this variable could be useful as it informs whether users’ credit card
purchases are capped. The rationale is that if users exhaust their credit card limit there might
be further potential to develop the relationship by enhancing their credit card limit.
PAYMENT_DIFFERENCE: This variable was calculated by subtracting
MINIMUM_PAYMENTS from PAYMENTS. The rationale is that customers who do not cover
their minimum payments pay much higher interests plus extra fees and therefore cumulate
debt, which might entail financial distress in the long-run. (Tipp: In your interpretation,
consider what low and negative values of this variable indicate!)
Data Preparation
You start mining the data. Please first load the data file into R. Then write a code to get an
overview over the data.
Since the data has been cleaned by a colleague, you decide not to delete any cases and
you do not replace any observations by NA.
Two-step Cluster Analysis
Next, you perform two-step cluster analysis. In the hierarchical cluster analysis, please use
the same methods to calculate distances between and within clusters you used in the tutorial
(i.e., euclidean, ward.D2). You use the following variables in your cluster analysis:
· ONEOFF_PURCHASES,
· INSTALLMENTS_PURCHASES,
· CASH_ADVANCE,
· ONEOFF_PURCHASES_FREQUENCY,
· PURCHASES_INSTALLMENTS_FREQUENCY,
· CASH_ADVANCE_FREQUENCY,
· CREDIT_EXAUST,
· PAYMENT_DIFFERENCE
As a general rule, please standardize all variables before you use them. Please produce the
clustering solution you find most appropriate. Once you have done that, please answer the
questions in the Moodle assignment.