how to increase customer loyalty using cluster analysis and decision tree analysis of customer...

How to Increase Customer Loyalty Using Cluster Analysis

and Decision Tree

Analysis of customer behavior and service

design

What is the most important factor in CRM or servicing

customers?1.

2. Identify Needs of customers

3. Loyalty

Questions on brand loyalty

• Why is brand royalty so important to most companies?

Perspectives of Brand Loyalty

• Customer loyalty as customer’s commitment or attachment to a brand, store, manufacturer, service provider

Or• Entity based on favorable attitudes and

behavioral responses, such as repeat purchases

• Ex) ‘Red Devil’ for national soccer team

Organizations and their loyal customers

• Airlines• Credit card companies• Internet stores• Banks• Car dealers• Cell phone

Brand Loyalty as Behavior

• Rate of repurchasing [examples] Chicago Bulls, Cubs, Heinz, Crispy Cream donuts, Starbuck• Proportion of purchase = the number of time the most frequently purchased brand

total number of times the product category is purchased

5 types of customer behaviors

• Undivided loyalty: A A A A A A A A A• Occasional switcher: A A A B A A A C• Switched loyalty: A A A A A B B B B B• Divided loyalty: A A A B B B A A A B B

B• Indifference: A B C D A B C D A B C D

Churn rate

• Switch from one brand to other brand

• Customers RFM (key variables in market segmentation, also understanding loyal customer)

- recency - frequency - monetary: average purchase size

Brand loyalty as attitude

• Why customer has loyalty on a brand?

[example] bank, internet shop, airlines, credit

cards• Brand loyalty is a behavioral

response to an attitude toward a brand

Loyalty versus inertia

Inertial loyalty

• Habitual

• Latent loyalty -strong commitment -low repeat purchase [example] SONY PS2, Nintendo

Factors that affect customer loyalty

(Intimacy)

Attitudinal and behavioral components of loyalty

15

Personalization of Service in the Web Using Intimacy Theory,

Cluster Analysis, and Decision Tree

: How to increase intimacy with customers

Introduction

• Face – to – face• Object – medium - object

– Digital interaction with Internet

• Setting Interpersonal Distance– Intimacy theory– Web interface development

Research Background• Designer , Web Master based pages…

– Personalization, categorization- User , customer based web pages

• Relations adjustment of interface by emplyee

Frequent Customer

Not Frequent

Clerk

Proxemics• People surround themselves with a

“bubble” of personal space(Hall, 1966)

Intimate distance: 0 ~ 1.5 feet(0.45 m)

Personal distance: 1.5 ~ 4 feet(1.2 m)

Social distance: 4 ~ 12 feet(3.6 m)

Public distance: more than 12 feet

person

Machine Learning Modeling

• Prediction(supervised learning)– Inputs output– Neural networks, rule induction,

regression

• Clustering(unsupervised learning)– Inputs similarity– k-means

• Association– Input output

Cluster Analysis of Customers

Cluster Distribution

Cluster

Ratio(%)

CountIntimacy

Level

A 20.86 34 2.41

B 25.77 42 3.02

C 24.54 40 3.85

D 28.83 47 2.87

• Cluster A– if (Rep = good) And (period = 6

months) Or (rep = excellent) Or (Rep = good) And (visit = weekly)

Rule Set for each cluster

Cluster B if (Rep = good) And (period = 1 year) Or (rep = good) And (visit = monthly) And (period =

1year) Or (rep = good) And (visit = monthly) And (period =

1month) Or (rep = good) And (visit = monthly) And (period =

2years) Or (rep = good) And (visit = monthly) And (period =

6months) Cluster C

if (Rep = good) And (visit = 1 year) Or (Rep = good) And (visit = > 1 year Or (Rep = good) And (visit = monthly) And (period = > 2

years) Or (Rep = good) And (visit = daily)

Cluster D if (Rep = middle) And (period = 1month) Or (Rep = middle) And (period = 2years) Or (Rep = middle) And (period = >2years

• Physical distance

Analysis from Rules/Decision Tree

object object

Psycholgical distance

reputationreputation No. of visitsNo. of visits

Membership periodMembership period

X

Y

Dynamic Web Page Personalized

Main Page

Logged/personalizing

Web Page Type IFor Cluster A

Web Page Type IIFor Cluster B

Web Page Type IIIFor Cluster C

Web Page Type IVFor Cluster D