DATA MINING IN CRM

ANALYTICS-INTELLIGENT
MANAGEMENT OF PRODUCT
LIFE CYCLE

Presentataion by Mr.Suresh Shirha 1

ttikar
 AN EFFECTIVE CRM

 companies need to match products &

campaigns to prospects & existing
customers.

Presentataion by Mr.Suresh Shirha 2

ttikar
 OPERATIONAL CRM Vs DATA
MINING (ANALYTIC CRM)

 Operational CRM- Software that

creates a database which provides a
consistent picture of customer’s
relationship with the company,
through specific applications. e. g.
Sales Force Automation, Customer
Service software etc.

Presentataion by Mr.Suresh Shirha 3

ttikar
 OPERATIONAL CRM Vs DATA
MINING (ANALYTIC CRM)

 Whereas Data Mining is a process

that uses variety of Data Analysis &
modeling techniques to discover
patterns & Relations in a Database
that may be used to make accurate
predictions.

Presentataion by Mr.Suresh Shirha 4

ttikar
 STEPS IN DATA MINING

 Describe the data- summarize its

statistical attributes (e.g. means &
standard deviations), visually review
it using charts & graphs & look at
distribution of values.

Presentataion by Mr.Suresh Shirha 5

ttikar
 STEPS IN DATA MINING
 Build a predictive model- Using
patterns from a known set of results
& then test that model on your
original sample. E.g. in Classification-
one predicts in what category an
element of data might fall. In
Regression one predicts a number
such as probability of that element of
data to respond to your action.
Presentataion by Mr.Suresh Shirha 6
ttikar
 STEPS IN DATA MINING

 Determine propensity or create a

profile- a score to give a measure to
the propensity of customer behavior
& create a set of characteristics that
could segment the customer into
groups with similar behavior.

Presentataion by Mr.Suresh Shirha 7

ttikar
 DATA MINING THROUGH
CUSTOMER LIFE CYCLES
 Data mining could improve your
profitability in the following stages of
life cycle either by integration with
Operational CRM or independently.
 Acquiring customers.
 Increasing the value to the customer.
 Retaining good customers.

Presentataion by Mr.Suresh Shirha 8

ttikar
 AQUIRING NEW CUSTOMERS
THROUGH DATA MINING
 A company normally could win new
customers either through change in the
method of solicitation by reaching
customers through different media or
offering attractive ‘freebies’.
 Or through data mining techniques to
build a predictive model of who is
likely to respond favorably or use the
method of scoring the existing
customers.

Presentataion by Mr.Suresh Shirha 9

ttikar
INCREASING THE VALUE TO THE
EXISTING CUSTOMERS

Technique of Cross Selling- The

Company may offer more features to its
product or facilitate better choosing &
buying facilities by
 The technique of “Clustering” to see
which products & add-ons grouped
more naturally

Presentataion by Mr.Suresh Shirha 10

ttikar
INCREASING THE VALUE TO THE
EXISTING CUSTOMERS

 The technique of “profiling” the

customers to see which group of clients
readily tried new variants or which ones
stayed loyal to old brands.
 Use the “personalization” to proactively
announce new products to a select group
chosen by data mining.

Presentataion by Mr.Suresh Shirha 11

ttikar
RETAINING LOYAL CUSTOMERS
 The cost of winning over a new
customer grossly exceeds that of
retaining a customer (Sometimes as
high as 7 times).
 The company at first will use data
mining to identify repeat buyers.
 Next it would study the attrition pattern
of once good customers (who left them
& why?).

Presentataion by Mr.Suresh Shirha 12

ttikar
RETAINING LOYAL CUSTOMERS

 Next they would design programs to

identify signs of discontent in time &
offer new products/feature to retain.
 Eventually they would build data mining
programs that would predict which offer
works the best for customer retention.

Presentataion by Mr.Suresh Shirha 13

ttikar
APPLYING DATA MINING TO CRM

 Define business problem.

 Build marketing database.
 Explore data.
 Prepare data for modeling.
 Build model.
 Evaluate model.
 Deploy model & research on results.

Presentataion by Mr.Suresh Shirha 14

ttikar
DEFINE BUSINESS PROBLEM

 Each CRM application has an end goal or

objective. Depending on the end goal
appropriate models need to be built. A
clear statement of the objective would
therefore also include the way of
measuring effectiveness of your CRM
project.
 The next four steps take the maximum
time as a lot of iterations of data building
& model building are done as you go along
& learn from the model.
Presentataion by Mr.Suresh Shirha 15
ttikar
BUILD MARKETING DATA BASE

 Independent of operational &

corporate database. Multiple
databases from customer database,
product database & transaction
database needs to be standardized,
integrated & consolidated.

Presentataion by Mr.Suresh Shirha 16

ttikar
 EXPLORE THE DATA

 To understand the data better,

various summaries such as averages,
standard deviations & pivot tables for
multidimensional comparison are
made. Graphing & visualization like
histograms, scatter plots give multi-
dimensional views.

Presentataion by Mr.Suresh Shirha 17

ttikar
PREPARE DATA FOR MODELLING
 There are 4 steps -
 Select the variables on which to build
the model.
 Construct new predictors derived from
new data.
 Select a sub-set or sample of the entire
data on which to build models.
 Transform variables in accordance with
the requirements of the algorithm.

Presentataion by Mr.Suresh Shirha 18

ttikar
 MODEL BUILDING
 It is an iterative process where alternative
models are tried to get the one most
suitable.
 Most CRM applications are based on a
protocol called supervised learning. One
starts with a customer information on
which desired outcome is already known.
This you try on a test basis on the new
data, first on a smaller group & then on
the whole of data.

Presentataion by Mr.Suresh Shirha 19

ttikar
EVALUATING THE RESULTS
 Some methods are,
 Lift- measures the improvement
achieved by the predictive model.
 Rise-Other measure is relative rise in
profits.
 ROI-Third measure which is most
acceptable is ROI (Return on
Investment).

Presentataion by Mr.Suresh Shirha 20

ttikar
INCORPORATING DATA MINING IN
CRM SOLUTION

 The data mining is most of the times

combined with the knowledge of
domain experts. The data mining is
built into the system depending on
the type of customer interaction
‘Outbound’ or ‘Inbound’.

Presentataion by Mr.Suresh Shirha 21

ttikar
OUT-BOUND INTERACTIONS & IN-
BOUND TRANSACTIONS

 Out-Bound-In this case the company

reaches out to the customer by a
desired action using any one of
mode of contact.
 In-Bound-In this case the customers
contact the company with a need for
action.

Presentataion by Mr.Suresh Shirha 22

ttikar
PRODUCTIVITY IN DATA MINING

 Intelligent transformations done in

the individual elements of data to
include multi-elemental features,
decide the productivity of any data
mining exercise.

Presentataion by Mr.Suresh Shirha 23

ttikar
 THE ROUTE TO A SUCCESSFUL
BUSINESS

 Depends on how effectively you can

use information about customer
needs to transmit into more profits
for you. For this it is necessary that
‘The Operational CRM ‘is aptly
supported by ‘Analytical CRM’, with
accurate predictive capabilities.

Presentataion by Mr.Suresh Shirha 24

ttikar
TECHNIQUES USED IN DATA
MINING

DECISION TREES
 NEURAL NETWORKS
 CLUSTERING

Presentataion by Mr.Suresh Shirha 25

ttikar
Cookin
17,95
g-
Total 16,126 4 DATA
19,542 ON COOKING
20,846 21,548 EQUIPMENT
21,917 23,315 SALES
25,390 23,809 24,641 24,961
Electri
c
Range
s-
Total 4,240 4,639 4,983 5,026 5,066 5,338 5,622 6,145 6,194 6,317 6,410

Freesta
nding 3,177 3,481 3,785 3,826 3,842 4,030 4,238 4,612 4,677 4,775 4,858

Built-In 617 652 705 706 726 780 841 963 975 993 997

Surface
Cookin
g Units 446 506 493 494 498 528 543 570 542 549 555

Gas
Range
s-
Total 2,744 2,950 3,137 3,176 3,036 3,268 3,419 3,719 3,755 3,809 3,836

Freesta
nding 2,391 2,543 2,698 2,729 2,580 2,781 2,897 3,124 3,131 3,170 3,193

Built-In 73 71 72 70 72 71 67 67 64 63 61

Surface
Cookin
g Units 280 336 367 377 384 416 455 528 560 576 582

Micro Presentataion by Mr.Suresh Shirha 26

wave 10,36 11,42 ttikar 13,31 14,27 13,86
Ovens 9,142 5 2 12,644 13,446 1 4 15,526 0 14,515 14,715
 OBJECTIVE OF THE STUDY
There is a considerable variation in
the sales of a set of local dealers in a
Metro city.
 The company expects that by & large
the sales of all the dealers in the
town should be comparable.
 The study wants to find out what
measures need to be taken to boost
the laggards.

Presentataion by Mr.Suresh Shirha 27

ttikar
 DECISION TREES

Decision Trees are a way of representing a series

of rules that lead to a class or value.
The basic components of a decision tree are,
 The Root node- This is the top decision node
by which a common test is applied.
 The branch node- There could be several
branch nodes, starting at root node & fanning
out in various branches.
 The leaf node- Through this, one finally
reaches the end of the tree.

Presentataion by Mr.Suresh Shirha 28

ttikar
 DECISION TREES
 The root node- Segregate the dealers
sales wise into groups of incremental
sale of 2000 units.
 Branch node- Segregates into

domestic/ commercial units wise

sales.
 Leaf node-Segregate into repeat

buyers/First time buyers.

Presentataion by Mr.Suresh Shirha 29

ttikar
 NEURAL NETWORKS
 Neural Networks offer a means of
effectively modeling large and
complex problems involving several
hundred predictor variables.
 Neural nets are commonly used for
Regressions.
 Neural nets start with Input layer,
where each node corresponds to a
predictor variable.

Presentataion by Mr.Suresh Shirha 30

ttikar
 NEURAL NETWORKS
 Input layer is connected to a number of
nodes in hidden layer. Each node in this
layer takes a set of inputs & multiplies
them by a connection weight age, adds
them together, applies a function (called
activation or squashing) & passes the
output to next layer.
 Finally on to Output layer which consists of
one or more response variables.

Presentataion by Mr.Suresh Shirha 31

ttikar
 NEURAL NETS
 Input Layer- e.g. take sales in the sets
of 1000.
 Hidden Layer-applies connection
weight age of seasonal variation to
each reading , adds them & applies
function of change in GDP, sends to
Final layer.
 Final layer- applies various variables
like a) competitive presence,
ineffectiveness of the dealer etc.
Presentataion by Mr.Suresh Shirha 32
ttikar
 NEURAL NETWORKS
 A typical example of neural network
is the ‘Back propagation’ in training,
where parameters like speed of
conversion are applied.
 Successful implementation of Neural
Networks involves very careful data
cleansing, selection, preparation &
pre-processing.

Presentataion by Mr.Suresh Shirha 33

ttikar
 CLUSTERING
 Classification is a way to segregate data
into already known groups &
Segmentation refers to identifying groups
that have common characteristics,
Clustering is a way to segment data into
groups that are not precisely defined.
 Clustering divides a database into

different groups. Clustering helps to

identify different groups, whose members
have common features.

Presentataion by Mr.Suresh Shirha 34

ttikar
CLASSIFICATION, SEGMENTATION
& CLUSTERING
 CLASSIFICATION- Segregate into known
groups, e.g. type of ovens sold i.e.
Microwave +convection+ grill.
 SEGMENTATION-Identify sales made to

groups that have common characteristics

e.g. households, hotels, industrial canteens
etc.
 CLUSTERING-Identify groups whose

members have common features, e.g.

locality wise, income group wise.
Presentataion by Mr.Suresh Shirha 35
ttikar
 A CASE STUDY
 ICICI Prudential a pioneering private sector
Insurance company was launched in India
in 2000 & today it is a major challenger to
LIC Corporation.
 Today it stands at 1900 branches,210,000
advisors & 7 ‘bancassurance’ partners with
10 million policies issued since inception .
 It’s USP is to offer customer a wide choice
of products.

Presentataion by Mr.Suresh Shirha 36

ttikar
 A CASE STUDY

 Their capital infusion stands at

Rs.4780 crores.
 Products developed after a thorough

research on customer needs.

 Easy accessibility to the customer.

 Smooth & hassle free claim

settlement procedure.
Presentataion by Mr.Suresh Shirha 37
ttikar
 A CASE STUDY
 The entire staff is regularly trained &
equipped with the latest sales
material .
 Robust risk management &

underwriting practices result in strong

core values of Integrity, Customer
First, Boundary less, Ownership and
Passion.

Presentataion by Mr.Suresh Shirha 38

ttikar
 A CASE STUDY
 The entire staff is regularly trained &
equipped with the latest sales
material .
 Robust risk management &

underwriting practices result in strong

core values of Integrity, Customer
First, Boundary less, Ownership and
Passion.

Presentataion by Mr.Suresh Shirha 39

ttikar
HOW THEY MANAGE CRM DRIVE?
 It operates on a multi-channel
distribution strategy including
advisors, banks, direct marketing
staff & corporate agents.
 It has made rapid strides in

Retirement & education solutions.

 Works on human emotions of trust,
togetherness & protection.

Presentataion by Mr.Suresh Shirha 40

ttikar