Analytical CRM

What is customer related data

◼ Any data related to the customer which is not

only about the customer but for the customer
also
◼ Can be historical, current and future
perspective
◼ Access to data about customer not only
restricted to functional areas such as sales,
marketing & customer service but also to third
parties
◼ Foundation for the execution of CRM strategy
Corporate Customer Related Data

◼ Most companies do not have just one customer

database – there are a number of databases
◼ These databases capture customer related data
from a number of different functions like
sales, marketing, accounts etc
◼ Might be managed by different people, have
past or future perspective, might be about
individual customers, groups
Exercise

State the types of structured & unstructured data

◼ Feedback form in a restaurant
◼ Sales Manager of an FMCG company
◼ Influencer on social media
 Structured Data
◼ Most of the data in databases is structured.
◼ Data stored in a fixed and named field in a record or file is called
structured data
◼ Structured data can be stored in 3 types of databases
▪ Hierarchical
▪ Network
▪ Relational
Relational Database

◼ Concept of primary
key like for sales
database each
customer is assigned
a unique no. which
is the primary key
◼ Share a common
structure of files,
records and fields (
tables, rows and
columns)
Hierarchical and Network
• Oldest form of database
• Child can only 1 parent but parent
can have many children.
• You may have to work through
various levels before arriving at what
you want
• For parent file to access titles, the
route is through authors and so on.

• Child can have many parents

• The Web is one such large network
structure
Unstructured data
◼ Unstructured data take the form of textual or non textual data
– unstructured data could be word docs, ppts, phone
recordings, agent notes
◼ Huge increase in unstructured data produced by users of social
media and elsewhere – this momentum in the growth of
unstructured data is expected to accelerate in the future and is
known as “Big data”
◼ Big data presents an opportunity only if the data can be
interpreted and acted upon
◼ 3 V’s of analysis
▪ Volume (amount of data)
▪ Velocity (the speed at which the data is generated)
▪ Variety (the kind of data available
Big Data

◼ Big data is the art of collecting large data sets

from traditional and digital sources to identify
trends & patterns
◼ Improve what they know about customers needs
and wants
◼ People are increasingly “ok” to hand over their
data for prodicts/services made for them
◼ 3 V’s of analysis
▪ Volume (amount of data)
▪ Velocity (the speed at which the data is generated)
▪ Variety (the kind of data available)
SAS Text Miner

SAS Text Miner is a Text Analytics application

that converts unstructured text information into
structured information that can be used for
generating insights.

Applications in CRM
Social media sentiment analysis
Automatic Call routing
Automatic email forwarding
Root cause analysis
Sentiment about brand, company, service, employees
Sentiment ANalysis
Data quality

◼ Verifying the data : The data should be entered exactly as

found in the original source
◼ Validation of data : Data should not have misspelt names,
incorrect titles and inappropriate salutations
◼ De-duplication of customer data – Customer data should not
be duplicated like Micheal Sethi is written as Micheal S and
Micheal Sethi though both are same customers
◼ Removing a record that should be retained
◼ Retaining a record that should be removed
◼ Merge and Purge : When 2 or more databases are merged ( eg.
Marketing and customer service) to avoid any duplication of
customer data
START – Desirable data quality
attributes
◼ S : Shareable - Multiple users/departments may require to
access same data at same time
◼ T: Transportable - The data needs to be available wherever
and whenever required by the users
◼ A : Accuracy - The data quality should be good and free
from any inaccuracies like duplication, misspelt etc
◼ R: Relevant – Data should be relevant to make better
business decisions
◼ T: Timely data – Users should have access to use the data at
right time for providing customer offers and making
business decisions
◼ S: Security – No compromises should be made on data
security about the customers
Datawarehousing

◼ Date Warehouse are repositories of large

amounts of operational, historical and other
customer-related data
Data warehouse

◼ Subject oriented – data organized around the essential

subjects of the business – customers and products
rather than around applications such as inventory
management or order processing
◼ Integrated – It is consistent in the way that data from
several sources are extracted and transformed
◼ Time-variant – data are organized by various time
periods
◼ Non-volatile – The warehouse data is not updated in
real time. There is periodic bulk uploading of
transactional and other data
Building a data warehouse
Identify the sources of
Data

Where are the data

stored

Extract the data from

these systems
ETL ( Extract
Transform the data in
Transform &
standardized and
Load)
clean format

Upload the data from

these systems

Update/Refresh the
data in the warehouse
 Benefits of data warehouse

◼ Data from multiple sources/databases can be stored at a

single place
◼ Enables a single view approach for searching/using data
◼ Historical data is available
◼ Improves data quality by removing duplication, errors
◼ Regular updating of data
◼ Restructures the data for business purpose
◼ Add value to CRM through customer analytics such as
decision support systems, queries to business questions
Simple DW Architecture
Data
Mart
Marketing

Sales
ETL
Integrat Data DM
Warehouse
ion
Billing Layer
DM

Supply Chain
Data Configuration for CRM
Analytics
Reporting

CRM
Data Analytic OLAP
Warehouse
s

Data Mining
Reporting – Generate Analytical
Insight
◼ Provides simple list of information such as key
accounts, annual revenues, product wise sales etc
◼ Can be standardized ( pre defined ) or query based
◼ Standardized reports are difficult to customize
◼ Query based reports provides a selection of tools
which can be used to construct a specific report
required by the users
◼ Can be generated to users in an array of visualization
tools such as tables, charts, graphs , plots, maps etc
◼ Standard report can be generated in an excel sheet for
further analysis
Reporting Example
OLAP – Online Analytical
Processing
◼ Extremely valuable to users from different functions like sales, marketing,
customer service etc who can ask different business questions
◼ For example – A sales person can analyze their territory for sales and
profit by customer, a customer service person can analyze call response
rates and resolution time by customer and a campaign manager can
analyze the campaign effectiveness day wise, product wise and campaign
type wise
◼ Can support decisions in real time, for example, propensity to buy
measures can be delivered to a call centre agent when the customer is on
the phone
◼ Information delivery mechanism is improved by making the information
available on the desktop in web browser interface with graphical layouts
and drill-down
◼ Some of the major vendors for OLAP are Qlik, Microsoft, IBM, SAS, SAP
and Oracle
Datawarehousing & Datamining

Data Warehousing
provides the Enterprise
with a memory

Data Mining provides the

Enterprise with strategic
intelligence
Meaning of data mining
▪ Non-trivial extraction of implicit, previously unknown and potentially
useful information from data
▪ Exploration & analysis, by automatic or semi-automatic means, of
large quantities of data in order to discover
meaningful patterns
◼ Data mining involves the use of sophisticated data analysis tools to
discover previously unknown, valid patterns and relationships in large data
sets. These tools can include statistical models, mathematical algorithms

◼ Data Mining also helps in predicting future trends and behaviour allowing
business to make pro-active and knowledge driven decision. It is an
important CRM enabler

◼ Enables data exploration, data analysis, and data visualization of very large
databases at a high level of abstraction, without a specific hypothesis in
mind
Lets discuss…

Visit any Retail store/Banks or any other

relevant establishments and understand
◼ What kind of customer data are being
captured by the retail store at their point of
sales?
◼ What are the different modes through which
these data are collected?
◼ List the usage of the data collected by the
company
Levels of data mining

-Depending on the level at which we are mining, data

mining operations can be divided into two categories:
-
Aggregate or Macro Level (without looking at any
customer in particular, we try to find general customer
behaviour, trends & preferences from a large database
such as market basket purchases from a retail store)

Individual or Micro Level (attempt tracking a particular

customer & try to further the relationship by acting in a
timely way to his/her choice and preferences)
Aggregate or Macro Level

◼ Gives broad overview of the data. While dealing

with issues pertaining to a customer segment.
Classification, Regression and link analysis are the
most common operations used at this level

◼ E.g., when customers of a retail store are

segmented by profitability criteria, we obtain
clusters who are profitable to various extents.
Mining can identify the characteristics of each of
these segments
Individual or Micro level
◼ As interactions of the individual with the firm increases, the firm
obtains more data about him/her. Relationship can be strengthened
by offering individualised value adding propositions
◼ For this, we need to track the customer and mine at indi/micro
level
◼ A firm takes up micro mining to build a detailed customer profile
of a regular customer
◼ Data mining at this level might be expensive if the data mining
tool has to cull out individual info from a large database. Having a
separate database for profitable customers will be helpful
Individual/Micro Level

◼ Knowledge obtained by mining at macro level is useful when:

- Firm wants to customise its offering (retail store offering
discounts on the purchase of a bundle of products that the
customer prefers buying together)
- Firm wants to take advantage of personal events in a customer’s
life to further cement the relationship
- Current patterns that go against usually observed customer
behaviour point to an interesting phenomenon (sudden switch in
brand – not satisfied with last purchase)
◼ Link analysis, Regression and deviation detection are the most
common operations used at this level
Why is data mining becoming so
important
◼ More demanding customers
◼ Growing business competition
◼ Large sizes of databases – Terabytes and
Petabytes
◼ Data coming from many channels – Online,
mobile, offline
◼ Quick decision making is need of the time –
Remember : Right offer to the Right Consumer at
Right time through Right Channel
◼ Decision making with maximum knowledge
Approaches to data mining

◼ Directed (supervised, predictive or targeted)

▪ Using input data to predict a specified output

◼ Undirected (unsupervised)
▪ Exploration of data sets to see what can be learned
Application of Data mining

Data Mining is used in almost every industry today

Datamining from a CRM
perspective
Can be used in How can it be used
Sales Tracking Next Best product basis purchase behaviour, demographics,
who to Cross sell to and who to upsell to, Same could be
tracked using sales force automation
Customer Retention Which retention strategy, impact of the same, tracking
reduction in RFM, etc
Customer Loyalty Which customer might attrite, how often do customers buy,
what triggers a buy
Purchase behaviour Similar to sales tracking

Cost Efficiency Which customers to offer what kind of services to, ABC
type of services,
Issue/complaint Resolution, Fraud prevention
Inventory Management, placing of products in retails stores (milk & Diaper, Noodles &
Sauce
Exercise 1:

How can data mining be used from a CRM

perspective in the following industry/Area?

- Films
- Sports
- Public Sector
- Healthcare
 Data Mining Applications in the
Retail Space
Applications Relevance
Database Marketing Develop customer profiles to run focused
and cost effective promotions like –
customers who prefers buying designer
shoes from the retail store
Sales Forecasting Use the time series pattern of customers to
predict the next purchase
Merchandise Planning and Allocation Demographic and psychographic data of
consumers can be used to change the retail
layout and merchandise planning
Clustering Creating customer clusters based on the
attributes ( demography, psychographic,
purchase pattern etc.)
Basket Analysis Which are the product the customers tend
to purchase together – example : Burger
and Cola
Tools/operations used for data
mining
Tools/operations used for data
mining
◼ K means Clustering or Segmentation
◼ Decision Trees
◼ Link Analysis
◼ Rule Induction – MBA
◼ Regression
◼ NNT
Segmentation

◼ The goal of segmentation is to identify groups

of individuals or entities that share similar
needs, preferences, behaviors, or other
relevant characteristics.
◼ Factors that can be used for segmentation
▪ Geography
▪ Product usage
▪ Length of relationship
▪ Revenue
▪ Demographics
▪ Customer value
▪ Attitude
 Segmenting customers based on
profitability
What segment spends more
Most profitable with us over time, costs less to
customers Platinu maintain, spreads positive
m word-of-mouth?

Gol
d

Iro
n
What segment costs us in time,
Lea effort and money yet does not
d provide the return we want?
What segment is difficult to do
Least profitable business with?
customers
 Activity - Offer Optimization matrix: Name the segments and

suggest action
LIKELY TO
CUSTOMER CHURN
PROFITABILI
TY Segment LOW MEDIUM HIGH

HIGH
REVENUE/HI
GH MARGIN
Type of offers

• High Cost HIGH

Incentive REVENUE /
LOW
• High Margin MARGIN
Incentive LOW
REVENUE/
• Low Cost HIGH
Incentive MARGIN
LOW
• Repricing REVENUE/LO
Action required
K means Clustering
◼ K-means clustering is a method of segmentation. It is a
popular technique used within segmentation processes to
divide a dataset into distinct groups or segments based on
similarities among data points.
◼ I n the context of customer segmentation, for example,
K-means clustering can be applied to group customers into
clusters with similar characteristics or behaviors. Each
cluster represents a segment of the customer base, allowing
businesses to tailor their marketing strategies and services
to the specific needs and preferences of each group.
◼ So, while segmentation encompasses various methods and
techniques for dividing data, K-means clustering is one
specific method used within segmentation processes to
achieve this goal.
Segmentation /K Means Clustering
and uses in CRM
K means can be used for Customer
Segmentation – this can further be used
for
◼ Targeted Marketing campaigns
◼ CLV Prediction
◼ Churn Prediction
◼ Product Reco and Personalisation
◼ Channel optimization
◼ MBA
Decision Tree

◼ There are hierarchical collection of rules that

describe how to divide a large collection of
records into successively smaller groups of
records. With each successive division, the
members of the resulting segments become
more and more similar to one another with
respect to the target
How it works?

◼ Decision trees recursively split data into smaller and smaller

cells which are increasingly “pure "in the sense of having
similar values of the target
◼ Decision tree uses target variable to determine how each input
should be partitioned
◼ Breaks the data into segments, defined by splitting rules at
each step.
◼ Taken together, the rules for all the segments form the
decision tree model
Decision Trees & CRM
◼ Can be used for segmentation - Decision trees can help
segment customers based on their characteristics,
behaviors, or preferences. By analyzing customer data such
as demographics, purchasing history, or interactions with the
company, decision trees can identify distinct customer
segments. This segmentation allows businesses to tailor
their marketing strategies, product offerings, and customer
service to meet the specific needs of each segment.
◼ Personalized Recommendations: Decision trees can be
used to provide personalized product or service
recommendations to customers based on their past behavior
and preferences
◼ Customer Service Optimization: Decision trees can aid in
optimizing customer service processes by guiding support
agents through troubleshooting steps or routing customer
inquiries to the appropriate department.
Link Analysis
◼ Link analysis, on the other hand, is a data analysis
technique that focuses on examining relationships or
connections between entities in a network. It is
commonly used in fields such as social network
analysis, fraud detection, and criminal investigation to
uncover patterns of relationships or connections
among individuals or entities.
◼ Information can be gathered by knowing how things are
connected. In an ever connected world, understanding
relationships and connections is crititcal.
◼ It can be used for both directed and undirected data mining
Link Analysis can be used for

◼ Helps in visualizing relationships

◼ Yields good results in
▪ Analyzing link between web pages
▪ Analyzing telephone call patterns to find influential
customers
▪ Understanding transportation planning
▪ Crime & terror links
▪ Social Network Analysis
▪ Physician prescription reference
▪ Market Basket Analysis (What all do people buy
together)
Application – Website Link Analysis

◼ Analyse the incoming links to

your page
◼ Evaluate the quality of link
◼ Analyse the link building
strategy
◼ Improve your page rankings
◼ Used in SEO
Application - Social Network
Analysis (SNA)
◼ Networks or graphs used to represent all kinds
of relationship between people such as kinship,
commerce, illness

▪ Linkedin
▪ Facebook (predicting the home address)
▪ Dating sites
▪ “Six degrees of Separation” (Stanley Milgram,
1967)
▪ Prevent epidemics
Rule Induction - MBA
◼ Link analysis deals with network data to uncover
connections between entities, while market
basket analysis focuses on transactional data to
discover associations between items purchased
together.
◼ Market basket analysis is closely identified with
association analysis.
◼ Market basket analysis is more than the content of the
shopping cart. It is also about the characteristics of the
customers- such as demographics and geography
◼ It is about key drivers of purchase.
Market basket analysis

◼ Each customer purchases different set of products,

quantities, different times
◼ MBA uses this information to
▪ Identify who customers are
▪ Understand why they make certain purchases
▪ Gain insight into the merchandise
▪ Fast & Slow movers
▪ Products which are purchased togethere
▪ Products which benefit from promotion
▪ Take action on store layouts, promotions
◼ Combining all of this with a loyalty card becomes
even more valuable
MBA and CRM

◼ Used for cross sell and upsell

◼ Personalised recommendation
◼ Formation of loyalty programs
◼ Bundling as a promotion
Regression

◼ Predict a value of a given continuous valued variable based on

the values of other variables, assuming a linear or nonlinear
model of dependency.
◼ Examples (not of CRM)
▪ Predicting sales amounts of new product based on advertising
expenditure.
▪ Predicting wind velocities as a function of temperature, humidity, air
pressure, etc.
▪ Time series prediction of stock market indices.
What is regression
◼ Predict a value of a given continuous valued variable based on
the values of other variables, assuming a linear or nonlinear
model of dependency.
◼ Examples (not of CRM)
▪ Predicting sales amounts of new product based on advertising
expenditure.
▪ Predicting wind velocities as a function of temperature, humidity, air
pressure, etc.
▪ Time series prediction of stock market indices.
◼ Estimate the value of the target variable
◼ Describe the relationship between the
variables
Regression & CRM

◼ CLV Prediction
◼ Churn Prediction
◼ Price Optimization
◼ Sales forcasting
NNT - Introduction

◼ Based on the concept of Collaborative filtering or

instance based learning or Similarity

◼ Uses past data instances with known output values to

predict an unknown output value of a new data
instance ( Memory based reasoning )

◼ Measures of distance and similarity are important to

nearest neighbour technique

So, NNT is a Classification and Predictive data

mining technique
Working of NNT

◼ A database of known records is searched to find

pre classified records similar to a new record.
These neighbours are used for classification and
estimation

◼ Applications
▪ Fraud detection
▪ Customer response prediction
▪ Medical treatment
▪ Classifying responses
NNT & CRM

◼ Segmentation Analysis
◼ CLV prediction
◼ Recommendation /Prediction systems
◼ Sentiment Analysis