Customer Relationship Management
Analytics
Week 13: Summary
2
What is Customer Relationship
Management?
• CRM is a strategy which utilises a combination of
– information – technology – policies – business processes, – employees
to develop profitable customers
3
Customer Life Cycle
Responder Target Market Responder High Value High Potential L o w V a l u e Voluntary C hurn Forced C hurn Product Usage Payment History C a m p a i g n R e s p o n s e s C a m p a i g n H i s t o r i e s Purchased Demographics Termination Reason (Berry& Linoff ,2000) 4
Objectives of CRM
• to acquire new customers • to retain good customers
• to develop long-term higher-spending customers from casual customers
• to drop bad customers
Motivation for CRM
• CRM is not new; it has always been practiced – but implicitly
• New business realities:
– globalisation, hyper-competition, technological change, customer expectations
– many more customers and products
– new product developments require timely knowledge of consumer requirements
– identifying what customers want and when has become a challenge
Three Types of CRM
• Operational: automation of customer-related business processes such as marketing, sales, and customer service
• Collaborative: provides channels for communication with customers (email, video conferencing, internet)
• Analytical: provides understanding of customer behaviour through in-depth analysis of customer data
7
Closed-Loop CRM
Using knowledge obtained from analytical CRM, collaborative and operational CRM can provide better customer service and more efficient processing
ANALYTICAL
Operational
Operational CollaborativeCollaborative
Customer knowledge Customer data Customer knowledge 8
Analytical CRM
Traditionally, companies focused on operational and collaborative CRM
• Problems: separate systems and databases that do not share information
• Analytical CRM consolidates customer data across the enterprise and provides unified view of the customer
• Objective is to increase customer profitability through better understanding of customers.
9
yesterday
functions dominated > data fragmentation > processes ineffective
Traditional Enterprise
functions
owned
databusiness process
data data data
business activity business activity business activity business
activity businessactivity
business
function functionbusiness businessfunction
business process
10
today
information integration > process improvement > business optimization
Intelligent Enterprise
•Operation al Excellence AND Customer Intimacy •Cross-functional optimization •Reduced IT infrastructure costcore processes
supported directly by datacorporate data warehouse
business
activity businessactivity businessactivity business
activity businessactivity
data data data
DATA OPTIMIZATION
business
function functionbusiness businessfunction
business process
business process
Analytical CRM process
Data Collection à Integration à Analysis à Action • Collection: Collect customer-related data from
– customer touch-points – operational systems – application forms, reports
– external data sources (credit agencies, marketing agencies).
Technologies for Analytical CRM
Data warehousing, OLAP and Data Mining provide an
• integrated • reliable • historical • analytical
13 customer contact points Retrospective analysis tools •OLAP •Query •Reporting Prediction and discovery •Data Mining Customer Data Warehouse
Architecture for Analytical CRM
Operational systems and external info
14
Analytical CRM process
• Integration:
– consolidate and integrate collected data into a central repository
• Analysis:
– analysis of data in the repository including customer profiling, customer profitability, customer retention, customer segmentation etc.
• Action:
– develop business processes and organisational structures to leverage the derived knowledge
15
Data warehouse
• “… a subject-oriented, integrated, time-variant, and non-volatile collection of data used in support of management’s decisions”
Inmon and Hackathorn (1994)
16
Analytical CRM
• DW integrates customer data collected from across all customer touch-points and systems • with the emphasis on data quality in a data
warehouse, DW can provide a reliable view of customers
• historical data in DW facilitates analysis of customer behaviour over time
• DW facilitates analytical information processing
CRM data warehouse
• customer dimension is critical for effective CRM • It is most challenging dimension for any data
warehouse
• Can have millions of rows(credit card companies, government agencies exceed 100000 millions of rows)
• Hundreds of attributes
• One leading marketer maintains 3000 customer attributes
• (Kimball 2002, The data warehouse toolkit,chapter 6)
Customer Data Warehouse design issues
• Data modelling issues • Capturing changes
• Consolidation and integration issues • Data quality
• Storage of unstructured information • Data storage requirements • Privacy issues
19
Data modelling for CRM data
warehouse
• Typically data for DW is modelled using dimensional modelling (Kimball) • Does it work for CRM?
20
Modelling data for CRM DW
• Dimensional model is Behaviour
centric (customer sales, bookings,
orders etc…)
• CRM data warehouse is customer-centric • In CRM we want to examine the effect of
change in customer circumstances
(Todman, 2001)
21
Customer-centric approach
• The emphasis is shifted from behaviour • More value attached to the customer’s
personal circumstances
• Todman proposes General Conceptual Model for customer centric warehouse
22
General conceptual model for a
customer-centric data warehouse
Customer Behavior Customer Customer Derived Segment Customer Changing Circumstances
(Adapted from Todman,2001)
Analysis of customer data
Customer analytics
• On-line analytical processing (OLAP),
query and reporting
• Data mining
Analysis of customer data, OLAP
• OLAP offers a set of graphical tools that provide multidimensional viewsof data and allow users to visualise, summarise, and analyse data
• An objective of OLAP technology is to provide users with the opportunity to perform complex analysis of data in an intuitive and simple way
25
Analysis of Customer Data
• OLAP, reporting, query
500 2000 1250 2 6 0 920 6 7 0 1 0 0 400 750 V o l u m e C o s t Revenue Month Product Measure Measures
Volume Cost Revenue
J a n u a r y 100 260 500
F e b r u a r y 400 920 2000
March 750 670 1250
Product 1
26
Analysis of Customer Data
Data Mining
• Data mining, or knowledge discovery, is the process of discovering valid, novel and useful patterns in large data sets
27
Predictive versus descriptive data
mining
• Predictive data miming
– discovering patterns for the purpose of predicting future trends and behaviours (sales forecasting, bankruptcy prediction, propensity models)
• Descriptive(exploratory ) data mining
– identifying understandable patterns for the purpose of obtaining an insight into the business domain. – example: association rules for market basket analysis
28
Knowledge discovery process
Target Data
Transformed Data
Patterns
(adapted from Fayyad(1996))
Data Selection Pre-processing Transformation Data Mining Interpretation/ Evaluation Pre-processed Data KNOWLEDGE and models Business
problem Many iterations may be required!
Data mining tools and techniques
• Neural networks
• Decision trees (C4.5, CART, CHAID) • Clustering
• Visualisation(aids understanding data and results)
• Rule induction • Text mining • Statistics
Neural Networks
• Neural networks are computing systems inspired by the structure and functioning of the brain. • Neural networks learn patterns in data through
the process of training on examples • Very good prediction/classification tools • Problems: If NN are used for decision making it is impossible to
31
Neural Networks
AGE Good account Bad account BALANCE Card holder data INCOMEInput units Processing units Output units
32
Decision Trees
• Tree-shaped structures • Used for prediction, classifi
cation, exploration • Can be converted to rules
IF State= Vic
THEN response = Low; IF State= NSW and Occupation = Sales
THEN Response = High; ………..
• Problems: tree can become very large - hard to understand VICT NSW QLD LOW sales transport sales teaching HIGH AVERAGE LOW AVERAGE
State Occupation Response
33
Clustering (unsupervised learning)
• Clustering algorithms group records with similar characteristics
• Clusters have to be analysed and classified •Income: very high
•Sex:M •Status: single •Age: 25-35 •Car: sport •Income: medium •Sex: F •Status: single •Age: 25-35 •Car: compact •Income: medium •Sex: F,M •Status: married •Age: 35-45 •Car: family •Income: high •Sex: M •Age: 35-45 •Status: married •Car: luxury 34
Typical data mining applications
for CRM
• targeted marketing • market basket analysis • trend analysis
• market segmentation • Credit risk analysis • customer retention
Examples of Data mining applications
for CRM
• Targeted marketing, marketing campaigns
– Data mining models (using, for example, neural nets or decision trees) can be built from historical data to predict:
• which customer segments are more likely to respond to a particular campaign
– Benefits: increased response rate, savings on marketing campaigns
Churn modelling
• Objective is to reduce attrition rate of valuable customers
• Data mining can be used to predict which customers are likely to leave (propensity to churn)
37 Market basket analysis
• Automated extraction of association rules from point-of-sales transaction data • Example rules
– IF bread then milk and bananas – IF soy milk then soy cheese
– Problems: too many rules, difficult to identify which one is significant, interesting
38
Integrating data mining in
business processes
Data mining models have to be evaluated and integrated in business processes • campaign management • Win-back programs • loan management • Call centre • etc.. 39
Integrating data mining into business processes example: Campaign Management
• Marketing dept. identifies a segment of interest • The records of customers from the segment are
selected from the data warehouse to a separate table
• The customers are then scored using a predictive model
• The score can, for example, represent the probability of responding to a particular promotional offer
40
Integrating data mining into
campaign management
• the customer records are then sorted by their score value
• The top 25% are selected to receive the offer
• The offer is then mailed out (by a mail house) to the selected customers
Trends in CRM
• Event Based Marketing
Customer Event
Significance Evaluation
Context & Lifestage
Service Need Proposition
Segment
Propensity
A Paradigm Shift from Traditional Target Marketing to Event Based Marketing
Offer
43
Types of Events
Life-stage Events Birthday Anniversary Marriage Product Events Maturity New Account Condition changes Behavioural EventsA significant change in a customers transaction patterns or behaviour that indicates a customer need
Significant Deposit, Withdrawal, Payment Changes,
Change in Purchase Patterns, Limit Utilisation, Change in Channel Usage, Increase/ Decrease in transactions (Type, Volume and/or Value)
44
Event Based Marketing has a Proven Record in Servicing Customer Needs
Traditional Target Marketing • Select Customers who fit a
certain criteria • Typically executed once
(single shot) or quarterly/ bi-annually, annually
Customer Needs Marketing • Identify Customers who fit a
certain criteria AND • ONLY Contact those
customers whose transaction behaviour suggests they have a need for additional credit
Tehan, D., Teradata,2004)
45
Effective CRM
• Develop closed-loop CRM processes to share knowledge obtained as a result of customer data analysis with all relevant business units
• Develop strategy and processes for leveraging this knowledge.
46
Closed-Loop CRM
Using knowledge obtained from analytical CRM, collaborative and operational CRM can provide better customer service and more efficient processing
ANALYTICAL
Operational
Operational CollaborativeCollaborative
Customer knowledge
Customer data
Customer knowledge
Why many CRM’s have failed?
• Too much emphasis on technology • Lack of strategy for developing new and
changing old business processes
• Failed to integrate the three types of CRM
“Doing CRM Right” - IBM study
(source: www.crm2day.com)
• Why many CRM projects have failed ?
– Fewer than 15% of global companies believe they are fully succeeding , 20-30 are having some success – Key reasons:
• Too much reliance on technology
• Underestimating the importance of senior management buy-in
49
Study Findings
• 75% of companies manage CRM at the functional level (marketing, sales etc..)
– CRM should be run at the corporate level or with a cross functional perspective
• Senior management in over 35% of companies regards CRM as useful but not critical
– it sends wrong message to employees that CRM is not a priority
• Over 75% of companies do not fully use CRM once it is Implemented
– Only 14% of employees fully use CRM
50
Study recommendations : Successful
CRM
– Right culture
– Creating broad acceptance and adoption Important:
• Viewing CRM as critical by senior management
• Employees understanding of CRM transformation and linking it to organisational objectives
• Aligning CRM objectives with the priorities of employees
“CRM initiatives transform a company culturally, structurally and strategically”
CRM Today (2004) “IBM Survey Shows Companies Gaining Strategic Value from CRM initiatives”, ww.crm2day.com, 29 April 2004
51
References
Berry, M. and Linoff, G. (2000 ) Data Mining Techniques for Marketing, Sales, and Customer Support, John Wiley & Sons
Berson A.., Smith S., Thearling K (2000), Building Data Mining Applications for CRM,McGrow-Hill
Fayyad, U., Piatetsky –Shapiro, G., and Smyth, P. (1996) (eds.) Knowledge Discovery and Data Mining: Towards a Unifying Framework , Proc. of the Second International Conference on Knowledge Discovery & Data Mining (KDD-96), AAAI Press, Menlo Park, California, 82-88
Jagielska I, Jaworski J.,(1996) Neural Network for Predicting the Performance of Credit Card Accounts, Computational Economics Vol 9, No. 1 February 1996, 77 -82
CRM Today (2004) “IBM Survey Shows Companies Gaining Strategic Value from CRM initiatives”, ww.crm2day.com, 29 April 2004