Transcript Chapter 16 - Data Miners

Chapter 16
Building the Data Mining
Environment
The Ideal Customer-Centric Organization
• Customer is king (not pauper)
• For B2C (business to consumer) - Combination of point-of-sale
transaction data and loyalty cards
• For B2B (business to business) – traditional approaches (purchase
orders, sales orders, etc.), Electronic Data Interchange (EDI) of
same, Enterprise Resource Planning (ERP) software with intranet
access for business partners
• Customer interactions are recorded, remembered, utilized (action)
• Corporate culture focused on rewards for how customers are treated
2
The Ideal Data Mining Environment
• A corporate culture that appreciates the value of
information
• Committed (human and $ capital investment) to
consolidate customer data from disparate data sources
(ECTL – extract, clean, transform, load) which is
challenging and time consuming
• A corporate culture committed to being a Learning
Organization which values progress and steady
improvement
3
The Ideal Data Mining Environment
• Recognize the importance of data analysis and
its results are shared across the organization
– Marketing
– Sales
– Operational system designers (IT or vendor software)
• Willing to make data readily available for
analysis even if it means some re-design of
software
4
Reality (where “rubber meets the road”)
• The ideal environments, organizations,
and corporate culture rarely exist all in one
organization!!!
• Don’t be shocked…it’s hard work!!!
5
Building a Customer-Centric Organization
• Biggest challenge to this is establishing a
single view of the customer shared across
the entire enterprise
• Reverse of this is also a challenge –
creating a single view of our own company
to the customer
• Consistency is needed for both the above
6
Building a Customer-Centric Organization
Corp. Culture
Data Mining Environment
Mining Customer data
Collecting the Right data
Single Customer View
Customer Metrics
7
Single Customer View
• Customer Profitability
Model
• Payment Default Risk
Model
• Loyalty Model
• Shared Definitions:
Figure 16.1 A customer-centric organization
requires centralized customer data
–
–
–
–
Customer start
New customer
Loyal customer
Valuable customer
8
Defining Customer-Centric Metrics
• Business metrics guide managers in their
decision-making
• Selecting the right metrics is crucial because a
business tends to become what it is measured by
– New customers – tend to sign up new ones without
regard to quality, tenure, profitability
– Market share – tend to increase this at the expense of
profitability
• Easy to say customer loyalty is a goal…harder to
measure the success of this
9
Collecting the Right Data
• Data collection should map back to defined
customer metrics
• Customer metrics often stated as questions in
need of answers:
– How many times/year does customer contact our
Customer Support (phone, web, etc.)?
– What is payment status of customers (current, 30, 60,
90 days, etc.)?
– Thousands of other questions
10
DM Environment & Mining Data
• Data Mining group (team) is needed
• DM Infrastructure to support is needed
11
Data Mining Group
• Possible locations for such a group include
– Part of I.T.
– Outside organization – outsource this activity
– Part of marketing, finance, customer relationship
management
– Interdisciplinary group across functional departments
(e.g., marketing, finance, IT, etc.)
• Each of the above have advantages and
disadvantages
12
Data Mining Staff Characteristics
• Database skills (SQL)
• Data ECTL (extraction, cleaning, transformation, loading) skills
• Hands-on with Data Mining software such as
PolyAnalyst, SAS, SPSS, Salford Systems, Clementine,
etc.)
• Statistics
• Machine learning skills
• Industry knowledge
• Data visualization skills
• Interviewing and requirements gathering skills
• Presentation, writing, and communication skills
13
Data Mining Infrastructure
1.
Ability to access data from many sources & consolidate
2.
Ability to score customers based on existing models
3.
Ability to manage lots of models over time
4.
Ability to manage lots of model scores over time
5.
Ability to track model score changes over time
6.
Ability to reconstruct a customer “signature” on demand
7.
Ability to publish scores, rules, and other data mining
results
14
The Mining Platform (example)
• Lots of
architecture
strategies –
this is just
one that
includes
OLAP also
15
Data Mining Software
Review “Questions to Ask” Side Bar
in book on page 533 (2nd edition)
16
End of Chapter 16
17