Selling vs. Marketing

1
Selling vs. Marketing

• Predicting Your Customers Needs Through Data
  Mining

Brian K. Chapman Marketing Database Analyst State
Farm Insurance Companies
2
Marketing Definition
Marketing is an organizational function and a
set of processes for creating, communicating, and
delivering value to customers and for managing
customer relationships in ways that benefit the
organization and its stakeholders. Source
American Marketing Association  
3
Enhanced Marketing at State Farm

• There is marketing.

• There is selling.

Persuading the customer to buy a product or
service
Making the right offer to the right person at the
right time.
4
Data Mining Definition

• Google Hits for Data Mining 49,400,000
• 20 Definitions at First Glance
• Definitions Vary
• The ability of users of a system to integrate a
  database ad hoc
• The analysis of relationships that have not been
  previously discovered
• Searching large volumes of data looking for
  patterns that accurately predict behaviors in
  customers and prospects
• A hot buzzword for a class of database
  applications that look for hidden patterns in a
  group of data
• Using advanced statistical tools to identify
  commercially useful patterns in databases

5
Keys to Successful Data Mining

• Know your industry
• Know your customers
• Know your data
• Know your tools and techniques

6
Data Mining for Marketing at State Farm

• Direct mail
• Right person
• Right time
• Right message
• Lead delivery to agents
• Direct mail to in-book
• Individual proactive personal contact
• Pivoting

7
Data Mining
Data Collection
Understanding the Data
Database Marketing
Modify
Model Creation
Model Implementation
8
Privacy

• Privacy is a concern to most consumers
• Public opinion can influence the legislative
  process
• Keep customer concerns in mind during all aspects
  of data mining.
• Good Neighbor policy Good business policy
• Understand the Legal Environment
• Federal
• State (SB1 California)

9
Privacy

• Do-Not-Share
• Gramm-Leach-Bliley-Act Financial Services
  Modernization Act of 1999
• Health Insurance Portability Act of 2005 (HIPPA)
• Fair and Accurate Credit Transaction Act (FACT
  Act)
• Do-Not-Solicit
• Controlling the Assault of Non-Solicited
  Pornography and Marketing Act of 2003 (CAN-SPAM)
• Telemarketing and the Telephone Consumer
  Protection Act (TCPA)
• Junk Fax Prevention Act of 2005 (Junk Fax Act)

10
Data Mining
Data Collection
11
Data Collection
Marketing Database (MD)

• Developed in 1998
• Built by IBM on Unix platform
• DB2 Relational Database
• Currently contains over 3000 data attributes
• Updated monthly - 24 month history
• Federated Data Warehouse

12
Data Mining
Data Collection
Understanding the Data
13
Understanding the Data
What data does State Farm have in MD?

• Demographics (age, gender, marital status,
  children, etc.)
• Geographics (city, state, zip)
• Policy information (qty, type of policy, when
  purchased)
• Item insured (vehicle, home, classic car,
  paintings)
• U.S. Census (neighborhood)
• Life Events and Change (marriage, birth, recent
  move)

14
Understanding the Data
Information State Farm does not have in IMD

• Name
• Street address
• Phone numbers
• SSN
• Ethnicity
• Psychographics motivations, preferences

15
Understanding the Data
Methods and tools of Data Mining (ad hoc)

• Structured Query Language (SQL)
• SAS
• SAS Enterprise Guide
• SPSS
• Cognos
• MapInfo
• Microsoft Access
• Microsoft Excel

16
Data Mining
Data Collection
Understanding the Data
Model Creation
17
Model Creation
Methods and tools of Predictive Modeling

• SAS Enterprise Miner
• SPSS
• IBMs Intelligent Miner for Data

Techniques of Predictive Modeling

• Logistic Regression
• Decision Trees
• Neural Networks

18
Model Creation

• Predictive models consider complex combinations
  of factors when looking at a group of customers.
• Those factors can include any factor that relates
  to both the customers and the primary market
  area.
• The factors that go into models are statistically
  complex and any particular factors importance
  can change depending on other factors present at
  the same time.
• Model creation is an iterative process, updated
  regularly, and subject to change.

19
Model Creation
Recent Life Events
Demographics
Policy Info
Agents Info
U.S. Census Data
20
Data Mining
Data Collection
Understanding the Data
Model Creation
Model Implementation
21
Model Implementation

• Direct mail
• Right person
• Right time
• Right message

ALWAYS TESTING!

• Lead delivery to agents (LM)
• Direct mail to in-book
• Individual proactive personal contact
• Pivoting

22
Model Implementation
Success stories
My staff used (the model) to target high
propensity renterssending out 100 postcards to
the high propensity customers. Of the 100
prospects, we quoted 12 policies and wrote 6 of
them!... My staff sent out 60 high propensity
Auto postcards last Wednesday and in one week we
wrote THREE autos! This is the best return on
investment! I became an instant believer in
(modeling) when Darwyn contacted our high
propensity Life prospects in order to set up IFR
appointments. As a result, we reached our Life
sales goal by October 31!
23
Data Mining
Data Collection
Understanding the Data
Modify
Model Creation
Model Implementation
24
Modify

• Model creation is an iterative process, updated
  regularly, and subject to change.

• Changing environment
• Changing customer base
• Changing data availability

25
Data Mining
Data Collection
Understanding the Data
Database Marketing
Modify
Model Creation
Model Implementation