WHAT IS DATA MINING?

1
WHAT IS DATA MINING?

• Data mining is the process of exploration and
  analysis, by automatic or semiautomatic means, of
  large quantities of data in order to discover
  meaningful patterns and rules.
• Need for analysis predictive modeling
• Massive data sets
• e-Business motivation

2
DATA MINING TASKS

• Classification, e.g. low, med, or hi risk credit
  card applicant
• Estimation, e.g. probability of paying off a home
  equity loan gt convert to 0/1
• Prediction, e.g. predicting churn, who will
  respond,

3
(data mining tasks cont.

• Affinity grouping or association rules,
  e.g.cross-selling opportunities, product
  placement, etc.
• Clustering, e.g. grouping like customers with no
  pre-set categories
• Description Visualization, e.g.intuitive
  graphical display of data for meaningful
  interpretation.

4
DATA MINING FOR MARKETING CRM

• Reduced costs, e.g. promotions targeted to
  relevant customers.
• Increased revenue, e.g. cross-sell and spot most
  profitable customers, up-sell
• 0ne-to-one marketing ability to implement
• Proactively anticipating customer needs and
  addressing them

5
DATA MINING DECISION SUPPORT

• Data mining is one DM tool.
• More than retrieval, intelligent queries
• Database requirements
• Transaction based vs. historical
• Decision Support Fusion and the VP of Marketing
  YOU!!!

6
SOCIETAL ISSUES

• Privacy
• Data Ownership
• Appropriate Use
• Implications of getting too good.

7
WHO ARE YOUR BEST CUSTOMERS????

• The Concept is very simple (RFM)
• Your best customers are those who
• Have brought from you most recently
• Buy from you most frequently
• Spend a lot on your products/services

8
Recency

• Most recent date customer has made a purchase
  from you.
• Most studies indicate that those consumers that
  have purchased from you most recently are more
  likely to respond positively to a new
  offer/promotion.

9
Frequency

• How many times they have purchased from you since
  their initial purchase or some set date.
• Customers who buy from you many times are more
  likely to respond positively to a new
  offer/promotion.
• Warning those that have just recently purchase
  may not have a lot of frequency but represent
  long term potential.

10
Monetary

• Total dollar value of customer purchases since
  they first started buying from your company.
• Amount of total purchases
• Amount of purchases in last 12 months.
• Amount of average purchase.

11
Creating RFM Codes

• Excellent scoring/segmentation system
• Convert raw values to categorical codes, e.g.
  divide recency days into five equal groups (code
  1-5).
• Simplifies profiling process and processing
  speed.
• Some loss of information.

12
Using RFM Cells to Predict Response

• Historical Data Overlay
• Test Group, e.g. 40000 out of 1 million
  customers, sent promotion/communication.
• Gauge response by RFM Group
• Demographic profiles of high response group for
  projections.