Transcript WHAT IS DATA MINING DM#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
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 => convert to 0/1
Prediction, e.g. predicting churn, who
will respond,
(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.
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
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!!!
SOCIETAL ISSUES
Privacy
Data Ownership
Appropriate Use
Implications of getting too good.
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
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.
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.
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.
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.
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.