NCDM Datamining Case Study 2010
-
Upload
jim-stafford -
Category
Documents
-
view
375 -
download
1
description
Transcript of NCDM Datamining Case Study 2010
NBA TEAM SCORES WITH DATA MINING:
A CASE STUDY IN MODELING AND PROFILING
Presented by:
James R. Stafford
What is modeling & profiling?
Who uses modeling and profiling?
Common approaches
7 steps to success
Case study - NBA team upsell study
Game Plan
Modeling & Profiling
Who will respond?
Identify cross-sell opportunities
Who is likely to lapse/churn?
What do my best customers look like and how can I get more?
Who should receive what message?
Increase revenues, profit, and maximize ROI on marketing $
What is Predictive Modeling?
Predicting outcomes and future events based on historical data relating to:
- past response- transactions/purchase history- geo-demographic- lifestyle, and other attributes
What is Customer Profiling?
Profiling is a data discovery procedure that uses standard queries and statistical analysis
to segment customers and prospects based on important
traits like R,F,M, transaction/purchase behavior,
and demographics.
Who uses PredictiveModeling?
Response Cross-Sell Lapse/Churn Reactivate Lifetime Value Most Profitable
Banks & Financial Services
Publications Retail Catalogers Telco’s High-Tech Hospitality & Gaming
The Industry The Problem
Which approach should be used?
RFM CHAID Linear regression Logistic regression Neural nets
Linear regression CHAID Neural nets
limited number of answers
If the business problem has a...wide range of answers
7 steps to successful modeling and implementation
Identify the business problem Data audit -- what’s available and
relevant? Create training and validation files Use best modeling approach and
appraise results Does the model make sense? Validate the model Test campaign
CASE STUDY IN MODELING
& PROFILING
The Business Problem
National Basketball Association Team Declining attendance Expanding to new stadium with more seats
Marketing Objectives Up-sell: Mini-plan to Season ticket holders Prospecting: identify Season ticket plan prospects
Applicability to you...
Retention and up-sell -- NBA franchise has products/services and desires repeat buyers
Desire to differentiate customers with different purchasing behavior
Desire to acquire new & profitable customers
Create marketing efficiency & cut promotion costs
Data audit - customer data
Street address
# of Seats
7 game mini-plans & 14 game combos
7A = “World’s Best” -- Dream Team players
7B = “Weekend Fest” -- Fri., Sat., Sun. games
7C = “Wild West” -- Western conference teams & Chicago Bulls
21 game mini-plans
Full season ticket holders
Data preprocessing & overlay
Correct and standardize addresses
Geo-code addresses to census neighborhoods
Append updated area-level demographics
Append PRIZM lifestyle cluster types
Create training and validation files
Training file - 1884 records (75% of file)
Validation file - 651 records (25% of file)
Must always use random sampling!
Use best modeling approach
CHAID Linear regression Logistic regression Neural net
Use best modeling approach
Appraise results – Gains chart for our best model
Let’s just mail to the 50% most
likely to respond, and we’ll get 70%
of the likely responders
_______Highly targeted
and saves money
Appraise results - Gains chart for our best logistic regression model
Appraise results - Gains chart for our best linear regression model
Does the model validate?Does the model validate?
Training Data Validation Data
Does the model make sense -- what do my customers look like?
Does the model make sense -- what do my customers look like?
Does the model make sense -- what do my customers look like?
Does the model make sense -- what do my customers look like?
Does the model make sense -- what do my customers look like?
Does the model make sense -- what do my customers look like?
PRIZM Cluster Groups
T1: Landed Gentry C1: 2ndCitySociety
S1: EliteSuburbs
U1: UrbanUptown
R1:CountryFamilies
R2:Heart-landers
R3:RusticLiving
T2:ExurbanBlues
T3:WorkingTowns
C2: 2ndCityCenters
C3: 2nd City Blues
S2: TheAffluentials
S3: InnerSuburbs
U2: UrbanUpscale
U3: UrbanCores
So
cio
eco
nom
ic S
tatu
s
Urbanization
PRIZM cluster composition for segments
Modeled C1 C2 S1 S2 S3 U1 U3Segment 1 1.6 2.4 31.2 4.0 12.8 12.0 34.4 2 2.4 16.3 56.5 1.6 0.8 13.7 4.0
10 5.5 28.4 11.0 5.5 18.1 1.6 4.7
19 2.8 2.8 10.1 32.1 4.6 2.8 0.020 5.5 0.9 18.4 22.9 2.8 0.0 0.0
TOTAL 6.0 9.5 24.0 14.9 11.0 6.1 4.9
EliteSuburbs
UrbanCores
Top demi-decile, i.e., those most likely to become
season ticket holders
Education
0
50
100
150
200
250
4 + Years of College
1-3 Years College HS Graduate < 12 Years
S1
U3
Household income
0
50
100
150
200
250
300
350
< $15,000 $15,000 - $34,999 $35,000 - $74,000 >= $75,000
S1
U3
Occupation
0
20
40
60
80
100
120
140
160
180
Professional/Mgr Other W/Collar Blue collar Service Farm/Ranch/Mine
S1
U3
Household size
0
20
40
60
80
100
120
140
160
1 Person 4 + Persons HH w/Child
S1
U3
Summary profile of “the best” segments
Wealthy whites, Asians and Arabic
High spending levels Highest income High education High investment
Multi-racial Multi-lingual Dense/urban Home & apartment renters High % of singles High % of single parents High unemployment Lowest income group
U3 - Urban CoresS1 - Elite Suburbs
Mostlikelytoo...
S1 U3Lifestyle
Country club Football games Contrib $50+ to PBS 15 + Lottery ticket/mo Sail Play pool Housekeeper Pro basketball Classical music Smoke Contract home improve Fast Mexican food
Mostlikelytoo...
Media S1 U3 News/talk radio Rock radio Murphy Brown The Simpsons Masterpiece Theater Rescue 911 Jazz radio Jazz radio Masters Golf Tourn Listen/football Travel & Leisure Car & Driver Fortune Ebony National Geographic Consumer's Digest Business section Classified section
Mostlikelytoo...
Buy S1 U3Cappuccino Malt liquorImported wine Domestic beerPita bread Fruit LoopsMontblanc pen Adidas shoesComputer Pepsi Saks K-Mart
How Can You Use This Information ?
Develop different messages Use different media/marketing approaches to
reach them Buy prospect lists based on best segment
profiles Develop retention and prospecting plans with
customized offers (e.g., free CD’s based on their particular tastes in music)
For each major customer segment, you can...
===>> improved customer up-sell and retention and better prospecting!
Potential marketing plans
S1 U3Giveaways
1,000 FF miles Mini-music systemCD - Classical/Jazz CD - Jazz/RockFree WSJ sub Free Consumer Report sub
Contests1 trip to the Master's, or… 1 trip to Super Bowl, or… the NBA finals the NBA finals50 Montblanc pens 50 pairs Adidas/Nike
AdvertiseJazz stations Jazz stationsClassical stations Rock stationsLocal Business Section Local Classified section
S1 U3Giveaways
1,000 FF miles Mini-music systemCD - Classical/Jazz CD - Jazz/RockFree WSJ sub Free Consumer Report sub
Contests1 trip to the Master's, or… 1 trip to Super Bowl, or… the NBA finals the NBA finals50 Montblanc pens 50 pairs Adidas/Nike
AdvertiseJazz stations Jazz stationsClassical stations Rock stationsLocal Business Section Local Classified section
Potential marketing plans
Potential marketing plans
S1 U3Giveaways
1,000 FF miles Mini-music systemCD - Classical/Jazz CD - Jazz/RockFree WSJ sub Free Consumer Report sub
Contests1 trip to the Master's, or… 1 trip to Super Bowl, or… the NBA finals the NBA finals50 Montblanc pens 50 pairs Adidas/Nike
AdvertiseJazz stations Jazz stationsClassical stations Rock stationsLocal Business Section Local Classified section
AdvertiseJazz stations Jazz stationsClassical stations Rock stationsLocal Business sections Local Classified section
Summary - why model & profile?
To identify those customers most likely to behave in certain ways (respond, cancel, etc.)
To see what those customers are like (high income, infrequent purchasers, etc.)
To identify what motivates our customers (price, frequency of contact, etc.)
To create mass personalizations
Expected results
Increased ROI on marketing dollars - e.g., only mail to those most likely to respond
Increased customer loyalty Decreased attrition rates Higher actual lifetime value
Maximize each customer relationship
NBA TEAM SCORES WITH DATA MINING:
A CASE STUDY IN MODELING AND PROFILING
Presented by:
Jim Stafford