Post on 14-Apr-2017
transcript
Data Mining Report for Blackwell’s E-commerce Team
RESULTS AND RECOMMENDATIONS
Our Mission
With this project we were tasked with using data mining methods to investigate customer buying patterns and assist in making cross-selling recommendations.
Our Data
Data from Blackwell
Age
Amount of Items bought
Amount spent
Region
Purchase made In-Store or Online
Data Processing Highlights Age separated into four target groups
Older Baby Boomers (68-85)
Younger Baby Boomers (52-67)
Generation X (35-51)
Millennials (18-34)
Region separated into East, West, South, and Central
Data Mining Objectives
Investigate the relationship between region of purchase and amount spent per transaction
Investigate the factors that predict the amount spent per transaction Understand the correlation between the age of a customer and
region Make cross-selling recommendations based off products already
purchased Decide if we can use the region of purchase to predict the customers
age
Relationship Between Region and Amount Spent
23%
8%
29%
40%
Sales Per Region
East Region
West Region
South Region
Central Region
What we learned: Central region spends most at $1284.05 West Region spends the least at $252.11
Recommendation: Compare and contrast results to
understand why Central Region spends so much more than other regions
Use marketing techniques and strategies in all regions to emulate the Central Region
Amount Customer Will Spend by Region
What we learned: In-store spending as a whole is much
higher than online spending Central Region online spends almost twice
the amount as second highest region
Recommendation: Focus on mirroring cross-selling
recommendations from in-store to online Consider adding brick and mortar stores to
the West Region to penetrate market East R
egion
Onli
ne
West R
egion O
nline
South R
egion O
nline
Centra
l Reg
ion O
nline
In-store
$-
$200.00 $400.00 $600.00 $800.00
$1,000.00 $1,200.00 $1,400.00 $1,600.00
$526.00
$253.00
$521.00
$1,023.00
$1,542.00
Regional Consumer Transaction Forecast
Predict Age Based on Shopping Method
18-34 35-51 52-67 68-850
5000
10000
15000
20000
25000
30000
35000
1283516305
82992561
9619
13418
10820
6143
In-Store Online
What we learned: 90% Success 10% Risk Key demographics (18-34,35-51) shop
over 25% more in-store
Recommendation: Offer cross-promotions for in-store
shoppers to make purchases online Focus marketing strategies towards key
demographics
Promoting Cross-Selling Recommendations
Report
Successfully recommended additional products to customers based off data from a limited product list
Conducted an attribute evaluation and included a merit metric to rank stronger correlations
Included items with direct correlation and separated irrelevant items with lower scores
Recommended other products for Blackwell Electronics to consider adding in their inventory
Report included rationale for each product given to help understand cross-selling decisions
Recommendations for Alienware AAR4-10000BK
Predict the age of a customer in region
What we learned: 40% confidence rate which is too low to consider
successful The Central Region has virtually no 18-34 and 35-
51 year olds
Recommendation: Offer more specific data sets, perhaps including
income, date, or items purchased, to help reduce error in predictions
Establish new marketing techniques in Central and South region to focus on 18-34 and 35-51 year olds18-34 35-51 52-67 68-85
0
5000
10000
15000
20000
25000
30000
4913 63633573
1151
2314
58445699
6143
4849
6479
5262
1410
10378
11037
4585
0
Age Breakdown per Region
East Region West Region South Region Central Region
Objectives Completed
Lessons learned:• Regional Customer Spending Trends
• Marketing Analysis and Direction
• Inventory Expansion
Results:Higher customer satisfaction rate and profitability for Blackwell Electronics
Successfully investigated the relationship between region of purchase and amount spent
Successfully able to predict how much a customer will spend in each region
Established age of customer based on shopping method
Successfully recommended particular products to launch
Consider this, with the proper data, here are some more questions data mining can answer: CAN WE PREDICT CUSTOMER PATTERNS AND THE DIRECTION THEY WILL
TAKE? HOW SHOULD WE DEFINE MARKETING GROUPS? HOW DO WE SEPARATE PROFITABLE CUSTOMERS FROM UNPROFITABLE
CUSTOMERS? COULD WE USE BUYING PATTERNS AS AN EARLY DETECTION SYSTEM
FOR FRAUD? CAN WE PREDICT NEW SHOPPING TRENDS BEFORE THEY HAPPEN?