Transcript Bank Segmentation for Marketing

Bank Segmentation for
Marketing
Randy Johnson & Roland Shabani , Western Kentucky University
BI 420 Data Mining
Question
• We’re looking to find customers’ usage patterns of banking methods
over the course of a three month study period. Using 100,000 case
examples. We’re wanting to find out which banking method is being
used and which ones aren’t. Using that information we want to make
the customers’ banking experience easier and more convenient.
Background
• 100,000 active customers were study.
• The dataset has six attributes. Which one, ID, being a special attribute used to identify the
customers.
Attribute Name
Model Role
Measurement Level
Description
Examples
ID
ID
Nominal
Customer ID
CNT_TBM
Input
Interval
Traditional bank
In-bank services;
method transaction deposit and
count
withdraws. Building
a relationship
CNT_ATM
Input
Interval
ATM transaction
count
ATMs
CNT_POS
Input
Interval
Point-of-sale
transaction count
MoneyGram,
Western Union
CNT_CSC
Input
Interval
Customer service
transaction count
Tellers, personal
bankers, financial
advisors
CNT_TOT
Input
Interval
Total transaction
count
Data Mining Approach
SAS Segmentation
Approach
• Using RapidMiner and SAS, we
ran different filters to attempt to
find any patterns or
commonalities within the
dataset.
RapidMiner Segmentation
Model
Results
• Cluster 0; traditional method the
most similarities.
• Clust 1-4 follower in order from
greatest to least.
• Most banking methods used
1. Traditional
2. ATM
3. Point-of-sale
4. Customer service
Conclusions
Recommendations for Segments
• Segment 1: Old-Fashioned; these customers were more likely use traditional making methods. To
offer more convenient banking, we propose telling them about online banking or offering them
debit cards.
• Segment 2: Transitionals; these customers also used traditional banking more, but they were
closer to the distribution centers on the other transaction methods. This seems to be the more
balanced segment. We wouldn’t offer them as much alternative methods because they are on
both side of the spectrum.
• Segment 3: ATMs; These were customers that used ATMs the most. We suggest informing them
about more ATM locations in the area, so they’ll have more access to them at their convenience.
• Segment 4: Cashless; these group of customers did the least amount of traditional banking. We
propose offering them one-on-one consulting section to get them into the bank and finding out
what kind of investments they might want to making in the near future.
• Segment 5: Service; this group had higher than average rate of customer service contract and
point-of-sale. We propose offering them the banks traditional methods they might have not
heard of.
Dataset Recommendations
• We felt that the dataset was insufficient in some ways. It didn’t offer
enough information about the customers.
• All the dataset told us was the amount of the bank methods each
customer used. Which doesn’t really give us much insight in what
they want and how the bank can better it.
• May if we knew information like their age, bank transaction history.
We might be able to group customers more accurately and find more
patterns within the dataset.
• We suggest that the bank survey their customers; in the bank or even
online to get a clear depiction of how their customers feel about the
bank and what they can improve on.
Sources
• Acknowledgements: Dr. Leyla Zhuhadar for finding the dataset and helping
us analyze it thoroughly.
• Data Source: Profile.csv
• Author Contact Info:
Randy Johnson, (270)421-3849, [email protected]
Roland Shabani, (270)320-1108, [email protected]