Recommendations to IDIDI Bank
June, 30 2011
Ashish Merchant
Summary of data accumulated
BACKGROUND
2
© Jigsaw Academy Education Pvt. Ltd.
IDIDI Bank wants to improve the profitability of its consumer loan
department. The bank has seen a high default rate on its loans between 2
to 5 years of tenure and wants to tighten its credit policy. You have pulled
the data for all bank loans in the last 2 to 5 years. In order to analyze the Decision Tree
data you have created the following decision tree.
Population: 10000
Present your recommendations to the IDIDI consumer loan VP based on
Good: 9500
your interpretation of the analysis. Bad: 500
Data Description: Yes No
Income>1.2
•Good – non-defaulters *
Population: 9800 installment Population: 200
•Bad – defaulters
•Income>1.2*installment – Was the applicant’s Good: 9450 Good: 50
Bad: 350 Bad: 150
monthly income > 1.2 times the amount of monthly
<2 # of other >=2
loan installment (Yes/No)
loans?
•# of other loans – Does the applicant have other
Population: 8800 Population: 1000
loans (<2,>=2)
•Income>2.5*installment – Was the applicant’s Good: 8550 Good: 900
Bad: 250 Bad: 100
monthly income > 2.5 times the amount of monthly
loan installment (Yes/No)
Yes Income<2.5 No Yes Default on a No
•Default on a prior loan – Has the applicant defaulted * prior loan?
on any other loan on record? (Yes/No) installment
Population: 400 Population: 8400 Population: 150 Population: 850
Good: 200 Good: 8350 Good: 70 Good: 830
Bad: 200 Bad: 50 Bad: 80 Bad: 20
3
© Jigsaw Academy Education Pvt. Ltd.
Executive Summary of Investigation
Legend: Parent Node: Node Total Good Bad % Good % Bad
Segments at Risk Segments Not at Risk 10,000 9,500 500 95.00% 5.00%
CHILD NODES
Descriptor 1 Descriptor 2 Descriptor 3 Node Total Good Bad % Good % Bad
Income > 1.2 *installment 9,800 9,450 350 96.43% 3.57%
Income < 1.2* installment 200 50 150 25.00% 75.00%
Income > 1.2 *installment less than 2 other loans 8,800 8,550 250 97.16% 2.84%
Income > 1.2 *installment 2 or more other loans 1,000 900 100 90.00% 10.00%
Income > 1.2 *installment less than 2 other loans Income > 2.5*Installment 8,400 8,350 50 99.40% 0.60%
Income > 1.2 *installment less than 2 other loans Income < 2.5*Installment 400 200 200 50.00% 50.00%
Income > 1.2 *installment 2 or more other loans Prior Defaulter 150 70 80 46.67% 53.33%
Income > 1.2 *installment 2 or more other loans Nor Prior Default 850 830 20 97.65% 2.35%
Theoretical Implications of Eliminating Segments at Risk 9,250 9,180 70 99.24% 0.76%
4
© Jigsaw Academy Education Pvt. Ltd.
Findings and Recommendations
1. This Recommends an approach that could potentially reduce default rates of
IDIDI bank from a historical value of 5% on average to potentially less than 1%
2. This approach suggests detailed investigations / outright rejection of 3 particular
types of highly risky customers who have traditionally comprised the bulk of
defaulters:
3. The 3 risky segments are described below:
a) All Customers whose monthly incomes are less than 1.2 times monthly installments
• 3 out of 4 such customers default on average
b) Of customers whose income is higher than threshold described above, those having 2
or more other loans AND a prior default history
• Over 1 in 2 such customers default on average
c) Customers having incomes between 1.5 times monthly installment and 2.5 times
monthly installment, but have less than 2 other loans
• 1 in 2 such customers default on average
4. In theory, if the bank had never loaned any money to these 3 customer
segments, the bank's average default rate would have been only 0.76%
5
© Jigsaw Academy Education Pvt. Ltd.
