This document discusses using alternative data sources and psychometrics to better assess credit risk for applicants. It proposes a framework called MAP (Metadata, Application, Psychometric) that leverages three scores: a metadata score using additional customer data, an application score using traditional credit data, and a psychometric score using a customer questionnaire. Together these scores are meant to provide a more well-rounded risk assessment, especially for applicants with thin credit files. The document outlines the development process for the psychometric score, which involves defining dimensions of creditworthiness, creating and refining questionnaire items, analyzing sample data reliability, and ongoing refinement. Implementing this MAP framework could provide lenders with an additional risk assessment layer and first mover advantage.
8. 8
Application Score
Behavioral Score
Collection Score
Traditional Data
Data Science & Analytics based solutions
Trusted and tested over time
Powerful distinction
Capability to Pay Assessment
Large acceptance and understanding
9. 9
Telco Score
Utilities Score
SMS Info Score
Psychometric Score
Social Score
New to credit
Unbanked
Thin Files/ Less Loans
Emerging Bureaus
Social media access – Social Score
Customer provided data - Psychometrics
Alternate Data
AI based solutions
10. 10
Alterate DataTraditional Data
Evolving Sciences
Correlation to Risk
Access to data
Privacy Issues
Thin File Customers
New to Credit
Short data history
No Data / Startups
Emerging/No Bureau
16. 16
Metadata
Score
Adding one more layer of Data Science and
Analytics in the overall assessment
Identification of
gaming/fraud
Reliability & Seriousness
Too fast or too slow?
Lie Detection
Collection of data during and after
response – constant refinement
17. 17
Application
Score
Comes along with Expert
Application score based
on CRIF Experience
Develop and fit in a
Bespoke* Application
Score
Embed your
Own Application Score
19. 19
Psychometrics
- Objective
measurement of
skills and
knowledge,
abilities, attitudes,
personality traits,
and educational
achievement
Application
- Vastly
successful in HR &
Recruitments
- Recent efforts on
applying to world
of credit risk
management
Correlation
- What are the
attributes that
correlate to my
repayment risk
- Some work
already being
done in this field
Success
- Successful test
case studies in
banking world
from EFL
- Metadata
elements and
integration with
data by CRIF
Challenges
- Fraud
identification
- Language and
Culture
-Empirical data
validation
- Time consuming
20. 20
C
O
S M
O
Scorecard
27 Dimensions, 200+ Questions, 500 responders, Final 40, 12 Traits
Reliability & Consistency, Factor Analysis, Lie Detection etc..
Honesty
Discipline
Responsibility
Overspending
Social Desirability
21. 21
1 Animator
2 Web Developers
2 Professional Psychologists
7 Data Scientists & Credit Analysts
12 months development cycle
Always in β Mode
The Team & Effort
◊Define Customer Segment Base
◊Dimensions of Credit Worthiness
◊Literature Review
◊Item Writing
◊Item Refinement
◊Sample Collection
◊Data Analysis – Reliability
◊Empirical Validation
◊Constant Refinement
The Quiz