- Conduct a job analysis to determine critical behaviors for success in CRM roles (e.g., customer service representatives, sales representatives, account managers).
- Gather input from managers, employees, and customers to identify essential behaviors.
- Align behaviors with company values and CRM goals.
2. Define Performance Levels:
- Establish clear and measurable performance levels for each behavior (e.g., unsatisfactory, needs improvement, meets expectations, exceeds expectations).
- Use specific examples to illustrate each level.
3. Create the Scorecard:
- Develop a visual representation of the scorecard, listing behaviors and performance levels.
- Use a simple and easy-to-understand format.
3. Purpose
The purpose of this use-case is to develop a
credit scoring model for loan approval in a
retail banking environment. Credit scoring
helps banks assess the creditworthiness of
loan applicants, allowing them to make
informed decisions regarding loan approvals.
4. Benefits
Risk Mitigation: A robust credit scoring
model helps the bank minimize the risk of loan
defaults by identifying applicants with a higher
likelihood of repayment.
Efficiency: Automated credit scoring
streamlines the loan application process,
reducing the time and resources required for
manual assessments.
Fairness: A well-designed model can
promote fair lending practices by evaluating
applicants based on objective criteria rather
than subjective judgment.
Profitability: By approving loans to
applicants with a higher probability of
repayment, the bank can increase its loan
portfolio profitability.
5. Data Sources
APPLICANT DATA: THIS
INCLUDES PERSONAL
INFORMATION (AGE,
INCOME, EMPLOYMENT
STATUS), FINANCIAL DATA
(CREDIT HISTORY,
OUTSTANDING DEBTS),
AND LOAN-SPECIFIC
DETAILS (LOAN AMOUNT,
PURPOSE).
CREDIT BUREAU DATA:
INFORMATION FROM
CREDIT BUREAUS ABOUT
AN APPLICANT'S CREDIT
HISTORY, INCLUDING
CREDIT SCORES AND
PAYMENT HISTORY.
BANK'S HISTORICAL
DATA: PAST LOAN
APPLICATION AND
REPAYMENT DATA TO
TRAIN THE MODEL.
EXTERNAL ECONOMIC
DATA: ECONOMIC
INDICATORS AND INTEREST
RATES THAT CAN IMPACT
AN APPLICANT'S ABILITY TO
REPAY LOANS.
7. Key Components of Behavioural
Scorecard
• Credit History:
An individual's credit history is a critical component of a behavioral scorecard. It includes
information about credit accounts, such as credit cards, loans, and mortgages.
Elements of credit history include the types of accounts, account opening dates, credit limits,
and current balances.
• Payment History:
Payment history is one of the most important factors in credit scoring. It reflects how
consistently an individual makes payments on time.
Late payments, missed payments, and delinquent accounts negatively impact the credit score.
• Credit Utilization:
Credit utilization measures how much of an individual's available credit they are currently
using.
High credit card balances relative to credit limits can lower the credit score, as it may indicate
financial stress.
• Length of Credit History:
The length of time an individual has had credit accounts is a factor. Longer credit histories
generally have a positive impact on credit scores.
It considers the age of the oldest account, the average age of accounts, and the age of
specific accounts.
9. Methodology of Behavioral Scorecard for
Credit Scoring in Loan Approval
• Data Collection:
Obtain credit data from credit reporting agencies (CRAs) .
These agencies collect and maintain credit information on
individuals.
Gather information on the applicant's credit accounts, payment
history, outstanding debts, public records (e.g., bankruptcies,
liens), and other relevant financial data.
• Data Preprocessing:
Clean and preprocess the collected data to ensure accuracy
and consistency. This includes correcting errors, handling
missing information, and standardizing data formats.
Normalize or scale data to bring all variables to a common
scale for fair evaluation.
• Model Training:
Divide the historical data into training and validation sets
to train and evaluate the model's performance.
The model learns the relationships between the
selected features and the likelihood of loan default.
• Model Validation:
Assess the model's predictive performance using
various metrics such as accuracy, precision, recall, and
F1-score. Fine-tune the model if necessary to improve
its accuracy and reliability.
10. Benefits of Using Behavioral Scorecards
OBJECTIVE AND
DATA-DRIVEN
PREDICTIVE OF
CREDIT BEHAVIOR
CONSISTENCY IN
DECISION-MAKING
FASTER LOAN
APPROVAL PROCESS
11. Key components of application scorecard for
Credit Scoring in Loan Approval
• Credit Score or Credit Bureau Information: The
applicant's credit score is a fundamental
component. It provides a snapshot of their credit
history, including payment history, outstanding
debts, and credit utilization.
• Income and Employment Information: Applicant's
income level, source of income, and employment
history are crucial. Lenders want to determine if
the applicant has a stable source of income to
repay the loan.
• Loan Amount and Purpose: The requested
loan amount and the purpose of the loan
are considered in relation to the
applicant's income and financial situation.
Lenders assess whether the loan amount
aligns with the intended purpose.
• Loan Term: The length of the loan term is
evaluated in terms of its appropriateness
for the applicant's financial circumstances.
• Residence History: The length of time the
applicant has lived at their current
residence may be considered. Longer
stability at one address can be indicative of
financial stability.
12. Methodology of application scorecard
for Credit Scoring in Loan Approval
• Data Collection: Gather detailed information from
the loan application form, which includes the
applicant's personal, financial, and employment
details.
• Collect any supporting documents, such as pay
stubs, tax returns, and bank statements, to verify
the information provided.
• Data Verification and Validation: Verify the
accuracy of the information provided by the
applicant. This includes confirming income,
employment history, and other financial details.
• Detect and address any discrepancies or
inconsistencies in the application data.
• Credit Report Retrieval: Obtain the applicant's
credit report from one or more credit
reporting agencies (CRAs) to assess their
credit history and credit score.
• Verify that the information in the credit report
aligns with what the applicant has provided in
the loan application.
• Feature Selection and Engineering: Determine
which applicant attributes and financial
variables will be included in the application
scorecard. Common attributes include income,
credit score, DTI ratio, and employment
stability.
• Create derived features if necessary to
improve the predictive accuracy of the
scorecard.
13. Collection Scorecard
Data Collection: Gather
information on the applicant's
credit history, including
collections, late payments, and
charge-offs.
Data Preprocessing: Clean
and preprocess data for
accuracy and consistency.
Feature Selection: Identify
relevant features like the
number and age of collections.
Scoring Model Selection:
Choose an appropriate scoring
model or algorithm.
Model Training: Train the
model using historical data
with collection information.
Model Validation: Assess
predictive performance and
fine-tune the model if needed.
Scorecard Development:
Create a scorecard based on
model coefficients and feature
importance.
Scoring Process: Calculate a
score for each applicant based
on their collection history.
Credit Decision: Use score
thresholds to determine loan
approval, denial, or further
review.
Documentation and
Reporting: Maintain detailed
documentation and provide
reports on the scoring
process.
Monitoring and
Maintenance: Continuously
update the scorecard to adapt
to changing conditions and
regulatory requirements.
14. Modeling Approach
Data Collection: Gather
applicant data, credit bureau
data, historical loan data, and
external economic data.
Data Preprocessing: Clean
and preprocess the data,
handle missing values, and
perform feature engineering
to create relevant features.
Feature Selection: Identify
the most influential features
for credit scoring.
Data Split: Split the data
into training, validation, and
test sets.
Model Selection: Choose a
classification algorithm suited
for credit scoring, such as
logistic regression, decision
trees, random forests, or
gradient boosting.
Model Training: Train the
selected model using the
training data, optimizing
hyperparameters for the best
performance.
Model Evaluation: the
model's performance using
metrics like accuracy,
precision, recall, F1-score,
and ROC-AUC on the
validation set.Evaluate
Hyperparameter Tuning:
Fine-tune the model's
hyperparameters to improve
its performance.
15. Conclusion
The use of behavioral scorecards, application
scorecards, and collection scorecards in credit scoring
for loan approval is essential for making informed
lending decisions. These scorecards leverage data
analytics, statistical modeling, and historical information
to assess an applicant's creditworthiness accurately. By
evaluating an applicant's financial behavior, credit
history, and collection actions, lenders can mitigate
risks, streamline approval processes, and ensure
responsible lending practices. These scorecards provide
a structured and data-driven approach to support both
lenders and borrowers in achieving mutually beneficial
financial outcomes while adhering to regulatory
compliance and privacy standards.