In what ways can we improve Modalku's risk assessment algorithm to better evaluate borrowers' creditworthiness?

Introduction

To improve Modalku's risk assessment algorithm for better evaluating borrowers' creditworthiness, we need to take a comprehensive approach that considers various factors and stakeholders. I'll outline a strategy to enhance the algorithm's accuracy, efficiency, and fairness, keeping in mind the evolving fintech landscape and the unique challenges of the Indonesian market.

Step 1

Clarifying Questions (5 mins)

• Looking at Modalku's position in the Indonesian P2P lending market, I'm curious about the current performance of the risk assessment algorithm. Could you share some insights on the default rates and how they compare to industry standards?

Why it matters: This helps us understand the baseline and set improvement targets. Expected answer: Default rates are slightly above industry average, around 3-4%. Impact on approach: Would focus on reducing false positives in credit approvals.

• Considering the diverse Indonesian market, I'm wondering about the data sources currently used in the algorithm. What types of alternative data, if any, are being incorporated beyond traditional credit scores?

Why it matters: Determines the breadth of our data inputs and potential areas for expansion. Expected answer: Currently using basic financial data, social media, and mobile usage patterns. Impact on approach: Would explore integrating more alternative data sources for a holistic view.

• Given the rapid evolution of machine learning techniques, I'm interested in the current technological stack of the algorithm. What ML models or techniques are currently employed in the risk assessment process?

Why it matters: Helps identify potential areas for technological upgrade or optimization. Expected answer: Using a combination of logistic regression and random forest models. Impact on approach: Would consider implementing more advanced techniques like gradient boosting or neural networks.

• Considering the regulatory environment in Indonesia, I'm curious about any recent or upcoming changes in fintech regulations that might impact our risk assessment practices. Are there any new compliance requirements we need to factor in?

Why it matters: Ensures our improvements align with regulatory standards and future-proofs the algorithm. Expected answer: New regulations requiring increased transparency in AI decision-making. Impact on approach: Would prioritize explainability and fairness in our algorithm enhancements.