Should Riskified's Chargeback Guarantee prioritize higher approval rates or lower fraud liability for merchants?

Introduction

The trade-off between higher approval rates and lower fraud liability for merchants in Riskified's Chargeback Guarantee is a critical decision that impacts both our business model and our customers' success. This scenario requires us to balance the potential for increased revenue against the risk of financial losses due to fraud. I'll approach this analysis by examining the product's core value proposition, evaluating key metrics, and designing experiments to inform our decision-making process.

Step 1

Clarifying Questions (3 minutes)

• Based on recent market trends, I'm thinking fraud patterns might be evolving rapidly. Could you share any insights on how fraud patterns have changed in the last 6-12 months?

Why it matters: Helps understand if current approval/fraud rates are still relevant Expected answer: Significant shifts in fraud tactics, especially in certain industries Impact on approach: Would prioritize adaptability in our solution and potentially segment our approach by industry

• Considering our revenue model, I assume we charge a percentage of approved transactions. Is this correct, and are there any other key revenue streams we should consider?

Why it matters: Directly impacts the financial implications of our decision Expected answer: Confirmation of percentage-based model, possibly with additional fees Impact on approach: Would influence how we balance approval rates against potential losses

• Looking at our user segments, I'm curious about the diversity of our merchant base. Can you give me an overview of our merchant distribution by size and industry?

Why it matters: Different merchants may have varying risk tolerances and needs Expected answer: Mix of small to large merchants across various industries Impact on approach: Might lead to a segmented strategy rather than a one-size-fits-all solution

• From a technical perspective, I'm wondering about our current fraud detection capabilities. How advanced is our machine learning model, and what data sources are we currently leveraging?

Why it matters: Informs the feasibility of improving approval rates without increasing fraud Expected answer: Details on model sophistication and data inputs Impact on approach: Would guide the level of investment needed in improving our technology

• Considering resource allocation, how much flexibility do we have to invest in improving our fraud detection capabilities if needed?

Why it matters: Determines the scope of potential solutions Expected answer: Some budget available, but needs strong justification Impact on approach: Would influence whether we focus on optimizing current systems or developing new capabilities