AI-Powered Chargeback Wins: More Wins, Less Time

Introduction

Chargebacks – the disputes customers raise with their credit card providers or banks – are a major headache. A leading provider of chargeback management solutions is currently reliant on manual dispute processing and faces challenges of operational cost and subjective decision-making based on team training. They embarked on a data-driven initiative to optimize its dispute resolution process. By leveraging machine learning models, the company aims to automate the triage of disputes, efficiently allocating resources by identifying cases best suited for automation. This initiative has the potential to significantly improve the dispute resolution success rate, and scale operations.

‍

Challenges Faced

• Time-consuming manual work: Chargeback analysts spend a significant amount of time gathering evidence, building cases, and communicating with customers.
• Increased costs: The process of manual dispute operations was expensive.
• Decision inconsistency: Subjective decisions based on team experience led to inconsistent outcomes.
• Difficulty identifying fraudulent patterns: Recognizing fraudulent trends and patterns was challenging, leading to missed opportunities to prevent disputes and recover lost revenue.
• Limited visibility into chargeback trends: Traditional methods did not provide sufficient insights into the root causes of chargebacks, making it difficult to develop preventative strategies.
  ‍

Solution Implemented

1. Data ingestion & model training

The company integrated their historical dataset, comprising thousands of resolved cases, into the RapidCanvas AutoAI platform. This included details like dispute reason codes, merchant categories, dispute amounts, currency, and card types. The details of resolution - win/loss outcome, and time to resolution also were brought in.

This data, once cleaned and processed, was used to train a ML model based on the learnings codified through all the previous case resolutions. The integrated data was analyzed and patterns are tracked in cases to judge them as winnable. The chargeback solutions provider uses an internal system to manage chargebacks, where data is gathered for each case. Once trained, the RapidCanvas ML model was integrated with this internal system and the model helps assign a winnability score for each case. The score is displayed directly on the company’s dashboard and helps chargeback analysts prioritize cases.

‍3. Win rate prediction

The trained model generates real-time win rate probabilities for each new case, ranging from "low" to "high." This allows the team to prioritize efforts on cases with the highest potential for successful defense. Alongside the probability, key factors influencing the prediction were also highlighted to enhance transparency and decision-making. As new data became available, the model was continuously updated to maintain its accuracy.

4. Lost case analysis

An analysis of the cases lost was also carried out to identify recurring reasons for declined chargebacks, particularly in cases where a lost case’s characteristics matched very closely with a case that was successful.

‍

Benefits

• Analyst time saved

‍With an accurate estimate of the ability to win disputes of each case, the time spent could also be allocated accordingly. High win probability cases and low probability cases were resolved with automated decisions leading the way , while cases with medium probability would be the ones analysts focus on to get better outcomes. The team saw a 50% reduction in operational time spent doing the reviews of the same number of cases. Case allocation between senior and junior agents is also done more efficiently and based on the ease of the case.

• Increased confidence

‍Armed with insights from analyzing the lost cases, chargeback analysts can approach disputes with greater confidence, potentially improving overall success rates.

• Better understanding of chargeback causes

‍By analyzing both winning and losing cases in a data-driven model-led process merchants gain valuable insights.

• Automated review

With an AI model that detects patterns and learns over time, the case review of known combinations of factors can be automated, with a manual validation step introduced to verify the model’s assessment. This reduces the manual workload for similar cases.

• Data-driven decisions

‍The objective predictions eliminated subjective biases and fostered informed decision-making throughout the chargeback dispute resolution process.

• Scalability and growth

AI enables the provider to handle larger case volumes efficiently, with a more efficient use of resources.

• Increased win rate

With this implementation, the company is looking to increase their overall win rate, boosting revenue recovery and client satisfaction.

‍

Conclusion

By replacing subjective assessments with AI-driven insights, the provider of chargeback management solutions transformed its chargeback management approach. RapidCanvas empowered the company to predict win rates accurately, optimize resource allocation, and ultimately recover more revenue while scaling its operations effectively. This case study exemplifies the transformative power of AI in addressing complex business challenges and unlocking new levels of efficiency and growth.