top of page

Transforming Telecom FinTech: A Success Story of Slashing Loan Defaults by 35%

Abhishek Sheth

How a major telecom company reduced fraud losses by 35% with Matics Analytics AI powered solution. The AI fraud detection system analyzed customers and transactions data to identify suspicious activity and minimize bad debt from fraudulent accounts.


AI Fraud Detection

INTRO AND CLIENT BACKGROUND


A leading telecom company from Africa with over 5 million subscribers, recently entered the Fintech industry where they provide credit services and loans to their telecom customer base.


However, faced with the challenge of mounting bad debt loans, which amounted to $50,000 in 2022, the company sought a strategic approach to understand and address the root causes. The ultimate goal was clear: reduce loan defaults, create a proactive collection strategy, and safeguard revenue.



BUSINESS PROBLEMS


> Finding Eligible Customers for Loans:
  • The telecom company faced several challenges in its new venture. One big problem was figuring out eligible customers for loans.

  • They used a basic credit scoring system, which was based on limited and outdated data, such as customer age, gender, and tenure. Resulting in high loan defaults.


> Lower Recovery with Customer Churn:
  • Creating good plans to collect payments was also hard, and the company struggled with the serious issue of customers not paying and switching to other networks. This resulted in customer churn and lower recovery rates.



KEY CHALLENGES FACED


  • The first challenge came from the overwhelming task of managing and analyzing vast, unorganized data, including transactions, customer, and varius mobile plans data.

  • Compounding these issues was the reliance on traditional, generic, rule-based methods that failed to consider individual customer needs and behaviors.

  • As expected, the positive class (fraudstres) is very less as compared to the negative class (non-fraudsters), creating a data imbalance issue. This made it difficult to train accurate and robust machine learning models that could handle the complexity and diversity of the fraud cases.


OUR SOLUTION: FROM ACTION TO IMPACT


Action to Impact

To tackle these challenges head-on, our team implemented a comprehensive solution driven by AI and Analytics. Here’s how we turned the tide:


Data-Driven Insights


  • Leveraging advanced analytics, we delved into the unstructured data to identify key patterns, trends, and customer segments, providing the company with actionable insights.

  • We used data visualization techniques, such as graphs and charts, to present the insights and patterns in the data to identify the pain points and opportunities for improvement.


Machine Learning

  • We employed machine learning models to predict which customers were most likely to default on their loans. This proactive approach allowed the company to stay ahead of potential defaults.

  • We used various machine learning techniques, such as anomaly detection, clustering, and classification, to assign a risk score to each customer and transaction, based on various factors, such as customer behavior, payment history, location, device, and network.

  • With techniques such as oversampling and undersampling to address the data imbalance issue and improve the model performance.



Strategic Collection Actions

  • Armed with data-backed predictions, we strategically defined targeted actions to address potential payment delays in advance. This personalized approach to customer engagement was a game-changer for the company’s collection strategy.

  • Suggested the optimal actions for each customer segments, such as approving or rejecting new loans, sending personalized and timely reminders, offering incentives or penalties to existing customers. We also used A/B testing and feedback loops to evaluate and optimize the actions.



VALUE DELIVERED: HOW DID THE SOLUTIONS HELPED?


Fraud detection and prevention

1. 35% Reduction in Loan Defaults:


  • By identifying and proactive addressing the accounts, the company achieved a substantial 35% decrease in loan defaults, from $50,000 to $32,500 , in 2023. This not only secured revenue but also ensured a more sustainable financial future.


2. Collection Strategy Redefined:


  • The introduction of risk scores empowered the collection team to take necessary actions well in advance, minimizing losses and streamlining the collection process.

  • The collection team was able to prioritize the high-risk accounts, send customized and timely reminders, offer incentives or penalties, and escalate or close accounts. This resulted in an increase in the recovery rate.


3. Enhanced Credit Risk Assessment:


  • Every customer received a precise risk score, revolutionizing credit risk assessment and providing the company with a robust tool for decision-making.

  • The company was able to approve or reject loans based on the risk score, rather than on limited and outdated techniques.



CONCLUSION


By using data-driven insights and machine learning, the African telecom company transformed its FinTech operations. This study demonstrates how AI solutions can overcome obstacles and create opportunities for long-term success and profitability in the dynamic telecom and finance industry.


Facing similar challenges? Let us know your business problems in the comments or reach out to us directly at: info@maticsanalytics.com

Comments


bottom of page