Our data and analytics team worked with one of the world’s biggest telecoms operator to identify and tackle wide scale customer non-payment worth €1.5m a year. 

Our telesales team were ten times more likely to sign up customers who default on their subscription payments. We needed a solution that would help identify those customers and address non-payment head-on.

Leading telecoms operator

Challenge

A major telecom operator realized that a significant number of its customers were defaulting on their mobile subscription payments. Of its more than 400,000 customers, 7% were likely to default representing a potential annual loss of revenue worth €1.5m.

The operator wanted to improve its fraud detection processes and identify ways to tackle wide scale fraud before it happened.

This meant a solution that would allow it to profile customers at activation to distinguish good and bad payers. It also meant analyzing subscriber activity after the first month to target early fraudsters.

Solution

Challenge

Based on thorough scrutiny of the available data, HyperCube created a predictive model that could be industrialized by the client. Analyzing data from over 400,000 customers, our model identified the three most influential activity drivers likely to suggest suspicious behavior.

Specifically, the model monitored:

  • Customer revenues. Higher revenues are an indicator of fraudulent conduct
  • The most frequently used (MFU) handset. Fraudsters are unlikely to put their regular handset to fraudulent use
  • The number of outbound and inbound calls made on the phone. A larger volume of calls, 350 or more, generally indicates a higher risk of default

Other fraud factors included customer and contract characteristics. Using this detailed analysis, HyperCube worked with the client to develop an action plan designed to discourage would-be fraudsters from signing up to the operator’s services.

Benefits

Challenge

The HyperCube predictive model outperformed existing methods used for fraud detection. It is:

  • Twice as likely to correctly classify dishonest customers in the highest risk category (class 5) at activation
  • Up to three times more accurate when activity variables based on the first month’s invoice payment were included

As a result of applying the HyperCube model, the operator avoided signing up 10,000 fraudsters per year and recruited an additional 20,000 good payers. This in turn ensured a healthy annual return on investment (ROI) of over €2m.

More use cases

Ready for the next generation of analytics?

Try HyperCube and experience its unrivaled range of features and ease of use Get started