Case Study

Machine learning solution for real time fraud detection and verification

Challenge

In an increasing online world and many devices / channels for placing an order, customer fraud is becoming the biggest reason for margin erosion. In the last two years, online order fraud is growing twice the rate of online sales and the industry average ranges from 14 – 27%.

As retailers grapple with this problem and implement varied solutions for fraud-detection and verification, the most common problem is “false positives” – orders which are not fraud but classified as fraud. This is resulting in loss of a different kind – declining customer satisfaction and increasing customer frustration; During the holiday season such scenarios will lead to a significant loss of loyal customers.

Approach

Convergytics deployed a machine learning solution built on a big data enabled data layer to enable accurate fraud detection in real time. The solution which used a combination of unsupervised, supervised and heuristic analysis increased prediction of accuracy from 82% to nearly 93%. 

 

The different levels of fraud severity have helped in implemented less intrusive tiered verification process.

Result

Our client has realized in nearly $200 Million in savings in the last holiday season and has seen an uptick in the customer satisfaction by 14%.

Savings up by

$200 Million

Uptick in the Customer Satisfaction by

14%

Interested in Learning More?