ai was approached by a major regional Telco that was experiencing significant fraud via online sales channels. Chargeback ratios in these channels were as high as 1.8% of total monthly transactions.
Due to the chargeback situation, the Telco was considering suspension of sales on a number of online sales channels. ai was tasked with reducing the chargeback ratio, whilst minimising the impact on genuine customer business.
ai was provided with 12-month historical transaction data set consisting of just 8 transaction data points, including fraud indicator. ai applied it’s machine learning solution, SmartRule, to create an optimal set of business rules. 4 separate rule set proposals were developed, to enable the client to select an optimal performance mix between Fraud Detection and Customer Acceptance that matched with internal risk tolerance. Following one-off configuration, the process to generate all 4 rule sets took only 6 hours.
The agreed optimal rule strategy consisted of 9 high performance business rules. The optimal strategy detected 52% of chargebacks whilst maintaining a genuine customer decline rate of just 0.7%. Performance of the rule strategy was proven via a blind-test data exercise.
The rule set was implemented in RiskNet Real-Time to ensure that suspected fraudulent transactions were declined at authorisation stage.
Comprehensive reporting on declined transactions was provided to the Telco client using dashboards within the SmartIntelligence reporting tool. This enabled the client to understand their decline rate and profile of declined customers in near real-time.