Case Study: eftpos achieves stronger fraud detection with Featurespace's machine-learning platform

eftpos cuts fraud with Featurespace using 33% true positive rate

eftpos, Australia's leading debit card payment network, faced the challenge of rising fraud, particularly in card-not-present transactions, as its dual-network cards grew in popularity. Existing systems struggled to detect this fraud effectively without disrupting genuine customer transactions. To address this, eftpos partnered with Featurespace to implement a comprehensive machine-learning fraud solution.

Featurespace provided its Adaptive Behavioral Analytics platform, deploying a bespoke model for card-present transactions and a innovative "cold-start" rules-model hybrid for card-not-present transactions despite lacking historical CNP data. This solution gave eftpos members a powerful network-wide defense. Featurespace's implementation achieved a strong 40.6% value detection rate with a very low 0.1% false positive rate, and its sub-classification codes demonstrated high accuracy, with some risk types showing fraud rates 10–100 times higher than baseline.

eftpos

Derek Kidd

Head of Fraud, Risk and Scheme Compliance, eftpos