What we learned since 2012
Practical lessons on losing less to fraud and keeping more good customers.
Real-time fraud decisions for instant payments
An instant payment cannot be recalled, so the decision has to be right in the second before it leaves.
Reaching customers before they leave
Customers rarely announce that they are leaving; their behaviour usually does, weeks before the account goes quiet.
Rule changes without surprises
Know what a new rule would stop, fraud and good customers alike, before any customer feels it.
Order tells the customer's story
What customers do, in order, shows who may leave and who is ready for more.
Measure fraud and friction together
A fraud rate that falls while good customers are turned away is a loss that does not show on the fraud report.
Explaining automated decisions
One record of each decision should satisfy the customer, the analyst and the regulator, each in their own words.
Predictions that hold up in use
A model that looks right on paper and misses in daily use usually learned from facts it could not have known at the time.
From AI pilot to production
Most AI pilots stall on data, cost and unclear value; choosing the right decision first avoids most of it.