Historically, card-not-present merchants have protected themselves against fraud using manual review of transactions, rule-based automated scoring, and more recently, machine learning-based scoring. While all three methods can be effective, all have liabilities. Manual review is too costly to be deployed on a large scale. For sellers offering same-day delivery, it is not feasible. Rule-based scoring denies too many good transactions (false positives for fraud). This technology and machine-learning engines rely on historical data to spot patterns, which leaves them a little late to discover the...