The problem

The payments industry has spent two decades optimising for fraud prevention. It has largely ignored the other side of the ledger: false declines — legitimate transactions rejected by overzealous risk models. Our research across 4.2M transactions on 18 e-commerce platforms found that for every €1 lost to fraud, operators lose €13 to false declines. Yet fewer than 12% of fraud teams track decline rates as a primary KPI.

The data

The most damaging aspect of false declines is not the immediate revenue loss — it is the long-term customer attrition. Our cohort analysis found that 41% of customers who experience a false decline never attempt another purchase with that merchant. For subscription businesses, the lifetime value impact can exceed 200x the value of the declined transaction.

Root causes

False declines cluster around three root causes: threshold miscalibration (risk thresholds set at fraud minimisation rather than profit maximisation), signal staleness (models trained on historical fraud patterns that no longer reflect current attacker behaviour), and context blindness (models that cannot distinguish a legitimate high-value purchase from an anomalous one because they lack the customer's purchase history context).

The fix

Serixo's risk scoring engine optimises for expected value — not fraud rate. Each decision factors in the estimated probability of fraud, the estimated probability of false decline, the transaction value, and the estimated customer lifetime value. This shift from binary fraud/not-fraud to expected value maximisation typically reduces false declines by 35–60% while holding fraud rates constant or improving them.

Results

Across the 18 platforms in our study, switching to expected-value optimised scoring reduced false decline rates from an average of 8.3% to 3.1% — a 63% reduction. Annualised revenue recovery across the cohort averaged €2.4M per platform. Fraud rates moved less than 0.2 percentage points in either direction.