Manual review is the step in fraud screening where a human analyst evaluates a transaction, account, or claim that automated systems couldn’t confidently approve or decline. The analyst weighs the evidence, sometimes contacts the shopper, and makes the call. Done well, manual review rescues good orders that rules would have rejected; done poorly, it becomes a slow, expensive bottleneck.
Where Manual Review Fits in Fraud Screening
Most modern fraud stacks decide the clear cases automatically and route the grey area to people. A risk score in the middle band, a first order from a new device with a mismatched address, or a high value order with mixed signals all land in a review queue. The analyst looks at what the model saw plus context the model may lack: order notes, communication history, and judgment about whether the story hangs together.
The Trade-Off Manual Review Manages
Auto-declining everything uncertain is the expensive path. PYMNTS Intelligence research found nearly half of merchants estimate up to 5% of the orders they decline are actually legitimate, roughly $50 billion in lost revenue industrywide, and 85% of merchants name preventing fraud without degrading the customer experience as their top challenge. Manual review exists to claw back those false declines: a person can approve the unusual-but-legitimate order that a rule cannot.
What Good Manual Review Looks Like
Three properties separate effective review operations from bottlenecks. Selectivity: only genuinely ambiguous cases reach humans, so queues stay small. Speed: decisions land in minutes or hours, not days, because a delayed approval is a degraded customer experience. Feedback: every human decision trains the automated layer, so the same pattern doesn’t need a human twice. Review capacity is also a staffing question, which is why many brands use a provider’s expert analysts instead of building an in-house team.
How Wyllo Helps
Wyllo’s screening decides the overwhelming majority of orders in real time and automatically escalates the true grey area to seasoned fraud analysts, a layer Wyllo calls Proactive Review. Merchants get the approval rate lift of expert judgment without staffing a review desk. Learn more about Wyllo Payment Fraud Protection.
Frequently Asked Questions
Does manual review slow down orders?
Only the small share routed to it, and a well-run operation clears most reviews quickly. The alternative is worse: auto-declining ambiguous orders trades a short delay for lost revenue and a turned-away customer.
What share of orders should go to manual review?
There is no universal number, but the direction is clear: as models improve, the reviewed share should shrink. A growing review queue usually signals that the automated layer needs better signals, not that more analysts are needed.
What do analysts look at during a review?
Identity coherence (does the buyer match the history), device and network signals, order composition, delivery details, past behavior for the identity, and, when needed, direct contact with the shopper to verify the order.
