Latency in Fraud Prevention: Don't Let the Numbers Fool You (Transaction Latency Explained)

If a fraud solution provider is boasting about their latency numbers, be cautious—it’s easy to be misled.

Latency is simply the time it takes to complete a task. It can be measured in two ways: average latency and percentile latency.

Average latency is straightforward, but percentile latency is more revealing. Even a few slow transactions can drastically skew the results, as Feedzai's Chief Science Officer Pedro Bizarro points out.

For example, in a system processing 100 million transactions, 99.95% might have a latency of 20 milliseconds, but if 0.05% spike to 500 milliseconds for just 50 seconds, the average barely changes, while the 99.95% percentile jumps to 500 milliseconds. This shows how even brief slowdowns can significantly affect metrics, making percentile latency a better indicator of system quality.

Also, consider what type of latency is being measured—server, network? To truly evaluate performance, you need to look at both percentile latency and communication costs.

#fraudprevention #riskmanagement #computerscience