It may be measured one-way or round-trip and expressed as a count or proportion. The definition must identify the endpoints, traffic type, timing threshold, and measurement interval.
Packets may be discarded because queues overflow during congestion, transmission errors occur, equipment fails, a route changes, or policy filters the traffic. Reliable transports may retransmit missing data, masking loss while increasing completion time; delay-sensitive or unreliable applications may expose it immediately.
Key points
Measurement choiceUse representative packet sizes, protocols, paths, and time windows. One-way tests require suitable clock synchronization; round-trip tests cannot reliably assign loss to the outbound or return direction.
Diagnostic contextCorrelate end-to-end loss with interface error and discard counters, queue utilization, route changes, radio conditions, captures, and application behavior before assigning a cause.
Service effectEven modest loss can reduce TCP throughput or degrade voice, video, gaming, and industrial messaging, depending on retransmission, buffering, redundancy, and application tolerance.
Important limitationMissing probes do not by themselves prove packet loss or malicious interference: filtering, rate-limited replies, asymmetric paths, sampling, and capture drops can produce similar evidence. Authorize and rate-limit active tests, especially on safety- or availability-sensitive operational technology (OT) networks.