Packet loss percentage is only meaningful with the target, duration and number of samples of the test. Losing one in four pings does not prove a 25 percent long-term loss.
Basic approach
In measurement, sent and received packets must be matched; The timeout limit also affects the outcome. ICMP responses may be low priority on some devices. It is possible to limit diagnostic packets when application traffic is operating normally.
Application steps
- Clearly record the missing observed target and time interval.
- Compare two significant points, such as the local gateway and your own application target.
- Distinguish sustained loss to the final destination from the mere lack of diagnostic response of the intermediate device.
Practical example
If one hop shows 60 percent loss while subsequent hops and the final destination are lossless, you cannot say that the device drops 60 percent of all traffic it forwards. The intermediate hop may be limiting only its own ICMP replies.
Interpret the result correctly
More samples reduce uncertainty, but do not reduce the entire result to a single average if test conditions vary. Store wired and wireless results separately. Instead of automatically labeling packet loss as an ISP failure, investigate where it starts.
Source and follow-up reading
Protocol or command details: RFC 2680. The steps and example scenario are IPScans editorial narrative.