MEASUREMENTS / 5 MIN READ
Read the Statistics Behind the Sound.
A number is useful only when you know what was measured, where, in which direction, and over what interval.
Use Targets as Prompts, Not Verdicts
| Observation | Broad Planning Target | Important Limitation |
|---|---|---|
| One-way end-to-end delay | Below 150 ms | Network transit alone omits encoding, packetization, buffering and playback. |
| Jitter | Below 30 ms | Different counters use different definitions; averages hide excursions. |
| Packet loss | Below 1%; aim near zero | A small burst can be more disruptive than isolated losses with the same average. |
These are broad voice-design references drawn from Cisco guidance, not a provider SLA or a universal boundary between good and bad audio. ABOUTVOIP marks values at or above these targets for review. The codec, loss concealment, buffering, listening conditions and measurement method still matter.
RTT Is Not Mouth-to-ear Delay
A ping reply measures a round trip involving an ICMP responder. Its outbound and return paths may differ, and a web or ping target may not be your media endpoint. Halving RTT does not establish one-way voice latency. The reviewer therefore labels RTT as context and does not apply the one-way target to it.
Even a measured one-way network value may exclude parts of the audio chain. Record whether a dashboard reports transport delay, round-trip delay or an end-to-end estimate before entering it.
Jitter Needs a Definition
RTP interarrival jitter is a smoothed estimate of variation in packet transit timing. In RTCP reports it is expressed in RTP timestamp units; software often converts it to milliseconds. A vendor’s “maximum jitter,” an average, and successive ping differences are not interchangeable.
A jitter buffer trades additional delay for more orderly playback. Late arrival can result in audio being discarded even when a network-loss counter appears low. Check the receiver’s discarded or late-packet counters alongside transport statistics.
Loss Needs a Time Window
Ask whether the percentage covers the whole call or just the affected minute, and whether it reports received media, sent media, retransmissions or discarded packets. One endpoint can hear a problem that the other does not. Preserve per-direction observations rather than blending them into a single average.
Why There Is No Invented MOS Score
Subjective opinion scores and model-based quality estimates are different things. Three manually entered counters do not supply the full basis for a defensible quality model. ABOUTVOIP shows the values and their limitations instead of manufacturing a precise-looking score.
A Useful Comparison
- Capture a short affected interval and a known-good interval using the same method.
- Keep device, codec and call destination consistent where practical.
- Record direction, time zone, sample window and whether the network was busy.
- Review spikes and consecutive losses, not just overall averages.
Further reading: Cisco voice delay, jitter and loss guidance · RTP and RTCP — RFC 3550.
Updated September 6, 2026. Corrections and field feedback: [email protected]. Please leave customer identifiers and sensitive call records out of your first message.