Route quality KPIs: ASR, ACD, PDD, NER, MOS, packet loss and jitter

Quality is what separates a good route from a cheap one. In wholesale VoIP, a small set of key performance indicators (KPIs) tells you whether calls connect, how long people stay on the line, how quickly calls start ringing and how good the audio sounds. This lesson defines each KPI, shows how to calculate it and explains how to read them together.

ASR – Answer-Seizure Ratio

ASR = answered calls ÷ total call attempts × 100%

ASR measures how often calls are answered. Example with invented numbers: 1,000 attempts, 420 answered → ASR 42%.

ASR depends a lot on factors outside the route: how many dialled numbers are invalid, how many subscribers are busy or switched off, the time of day and the type of traffic. That is why ASR should be compared between routes to the same destination with similar traffic, not against a universal target.

NER – Network Effectiveness Ratio

NER = (answered + user busy + ring no answer + terminal rejects) ÷ attempts × 100%

NER tries to separate network problems from user behaviour. A call that reaches a busy subscriber or rings without answer is a network success — the network did its job, the person simply did not answer. NER therefore counts those outcomes as good, and only counts failures caused by the network (congestion, unreachable routes, server errors).

In SIP terms, responses such as 486 Busy Here or an unanswered timeout usually count as network successes, while 503 Service Unavailable or 500 errors count as network failures. Exact classification differs between tools, so document the rules you use.

A route with low ASR but high NER is probably fine — the traffic itself has low answer rates. A route with low ASR and low NER has a network problem.

ACD – Average Call Duration

ACD = total answered minutes ÷ answered calls

ACD shows whether people actually talk. Example: 420 answered calls with 1,470 total minutes → ACD 3.5 minutes. Poor audio, one-way audio and dropped calls all reduce ACD because people hang up. Like ASR, normal ACD depends heavily on traffic type: call-centre traffic naturally has much shorter calls than family calls.

Look at the distribution as well as the average. Many calls clustering at one short duration can indicate a technical drop or false answering.

PDD – Post-Dial Delay

PDD is the time between sending the INVITE and receiving the first indication of ringing (180 or 183). It is what the caller experiences as silence after dialling. Long PDD usually means many hops or a vendor retrying through several of its own routes; callers may give up before the phone rings.

Measure PDD from your own INVITE, per route. When you use failover, remember that PDD on the final route includes the time spent trying earlier routes.

Packet loss

Packet loss is the percentage of RTP packets that never arrive. It can be measured from RTP sequence numbers or from RTCP reports. Loss that arrives in bursts sounds worse than the same amount spread evenly. Compressed codecs such as G.729 tend to be more sensitive to loss than G.711.

Jitter

Jitter is the variation in packet arrival time. Receivers use a jitter buffer to smooth it out; when jitter exceeds what the buffer can absorb, late packets are discarded and the effect is the same as packet loss. Jitter often increases at busy times of day on congested networks.

Latency

Latency (delay) is how long audio takes to travel one way. High latency does not distort audio, but it makes conversation awkward: people interrupt each other and pause. Long transcontinental paths, satellite links and extra transcoding all add delay.

MOS – Mean Opinion Score

MOS expresses perceived voice quality on a scale from 1 (bad) to 5 (excellent). In live networks it is usually estimated from measured loss, jitter, latency and codec using a model such as the ITU-T E-model, rather than from people listening. Two points to remember:

• The best possible score depends on the codec; G.729 cannot reach the same maximum as G.711.
• MOS is an estimate. Use it to compare routes and spot changes, and confirm problems by listening to recordings.

Reading the KPIs together

Practical measurement tips

• Use enough calls. KPIs on a handful of calls are noise. Compare periods with meaningful volume.
• Use consistent time windows. Compare the same hours on different days, since traffic changes during the day.
• Segment. Break down by destination, vendor and customer. A route-level average can hide one bad sub-destination.
• Alert on change, not only thresholds. A sudden drop compared with the route’s own history is often more meaningful than a fixed limit.
• Keep evidence. When KPIs point to a problem, collect example calls with Call-IDs, traces and recordings before contacting the vendor.

For a shorter introduction, see our blog post How to check VoIP route quality. VoxInspector calculates these KPIs per route from live traffic and refreshes them every few seconds.

Examples in this course use invented names (Vendor X, Customer Y, Country Z) and documentation IP ranges. Menu names and options differ between softswitch versions; always check your own system and vendor documentation.