Why Quality Metrics Matter
In wholesale voice, price is visible and immediate; quality is often only apparent when something goes wrong. Understanding and monitoring quality metrics transforms voice performance from a reactive complaint-handling exercise into a proactive management discipline. Carriers and enterprises that track the right metrics can identify degradation before it reaches the customer, negotiate SLAs with confidence, and make data-driven routing decisions.
ASR — Answer Seizure Ratio
ASR is the percentage of call attempts that result in a successful answer. It is calculated as:
ASR = (Answered calls / Total call attempts) × 100
ASR is the single most widely cited quality indicator in wholesale voice and the primary output of LCR quality monitoring. However, it must be interpreted in context:
- Normal ASR ranges vary by destination. Consumer mobile traffic to the UK typically achieves 55-70% ASR; traffic to developing markets with high call failure rates may see 30-45% and still represent good quality for that destination.
- Low ASR is not always a carrier quality problem. A contact centre dialling unanswered leads will naturally produce low ASR. Segment your ASR analysis by traffic type.
- Sudden ASR drops indicate route issues. A fall of more than 10 percentage points within a 15-minute window on a stable route is a reliable indicator of route degradation or carrier switching events.
ACD — Average Call Duration
ACD measures the mean duration of answered calls. Short ACD — particularly for routes that normally produce longer calls — is a strong secondary indicator of call quality issues. If calls are connecting but users are hanging up quickly, the route may be experiencing audio problems that aren't captured in ASR alone.
ACD benchmarks vary dramatically by traffic type. Outbound sales calls may average 3-5 minutes; customer service calls 8-12 minutes; contact centre agent-to-agent transfers much shorter. Establish a baseline ACD for each traffic segment and alert on deviations greater than 20%.
PDD — Post-Dial Delay
PDD is the time elapsed between the caller completing dialling and the first audible ringback tone. It is a direct measure of routing and signalling speed. Excessive PDD causes callers to assume the call has failed and hang up — inflating call attempt volumes and reducing effective ASR.
- Excellent: Under 2 seconds
- Acceptable: 2–4 seconds
- Poor: 4–7 seconds
- Unacceptable: Over 7 seconds
High PDD is often caused by slow SIP routing through multiple intermediary carriers, excessive DNS resolution time, or overloaded signalling servers. Direct interconnect routes — such as those Mokrina maintains with Tier-1 carriers — consistently achieve sub-2-second PDD.
MOS — Mean Opinion Score
MOS is a measure of perceived voice quality, expressed on a scale from 1 (unacceptable) to 5 (excellent). It is derived from objective measurements of jitter, packet loss, and latency — translated into a score that approximates subjective human perception via the E-model algorithm (ITU-T G.107).
- 4.3 – 5.0: Excellent — perceptually identical to a landline call
- 4.0 – 4.3: Good — minor degradation, acceptable for all use cases
- 3.6 – 4.0: Fair — noticeable but not distracting quality reduction
- 3.1 – 3.6: Poor — significant degradation, not acceptable for business use
- Below 3.1: Unacceptable
MOS scores are calculated per-call by your SBC or media proxy using RTCP-XR reporting. Not all carriers report MOS — Mokrina's CDR platform includes per-call MOS scoring for all traffic, enabling systematic quality monitoring.
Jitter, Packet Loss, and Latency
These three network-layer metrics feed directly into MOS calculations and should be monitored independently:
- Jitter (variation in packet arrival times): Under 20ms is acceptable; above 50ms causes audible audio quality degradation. Jitter buffers in SBCs and endpoints compensate for moderate jitter but introduce additional latency.
- Packet loss: Under 1% is acceptable for voice; above 3% causes noticeable audio clipping and dropouts. Unlike data protocols, VoIP does not retransmit lost packets — each lost packet is simply silent audio.
- Round-trip latency: Under 150ms is the ITU-T G.114 recommendation for acceptable voice quality. Above 300ms round-trip, conversations become awkward due to echo and talk-over effects.
Building a Quality Monitoring Framework
- Define quality thresholds for each metric by destination and traffic type, not as universal values.
- Measure over appropriate windows: 15-minute intervals for real-time alerting; daily and weekly averages for trend analysis.
- Correlate ASR drops with ACD and PDD data to distinguish call quality issues from traffic pattern changes.
- Review route performance weekly and remove routes that consistently underperform against their SLA commitments.
- Require CDR data from your carriers that includes PDD and — where available — MOS per call for full visibility.