ASR, ACD, PDD and NER: Voice Quality Metrics Explained

The four metrics at a glance

ASR: answer-seizure ratio

ASR is the share of call attempts that were answered. If you send 1,000 attempts and 420 are answered, ASR is 42%. It is the metric buyers watch most, and the one most often blamed on the wrong party.

ASR depends on two things at once: whether the route can complete calls, and whether the people you call want to pick up. A predictive dialer calling an aged lead list will see low ASR on the best route in the world. A service callback to customers who just requested it will see high ASR on a mediocre one. That is why ASR should always be compared for the same traffic on different routes, never across campaigns.

• Agent-dialled business calls: often 45–60%.
• Consumer outbound from a dialer: commonly 20–35%, heavily list-dependent.
• IVR / press-1 reminders to opted-in customers: typically 35–55%.

These are rough guidance ranges from our own monitoring, not promises. Each country page, for example France or Brazil, lists the ranges we see on healthy routes for that destination.

ACD: average call duration

ACD is the mean length of answered calls, in minutes or seconds. It is calculated only over answered calls, so a campaign with many unanswered attempts can still have a long ACD.

ACD is the best early warning of false answer supervision, where a carrier in the path answers the call itself (playing ringback or a recording) so the call becomes billable. False answers produce a cluster of very short calls and drag ACD down while artificially lifting ASR. If ASR rises and ACD falls on the same destination at the same time, investigate immediately. Our FAS guide explains how.

PDD: post-dial delay

PDD is the time between sending the INVITE and receiving the first indication that the far end is alerting (a 180 Ringing or 183 Session Progress with media), or an answer. Most buyers measure to first 18x; be consistent so values are comparable.

PDD rises with every carrier hop and every failed attempt inside the route. A provider that tries three partners in turn before one accepts the call will show long PDD even if the call eventually connects. Long PDD matters most for dialers: callers hang up, agents sit idle, and abandoned-call ratios go up. As guidance, 2–5 seconds is healthy for most international mobile destinations, and anything consistently above 8 seconds is worth raising with your carrier.

NER: network effectiveness ratio

NER was defined by the ITU (E.425) to isolate the network from user behaviour. It counts a call as successful if the network delivered it to the called party, whether they answered, were busy, did not answer, or rejected it. Only network failures, such as congestion, routing errors or timeouts, count against it.

In SIP terms, a practical mapping is: answers (200 OK), 486 Busy Here, 480 Temporarily Unavailable when caused by no answer, 603 Decline and cancellations after ringing are user outcomes. 503, 502, 500, 408 and 404 on valid numbers are network outcomes. See SIP Response Codes for Dialer Operators for the full list.

A healthy route usually shows NER above 95%. If ASR drops but NER stays high, the route is fine and the change is in your traffic. If both drop together, the route is the problem.

Reading the four metrics together

Always compare like with like: the same destination, the same hour of day, and a meaningful sample. A few hundred attempts is the minimum before drawing conclusions; for busy dialer destinations, use at least an hour of traffic.

A worked example

Suppose a campaign sends 2,000 attempts to German mobiles in one hour. The results are: 640 answered, 380 busy, 520 no answer, 60 rejected by the user, 300 failed with 503, and 100 failed with 404 on numbers that turn out to be disconnected. Total billed time is 1,856 minutes.

• ASR = 640 ÷ 2,000 = 32%.
• ACD = 1,856 ÷ 640 = 2.9 minutes.
• NER = (640 + 380 + 520 + 60) ÷ 2,000 = 80%, or 84% if you exclude the 100 disconnected numbers as list errors.

An ASR of 32% looks unremarkable for consumer traffic, but an NER of 80–84% is poor: 300 attempts, 15% of the hour, failed inside the network. The 503s are the issue to raise with the carrier, after checking that your own CPS was within limits. Without NER you might have blamed the lead list. See the country guidance for Germany CC routes for typical ranges.

Sample size and time of day

Small samples mislead. With 100 attempts, an ASR of 30% has a margin of error of roughly ±9 percentage points; with 1,000 attempts it is about ±3. Compare routes on at least several hundred attempts each, during the same hours, to the same number types. Mobile and fixed destinations behave very differently, and so do mornings and evenings: answer rates on consumer mobiles often peak in the early evening, while business lines are best mid-morning.

Building a useful dashboard

Show the four metrics per destination and number type on one screen, hour by hour, with the previous week's line for comparison. Add attempts per hour so you know when a sample is too small to trust, and an alert for any destination where NER drops by more than five points in an hour.

Calculating the metrics from your CDRs

Most dialers and PBXs export CDRs with a disposition and a billed duration. The query below works on a typical Asterisk CDR table and gives you ASR, ACD and an approximate NER per destination prefix. Our ASR / ACD / NER calculator does the same from totals you paste in.

SELECT
  LEFT(dst, 4)                                   AS prefix,
  COUNT(*)                                       AS attempts,
  ROUND(100.0 * SUM(disposition = 'ANSWERED') / COUNT(*), 1) AS asr_pct,
  ROUND(SUM(billsec) / NULLIF(SUM(disposition = 'ANSWERED'), 0) / 60, 2) AS acd_min,
  ROUND(100.0 * SUM(disposition IN ('ANSWERED','BUSY','NO ANSWER')) / COUNT(*), 1) AS ner_pct
FROM cdr
WHERE calldate >= NOW() - INTERVAL 1 DAY
GROUP BY prefix
HAVING attempts >= 200
ORDER BY attempts DESC;

In the CLIRoutes portal these metrics are calculated live per destination and per route, so you can see a change within minutes rather than at the end of the day.