The average length of answered calls over a period, used to gauge route quality and to spot traffic patterns associated with robocalls and fraud.
ACD is calculated by dividing the total duration of answered calls by the number of answered calls. Like ASR, it is a standard wholesale metric: a sudden drop in ACD on a route can indicate poor audio quality or false answer supervision, where calls are marked answered when they are not.
In fraud detection, ACD tells a story when combined with other signals. Very short average durations with high volumes suggest robocalls that consumers hang up on. Very long, uniform durations to a single destination can point to traffic pumping or revenue share fraud, where calls are deliberately held open.
Looking at the distribution of call durations, not just the average, is often more revealing, because fraud tends to produce unnaturally regular patterns. For example, a large cluster of calls lasting almost exactly the same number of seconds is rarely the result of real conversations and deserves a closer look.
Why it matters to carriers
ACD is easy to compute from CDRs and changes quickly when a customer's traffic changes character. Tracking it per customer and per destination gives you an early warning of robocalling, IRSF, and pumping schemes, as well as quality problems on routes you are paying for.