The Hidden Cost of VICIdial AMD: Extra Time on Every Call

Answering machine detection (AMD (answering machine detection)) is usually sold as a way to keep agents on live people. That part is true. What gets left out is the bill: AMD adds processing time to every answered call, not only the ones that turn out to be machines. Before you switch it on for a whole campaign, it helps to see exactly where that time goes and whether you actually need it.

Where the extra time comes from

You turn AMD on for a campaign by setting the Campaign VDAD exten field to 8369, the AMD-enabled call processor. When a call hits that extension, the dialplan (Dialplan) runs a short sequence before anyone is connected: it plays a slice of silence, logs the call, runs the Asterisk AMD analysis, runs the VD_amd routing script, then hands the call to the outbound routing agent. Every one of those steps costs a fraction of a second, and the analysis step in particular has to listen to the line long enough to decide what answered.

That listening window is the catch. The analyzer cannot know it has a human until it has heard enough audio to rule out a machine. So even a real person who picks up and says hello sits through the same detection pass that a voicemail greeting does. Multiply that delay across thousands of answered calls and you have measurably slower connects and a touch more dead air before the agent says a word.

flowchart TD
  A[Customer answers] --> B[8369 plays brief silence]
  B --> C[Log the call]
  C --> D[Run AMD analysis]
  D --> E[Run routing script]
  E --> F{Human or machine}
  F -->|Human| G[Connect to agent]
  F -->|Machine| H[Hang up or play message]

Why the cost is paid by everyone

It is tempting to think the time only matters on machine calls, since those are the ones you divert. It does not work that way. Detection has to finish before routing can happen, so the live human pays the same toll. On a busy Predictive dialing campaign, that small per-call delay also nudges your pacing math, because the dialer is holding calls slightly longer between answer and agent connect.

There is a second cost that is easy to miss: tuning. AMD is not perfect and it is sensitive to your leads and your telecom lines. A setup tuned for clean landlines will throw an AMD false positive on noisy mobile lines, dropping real customers as if they were machines. Fixing that takes attention you might rather spend elsewhere.

The leaner alternative

For many campaigns you can skip AMD entirely. Set the Dial timeout to somewhere in the 22 to 26 second range. Most answering machines pick up inside a predictable window, so a tighter ring time quietly dodges a large share of them without analyzing a single call. The trade-off comparison is laid out in our note on AMD versus Dial Timeout. One caveat: USA FTC Safe-Harbor rules require at least 15 seconds of ring time on B2C calls, so do not set the timeout too low.

AMD earns its keep when the value of an extra live connection clearly beats the per-call time you spend finding it, or when you genuinely need to drop voicemail on machines. For everything else, weigh the cost first. Remember too that AMD only runs under auto-dial methods like RATIO or ADAPT; it does nothing in MANUAL or INBOUND_MAN Dial method modes, so turning it on there just wastes the field.

Decide with the full picture

If you want the complete walkthrough of how detection, routing, and message drops fit together, read our AMD and CPD complete guide, and if you are still deciding whether AMD belongs on your campaign at all, start with when to use VICIdial AMD.

Not sure your servers are sized for the extra processing AMD demands across a full campaign? We tune and host VICIdial so the math works out for you. Take a look at VICIfast pricing and pick the plan that fits your call volume.