How does answering machine detection (AMD) work?

We explore how computers can detect voicemails and answering machines during outbound dialling and whether you should enable this technology on your campaigns.

Date: 25th July 2019


One of the biggest factors you have to battle against when outbound dialling is the contact rate of your dialling lists. As most communication moves from land lines to mobile communication, and with Caller ID now near universal, it has become more acceptable in society to ignore calls than it was when you had a land line and didn’t know who was calling.

This has meant for those trying to contact their customers, the volume of failed connects and voicemails has increased significantly. Whilst any dialler can easily filter out calls which are disconnected or unanswered, voicemails pose a unique challenge.


To explain how voicemail detection works, let’s look at the typical difference between a human answering the phone and a machine:


"Hello! Sarah speaking?"


"This is the voicemail service for..."


So typically a human answers the phone with a brief salutation, and then goes silent waiting for you to respond.

Whereas a voicemail or answering machine would tend to provide a much longer salutation. This is the way near enough every answer machine detection works - by learning the difference in length of salutation between humans and machines. They may add fancy words such as Artificial Intelligence or Neural Networks to their marketing packs, but you can be pretty confident underneath the marketing is the same concept.

Now knowing how it works, we must understand the inherent problem with the technology.



For us to understand the length of the salutation and to filter out the voicemails, we must listen to the first few seconds of the call. The longer the period we can listen, the more accurate the computer can be in it’s prediction, but for live customer calls they simply aren’t going to stay on the line listening to silence for more than a couple seconds. Further both the FCC in the US and Ofcom in the UK dictate that all calls must be connected to a live agent within 3 seconds.

Secondly, no prediction is 100% accurate. Whilst the prediction would be right in the vast majority of cases, nothing can be guaranteed which presents a potential conflict with the FCC & Ofcom rules in regards to never ending a call to a live human without an agent handling it or a message being played. Ofcom have even provided specific guidance in this regard that in their opinion that using voicemail detection would be incompatible with the rules on silent calls.

Finally, when using voicemail detection it is important for you to train your contact centre staff on how the technology works. They must be aware that using detection means you don’t hear the first few seconds of a call, and therefore they don’t hear the salutation.



As you can see, to give a good experience to your customers it’s important your agents start their pitch immediately on connection of the call to them.

When clients ask us about voicemail detection, we confirm we have the functionality built into the product. However we also make sure our clients understand the limitations of technology. We would always advise customers to avoid using the technology.

We can work with customers who are experiencing high levels of voicemails, and using the powerful data management functionality built into our product, can help naturally reduce the amount of calls hitting voicemail rather than trying to filter them out once it’s too late. It provides a better experience to your customers, a better experience to your staff and ensures you remain compliant with the law.




Ready to get started?
Get in touch or ask for a free demo today

Free Demo
Contact Us