The Future Of Network Answering Machine Detection

Answering Machine Detection (AMD) remains a hot topic in the UK, where the issue of AMD false positives is an unwelcome distraction. It relies upon cadence-based AMD and as we have pointed out in a previous blog, this is simply not viable under compliant dialing conditions.

Is there an alternative? Let’s consider the calling landscape for a moment. When calling to consumers (landline/ mobile), of those calls where an answering machine is used –

• 15% are on-premise hardware answering machine
• 85% are carrier voicemail

For the home answering machine, the choice is to use cadence detection, or don’t do AMD at all. For carrier voicemail, yes, there is an alternative – byte pattern recognition. (See our previous blog for details.) But as we have also stated, there are several reasons why this is not viable:

1. User custom message
   All carrier voicemail offers the possibility of user customisation. Even though the majority of users leave the default carrier message, there will always be a proportion who will customise. These custom messages can never be pattern-matched.
2. Carrier custom message
   Smaller carriers in the market piggy back onto large networks but do their own custom voicemail message. There are potentially infinite number to match against.
3. Legacy equipment
   Older equipment, in converting digital signalling to analog, sometimes several times, can corrupt and distort any message so that it no longer matches

Is there an alternative to the alternative?? Yes! Or rather, almost! Network-based detection of answering machines (or, more accurately, voicemail) is the very bright hope on the horizon.

Since we published this blog stating that Network AMD does not exist, we along with several other major vendors, have been working to bring it into existence. Sytel has been taking the technical lead on a working party to define and establish standards for Network AMD. Read more about the Network AMD Working Party.

What is the NAMDWP proposing?

At the moment, what happens is this:

1. A call is placed to the carrier
2. The carrier tries to connect to the called party but finds he is unavailable
3. The carrier diverts to voicemail
4. The carrier returns an ISDN ‘Connect’ signal (so it can charge for the call)

The proposal being put forward by the NAMDWP is this:

The above process is repeated, but at no. 4, the carrier returns an ISDN ‘Network AMD’ indicator, followed by ISDN ‘Connect’.

The good news is that there is no bad news! It produces winners all round.

• For call centers
  Dialers would have the option to drop the call after the NAMD signal. This would mean not having to tie up trunk for 3 secs, which over the course of a day, results in a significant performance boost. Also, customers who have not been kept waiting for 3 secs are more inclined toward a positive conversation.
• For customers/ Ofcom
  No more waiting with dead air for the AMD to do its thing, greatly reduces the number of nuisance calls produced and therefore reduces complaints to Ofcom.
• For carriers
  The standard equipment operated by the vast majority of carriers throughout Europe is quite capable of following these recommendations with minimal outlay. Carriers will still be able to charge for the same number of calls but with 20% less load on the network, as the transaction happens at ISDN signal level, not audio. Carriers may even be able to levy additional charges.

Successful implementation of these standards may be followed by a proposal to outlaw cadence detection for the 15%. Both developments can’t come a moment too soon in our view, and should be welcomed by anyone looking for healthier outbound markets – not just in the UK but everywhere.