In all the years I've been working on, in, for, with, or around outbound call centers, I can't think of a single subject that has been debated, praised, scorned, cursed, or misunderstood, more than "answering machine detection", or "AMD"...
Answering Machine Detection is the process by which high volume outbound dialers attempt to determine, in some automated fashion, when the phone call connects, if the party answering is indeed a human being, or if it is a voicemail. If the determination is a human being, the call is connected to an available agent. If the determination is a voicemail, the call is usually terminated...
As mentioned in our previous article, one of the behaviors anti-spam and anti-scam algorithms look for is "short duration calls". While every carrier and technology platform has its own definition of this, placing thousands of calls, and screening out the answering machines without leaving a voicemail - over, and over, and over again - creates a pattern of dialing outside phone numbers, connecting, and then immediately disconnecting, en-masse - clearly a behavior that any worthwhile anti-spam algorithm would frown upon...
A common argument I hear from many customers is that adding voicemails to their outbound calling process would add cost in the form of minutes, and resources in the form of outbound lines and "dialer waits" that have the potential to slow down raw dialing. Although these are fair points, every customer needs to weigh the cost of lower contact rates, vs the cost of those additional minutes, and find the balance that works best for them. For example, if you configure your platform to leave voicemails on every third attempt, then your call patterns will look 30% better to the algorithms...so on and so forth...
But, to me, by far the biggest reason to be leery of Answering Machine Detection is pure consumer experience...
In order to be effective, AMD needs to analyze the prospect's side of the call prior to handing the call off to an available agent. This analysis, no matter how you slice and dice it, takes time, and to the prospect that time sounds like dead air - like nobody's on the line...like a "call center". And the moment the prospect hears dead air, even for a moment, your odds of losing them skyrocket to almost 100%.
And if there's anything worse than not making contact with your prospect, it's making contact with your prospect, and blowing it...
No matter how fast your AMD is, never underestimate how incredibly sensitive the human ear is to micro-second lags, delays and pauses from automated systems, AI virtual agents, or even real life human beings. The moment they are given a reason to do so, the prospect will hang up and bail on the call. And to me, this outcome is a most tragic one in today's outbound calling environment...
The last piece I want to touch on, not entirely AMD related, but close, is your dial timeout - how long, or for how many rings, you let the prospect's phone ring, before you decide to mark it a no answer...
My advice here is simple - from a PURE contact rate perspective, wait as long as your model can tolerate...not because there's a tremendous reward in live person answer rates at the tail end of the timeout, of course, but because that is exactly what an outbound caller trying to communicate with a high intent prospect, would do - they would wait...and the algorithms know it...
Next up - we will dissect dial and retry strategy, itself, and give some best practices and guidance for how often and in what manner you should attempt to contact your prospects...although every data set, contact center, and business model calls for it's own unique approach, there are some common fundamental guidelines we share with all of our clients, and in our next article, we will share them with you... :-)