Answering Machine Detection Working

How Answering Machine Detection Works

Before understanding how answering machine detection works, we have to understand how answering machine works. Answering machines are devices connected to telephones which answers calls when no one is available. It determines whether there is a person to answer or not by the amount of time it lets the phone ring before it kicks in. Whenever it does kicks in, it would play a pre-recorded voice of the owner and allows for the recording of the message of the caller. It usually gives a hint (tone) to the caller before it starts recording.

There are three major variables in how an answering machine works that is very important for answering machine detection. This are the time a telephone rings before answering machine kicks in, the recorded voice played by the answering machine, and the tone or beep before the answering machine records the callers message.

First Level Answering Machine Detection
The first level in determining whether a call is answering machine or not is by calculating the time before an answering machine kicks in. This is usually between 4-5 rings or about 25-30 seconds. If a pbx system wants to detect answering machines, it should only let a call continue to ring for less than 25 seconds before disconnecting and tagging it as answering machine. This first level answering machine detection would of course fail if the answering machine was configured to answer calls automatically in a more lower number of rings or seconds. Again, you can compensate with this by lowering the number of seconds you would let a call ring before tagging it as answering machine, but this has a negative effect on your overall detection. If you lower to much the time you would let a call ring, say 10-15 seconds, a lot of calls wouldn't be given a chance to be answered by a real person. The most efficient time before tagging a call as answering machine is between 25-30 seconds. If you are still recieving a lot of answering machines, you could lower this to about 20 seconds.

Second Level Answering Machine Detection
The second level in answering machine detection is the pre-recorded voice played by the answering machine. If you would take a close look on how a live person answers a call, you would notice that 95 percent of live person would start the call with the word "hello" or "surname's residence" or "company name, hello", whatever it is, a live person on the other line would start the call with 1-3 words then waits for a reply. Answering machines however would have a pre-recorded voice with more than 3 words then ends with a beep or tone. This is how the second level answering machine detection works. It first determines the initial voice length of the caller which is about 1-2 seconds followed by the silence where the caller waits for the reply. The second level answering machine detection of course would fail if the called person has a noisy environment. For example if the called person is watching TV or listening to radio, the answering machine wouldn't detect the silent background and would interpret the noise it receives from the TV or radio as the callers voice. Second level answering machine detection is more delicate in the field of answering machine detection even though only a small fraction of calls would be detected wrongly because the calls it can detect wrongly are that of live ones. This would cause an increase in call abandon without the pbx system knowing it. Many countries already have strict guideliness in call abandonment and using second level answering machine detection might cause your system to do more harm than good. Also second level answering machine detection requires to analyze first the initial voice of the called party before it can pass the live calls to an agent. This would cause a delay of anywhere from 1-3 seconds before the the called party is transferred to the agent and can initially irate the called party. Also, if the called party has been recieving a lot of call center calls, chances are, it already knows that a delay in calls would mean the call is from a call center thus lowering the chances of conversions (sale).

Third Level Answering Machine Detection
Third level answering machine detection works by analyzing the initial response from the called party and determining if there is a beep or a tone. It tags a call as answering machine if it recognizes a beep or a tone from the initial response of the called party. Third level answering machine detection wouldn't fail as easily as the first and second level but it also requires analyzation of the initial response of the called party and again would cause a delay of anywhere from 1-3 seconds before the called party is transferred to the agent.

There may still be other ways to detect if a call is an answering machine or not but this three levels discussed here are mostly what predictive dialers or pbx system uses. If you are building a system with answering machine detection, you have to put in mind that accuracy and speed is more important than higher detection ratio. If your system is not sure if a call is an answering machine or not, always pass it to the agent and let him determine the call.
 
Back
Top