So the world hates telemarketers. They are calling calling calling you. It seems like they have this magic mantra to call you constantly and interrupt your dinnertime. And you think: Why is it so easy for these "bad" people to call me all the time? Why don’t we make it tougher for them to reach me? Even illegal.
Newsflash: Reaching people for phone solicitation is one of the toughest jobs around. There are of course the usual hurdles like National Do-Not-Call registry, new rules churned out by the government restricting telemarketing calls and stiff fines for errant calls. But the real challenge is simply to, like I said, reach people.
Expanding customer base and retaining the existing customers has become central to any Contact Center operation. To do this, a Contact Center needs to proactively reach out to its customers (current or future) for: telemarketing, customer communication, callbacks or even debt collection. In order to effectively do this, companies use Dialers to make calls automatically based on pre-configured parameters. But the actual connect-rate to a live person is pretty low. On an average, for every 100 calls dialed, they only reach about 10 people. Sometimes this can be as low as 1%. This means that a lot of calls need to be placed simultaneously, then processed (hopefully without human intervention) to detect a live person, and only the live person calls should be passed back to the Contact Center agents.
Once the call is placed, Call Progress Analysis (CPA) is performed to determine if the call was answered by a live person and should be connected to an agent. If Call Progress Analysis determines that an answering machine answered then an automated voicemail message could be left. If CPA detects a busy tone or network message then it can be programmed to categorize the call accordingly and perform other actions. It sounds easy enough to detect these, but it takes a lot for a machine to do this, and to do it well.
There are many solutions that provide CPA, based on different algorithms, but most cannot provide a high level of accuracy in the results, or they take a long time to return the results. Inaccuracy and call latency lead to lower overall productivity and higher wastage of agent times.
- Colored Ringback Tones: One basic trick of old-time CPA solutions was to listen for the standard ringback and perform analysis accordingly. Colored Ringback tones (CRBT) means you can greet your callers with your own choice of music or greeting instead of the standard ringback. This means that the CPA now has to be intelligent enough to recognize the music as a ringback and not as voice or answering machine. An intelligent CPA should be able to train itself and recognize a CRBT from a live speaker or answering machine.
- Hello…..<pause>….: People will often leave a long pause after the "hello" on their answering machines. This can often mislead the CPA to believe that it is a live speaker. A Good CPA solution should be robust enough to detect this without increasing the overall latency.
- Network messages: Some networks play a SIT tone (easily detectable by CPA) before they play their messages about number being out of reach, or number disconnected etc. But a lot of networks do not play the SIT tone any longer. That means the CPA needs intelligence to differentiate a network message from a live speaker or answering machine.
- Background noise: Isolating the background noise from the relevant audio is crucial to increasing accuracy and decreasing latency. Robustness to background noise has been a big challenge for CPA solutions till now, but new algorithms are now being developed that use special speech patterns to eliminate background noise and perform analysis on only the relevant audio.
So, while some work on coming up with Intelligent CPA, I propose this: A National Please-Call-Me list. By placing yourself in this list, you swear to always put down your fork and pick-up your phone when that poor agent calls you.
Share your ideas and experiences regarding the accuracy of Call Progress Analysis, especially in a SIP environment.