You are here: PositiveIndians Blog -> Category: Call-Centres

Disadvantages of Predictive Dialers

By Abhijit Phadke

While predictive dialers are a proven technology for boosting outbound call center productivity, there are a few drawbacks and pitfalls to it as well. In this article we take a look at the disadvantages of predictive dialers.

The main working principle behind a predictive dialer software is its algorithms that allow it to mathematically calculate and determine, when an agent will become free to take the next call and when the next dialed call is most likely to be connected. Based on this calculation the dialer initiates the process of dialing the next numbers on the list while all the agents are still busy speaking on their earlier calls.

Prediction failure cases

The drawback here is that what if this prediction algorithm fails? After all its only a prediction, which cannot be 100% accurate. In such cases two things can happen.

1. New call dialed is picked up, but none of the agents is yet free to take the call. In such cases the person at the other end of the line can get irritated because of the 'blank call' and hang up the phone.

2. If a deliberate lag is introduced in the dialer, such that it will wait few extra seconds to ensure 100% agent availability, then this again defeats the very purpose of having a dialer which is to make calls at the fastest possible speed.

This is one of the disadvantages of a predictive dialer, though its not very severe, since these days the prediction algorithms are fairly robust.

Minimum number of Agents

The next point to consider is the number of agents making the calls.

This is important because prediction algorithms work well only when the number of agents is high. The general consensus amongst those in the industry is that the minimum number of agents should be at least 10 or 12, for the predictive dialer to work effectively. So for smaller call centers who have only a few people making calls, predictive dialers do not provide much of an advantage.

AMD (Answering machine detection) issues

Predictive dialer do the task of auto-dialing numbers and when the calls are picked up by a person, these dialers immediately connect a free agent to the call, who can speak to the person dialed.

For doing this, the predictive dialers listen for a response from the other end, for every number that they dial. In cases of phone not picked up, or busy tones, the software is usually good enough to catch and interpret the signals accurately, however in cases where the call goes to an answering machine things are not so simple.

The dialer software has to then rely on basic voice recognition techniques to identify whether its a human response or simply an answering machine. Techniques such as listening for the timing of the 'pause' (e.g. 'Hello'...pause) are used by the dialer for this purpose. The problem here is that this voice recognition technology is not 100% accurate, and the dialer can mistake an answering machine for a human being and vice versa.

If an answering machine is mistaken for a human being, the call will be handed over to the agent (known as a 'false positive' case), and it will be an unproductive waste of time, since the agents cannot do anything when they are faced with a machine response.

On the other hand, when the voice of a human being is mistakenly thought of as an answering machine, the dialer will simply disconnect the call, instead of handing over the call to an agent. (sometimes recorded messages are also placed by the dialer in such cases). In any case, the call will not be handed over the agent, and the purpose of making an outbound call fails. Also the person whose number has been dialed will get irritated because of receiving a blank call.

This is considered to be the most significant disadvantage of a predictive dialer, and hence to overcome this problem to some exent at least, several dialer vendors allow for the AMD option to be switched off altogether, so that all picked up calls (human or machine) are handed over to the agents.

The AMD detection failure cases described above, also result in an increase in abandoned call rates which have to be maintained below stipulated levels.