Sober Thoughts On Predial

Be sure to read this topic before you spend any time on dialing methods which involve predial.

In the first release of Oceanic® we looked in depth at the different kinds of dialing algorithms that you may encounter in the more sophisticated dialing methods, especially predictive, and we defined two main types:

The key to good predictive dialing is an overdial algorithm that takes account of all campaign parameters. Development of good overdial algorithms is non-trivial, whereas it is quite straight forward to develop predial algorithms. For this reason, predial algorithms are quite common, and hence we devoted a lot of help file space to them in Oceanic®.

The effectiveness of a predial algorithm is dependent on the ability to predict when an agent will complete a call and be free to take another one. The Extended Properties page in the Oceanic® wizard allows you to simulate predial activity, either on its own, or in combination with an overdial algorithm.

Since the first Oceanic® release, we have analyzed the call completion phases for agents (both talk alone, and also talk followed by wrap) in several call centers. The results show in all cases that it is notoriously difficult, on any basis, to predict when an agent will finish a call, with the accuracy required for effective predial activity (especially if the aim is to avoid abandoned calls). Here's why:

Agents themselves are poor predictors, and in any case should not be asked to indicate when they expect to be free, since it diverts them from the task of completing a call.
Using a position in a script to indicate when an agent will be free is too restrictive in terms of how agents complete calls.
The method we thought might have some merit is that of tracking agent wrap times and using these to predict when an agent will become free. The idea is that wrap times can be tracked for each agent and on the basis of their distribution, a determination can be made as to when to predial. It's a good idea in principle, but in practice is flawed in that it requires both the average wrap times to be well known for each agent, and also for the distribution of wrap times to have a low variance, and this in our experience is rarely the case. If you doubt this, spend a little time, as we have done, watching how agents complete their calls on a range of different campaign types.

So our advice is beware the benefits of predial, unless you have firsthand experience of them, or, unlike most, you have found a way of really controlling the call completions of your agents.

Note

You occasionally see comments to the effect that a predial algorithm can be based on consideration of talk times only, when there is no wrap; i.e. as soon as the agent has finished talking, he is ready to take another call. Any time you are predisposed to believe that, then our advice is think again. The same arguments apply as for the third bullet point above, but with the added complication that talk times for individual agents are even less predictable than wrap times.