If you are going to get the best out of Oceanic®, you will want to know what the performance of a (your) dialer is relative to what Oceanic® does. Let's begin by looking at how both Oceanic® and a real dialer cope with the flow of events on a campaign.
With evenly balanced outcomes
Since variations in call outcomes are the major source of differences in overdialing levels, we will use the same example as in Uncertainty where 50% of all calls are live calls, and 50% no answers.
| Oceanic® | Predictive dialer | |
|---|---|---|
| How is event information measured? | ||
| It gets defined by you in the Campaign Wizard. Oceanic® knows that the probability of the next call being a live one is 0.5. | The dialer will be tracking call outcomes, working back from the most recent call attempt, on either a timed or call volume basis, to build up a picture of live call and no answer rates. | |
| How does prediction work? | ||
| Oceanic® knows the probability of the next call being a live one, and also the likelihood that some call sequences may be all live calls, or all no answers. See also Uncertainty. | The actual prediction method will vary from one dialer to another, but the essence of it, is that a dialer will use the data it gathers to calculate the likelihood of live calls or no answers over the sequence of calls following. Because of actual variations in call outcomes a dialer can estimate the probability of the next call being say a live one, but unlike Oceanic® it doesn't know this probability with certainty. | |
| How efficient is the prediction of call outcomes? | ||
| As efficient as is possible, given the natural uncertainty inherent in call outcomes. | Provided that the call outcome data is reasonably uniform, even though a dialer won't actually know the distribution of future call outcomes in the sense that Oceanic® does, it should still be able to track any movements in the live call and no answer rates, virtually as well as Oceanic®. For example, as an afternoon shift approaches the meal hour, chances are that the live call rate will go up. But this is a gradual change and one that a dialer should cope with easily. | |
With erratic outcomes
What if the data is not fairly uniform, what then?
In real life, fluctuations in live call and no answer rates might be both greater and more sudden than we have been considering, forcing a dialer to be more conservative in its overdial algorithms, in order to avoid unwanted abandoned calls. A dialer will try and respond to such fluctuations, by continually updating its predictions. But consider the following:
Imagine that you are targeting a local community during the day, and it's snowing. The business people and their families have all managed to get to work, school and so on, but the retired people have all stayed in. Sounds reasonable? OK. Now let's assume that the retired people live mainly in small groups and apartment blocks within the community. If your calling list was compiled by street address, and left unsorted, then you can spot what's going to happen. Overall you might have 50% live calls, but in fact you will have swings between zero response and 100% answers. Tough stuff for a dialer to handle. If you allow this kind of bias to exist in your calling lists, then your dialing performance will suffer. Best practice is always to sort calling lists to reduce the likelihood of major swings in the profile of call outcomes
Summary
Which brings us to a summary of the two possible differences between Oceanic® and what predictive dialers do in practice.
| Difference | Oceanic® | Predictive dialer |
|---|---|---|
| Quality of data | All campaign data is defined by you, and although it is subject to variance, Oceanic® knows how to measure and allow for this. | All campaign data has to be tracked by the dialer and used to forecast future values. Most dialers should cope well in doing this, provided that the campaign data doesn't contain the kinds of distortion described above for call outcomes. |
| How does prediction work? | Given the natural uncertainties in the data you provide it with, the Virtual Event Machine (VEM®) uses simulation techniques to find the best dialing performance possible. | There will be some great dialers around, and there will be some that are not quite so great. If you scout around enough, you may get some clues as to how the algorithms for the XYZ company's dialer work, but it's the wrong question to ask, since it is unlikely that you'll get the lowdown on all their secrets; consider instead some of the techniques suggested below. |
Oceanic® provides you with two controls for bridging the gap between its performance and what you can expect from a predictive dialer. See Overdial Setting in Oceanic® . We are going to suggest two ways in which you can use one of these controls, the Overdial Control, to benchmark a dialer. See Benchmarking Against Oceanic®.