Comparing Overdial and Predial
When both algorithms are used on the same campaign, how do we evaluate the contribution of one compared with the other? And should the application of one algorithm take precedence over the other? This topic compares the two algorithms, and in Combining Overdial and Predial we look at what happens when they are combined.
1) Campaigns with very low levels of no answers and answering machines
Under these circumstances, you may find that a predial algorithm will be more effective than an overdial one. If you want to check this out, here's what you do:
- On the Outbound Selection page, select the Cookbook method
- On the Call Outcome Properties page, set up your campaign with no answer/ answering machine rates of say 10% or less
- On the Extended Properties page, put all your agents into the predial state only, and fill in the time values
- Run the campaign, varying the Predial time values (see #3) until you are happy with the balance between the abandoned call level and average wait time. Make a note of both values. You will need to iterate a few times to do this, but it should take you less than a minute.
- Then on the Extended Properties page,
- move all the agents into overdial mode
- insert the abandoned call level as your target
- select predictive mode
- Run the campaign and then compare the average wait time with that obtained from the previous run.
Assessment 1
Compared with overdial, predial can give lower average wait times between calls by as much as several seconds. Under these circumstances, if you have a dialer with a good overdial algorithm (since you will need it for campaigns with higher no answers and answering machines), stick with it, and if you are absolutely ace at getting your predial interval values right, make sure your dialer has a predial algorithm that can be turned on as well.
If you have worked your way through the example doing a number of iterations with the predial interval, you'll have noticed that at these low no answer/ answering machine levels, the abandoned call percentage is very sensitive to movements of as little as a second in the predial interval.
So proceed with care. If your guesstimate of the average predial interval is likely to be in error by up to a couple of seconds, then halve the predial interval figure that you get from the example above, and use that. If you think that your error could be greater than this, then think twice about using predial at all, or you may get unwanted abandoned calls.
2) Campaigns with medium to high levels for no answers and answering machines
Repeat the above exercise, gradually increasing the no answer/ answering machine rates.
Assessment 2
You will find that predictive dialing shows increasingly better relative performance, the higher the no answer/ answering machine rates.
A good overdial algorithm will be measuring everything that moves, even including the times to answer. Thus, even at a high no answer level, it will often be able to keep both wait times and the level of abandoned calls down to acceptable levels.
On the other hand, in practice most predial algorithms are simple. They will take little, or no account, of movements in parameters such as no answer rates, and at anything beyond single figures, the predial interval figure is likely to be very difficult to estimate with any degree of precision. So, if you want the most productive dialing method at higher no answer levels, our advice is to choose a product with a good overdial algorithm.
If you fine tune your overdial algorithm with predial under these circumstances (by putting all agents into the agent dialing state marked overdial and predial), you'll find that for whatever combination of inputs you have chosen, you may get 1 - 2 seconds improvement at most in wait times, beyond what overdial alone provides.
Predial and hybrid algorithms
Even with good scripting systems it is very difficult to estimate the average predial interval precisely, and in practice you will find few vendors, if any, willing to recommend that you set it at anything but single figures. If there are exceptions, it will be vendors offering hybrid dialing algorithms, taking say a predial algorithm, and adding to it allowance for other parameters, such as the no answer rate.
Any hybrid, based on predial won't be as good as the choice of overdial algorithms offered by predictive vendors, since these should be taking account of all significant dialing parameters. Alternatively, a hybrid might resemble one of the forms of power dialing, in which case you might want to select one of the power dialing settings, and combine it with the predial interval, to get the dialing combination you want.
We don't know of anyone producing hybrid algorithms, but we would be happy to comment on any definitions, and incorporate them in Oceanic® if demand warrants.