Other Examples

These examples give an illustration of the impact answering machines can have on a campaign. We have varied 3 of the campaign inputs:

1. percentage of call outcomes
2. detection percentage
3. ring time

Before you get carried away and think that you must have answering machine detection, check out the assumptions on the Call Outcome Properties page that generate the differences, and change them to suit your circumstances. The relevant assumptions are

• dialer detection percentage for answering machines
• answering machine detection times (both of them)
• agent wrap time

See also Answering Machine Detection.

These examples set higher and lower bands of 30% and zero respectively for busies.

In the case of predictive dialing and auto preview, there is no significant difference between talk and wait times, which is what you would expect, given the short dialing setup time and the fact that the agent is not involved in detection.

If you have a predictive dialer that can dial efficiently at high levels of no answers and answering machines, then you may get a pleasant surprise at just how tolerant your dialer is of being asked to continually retry these types of call outcomes. In the two examples, we start with call outcomes for no answers and answering machines combined of 40% in the first cycle.

• In the case of Predictive with 40 pct no ans and AM, and same recycling we keep the call outcomes for no answers and answering machines, when these call outcomes are retried, at the same level for all cycles.
• In the second case, with the campaign Predictive with 40 pct no ans and AM, and low recycling we have assumed that the retry percentages for no answers and answering machines will be substantially worse than this. But we have still continued to recycle these call outcomes for the same number of cycles, as for the other call outcomes; seven cycles in each case.

The surprising outcome, perhaps for some, is that when we inspect the Forecast by Cycle for each campaign, it is difficult to spot major differences between them. In the case of the campaign with the low recycling rates, a lot more calls are made, but the agent hours are shorter, because fewer people have been spoken to. On the results and profit front, there is effectively little to choose between the campaigns.

If you don't have an efficient predictive dialer, then you will need to be realistic about how much recycling you will do in practice on no answers and answering machines.

Try running both campaigns in say auto preview mode to see what happens then. And try too activating answering machine detection by the dialer.

For wide variations in the percentages of live calls, these examples show that you can expect wide variations in both talk and wait times.

You'll notice that the campaign time is a lot less in the case of 60% live calls.

Relative to the Base examples, no answer rates have been doubled for all methods. Agent talk times decrease and wait times go up in all cases.

Note that the agent talk time per hour differences between predictive and power dialing on the one hand, and the more basic forms of dialing, have increased relative to the example campaigns in the Base folder, which have a no answer rate of only 25%.

Increases in Not Ready time are reflected almost proportionately in changes in campaign lengths. If this percentage is set high to accommodate long breaks, then it will depress the average number of agents on a campaign at any one time, and this will push up wait times and lower average talk times.

See also Talk below.

These three campaigns take the Base example for power dialing and show the overdial rate at 1.5, 2 and 2.5 respectively.

Note how the numbers of standby agents increase.

These campaigns look at the performance of a predictive dialer when we vary the abandoned call target, starting at 1%, and then going up to 5% in 1% increments.

If you expect to make abandoned calls, but want to keep the level of them low, these campaigns give you an idea of the trade-off between levels of abandoned calls and talk and wait times. Try rerunning the same examples, but with a higher proportion of call outcomes devoted to no answers and answering machines, and see what the trade-off is then.

These examples take both the base and the high no answer campaigns for predictive dialing and reduce the overdial percentage below the maximum available, to reflect the conditions you may find in real life. For a well performing dialer, a 10% reduction is probably the maximum you need for campaigns with low levels for no answers and answering machines. For most dialers you will probably need to apply a bigger reduction for higher levels.

See especially

• Overdial Control
• Benchmarking Against Oceanic®
• Simulator versus Dialer 'Makeup'

These examples show a spectrum of predial interval times for the base progressive campaign. Small miscalculations in the predial interval can easily lead to quite high levels of abandoned calls.

The examples are a clear reminder of the inherent limitations of a predial only approach in a world where network delays are now largely a thing of the past. But if your operation is located in a territory where network delays are commonplace still, then predial may have some real advantages for you; but even then don't think of it as being a good substitute for a dialing method incorporating a good overdial algorithm. It's not.

See also Comparing Overdial and Predial.

These examples take the campaigns from the Base folder and for each dialing method first double and then halve talk times.

There are big differences in agent talk times, especially for those methods with high setup times. Some differences in wait times are quite marked as well.

Variations in the time taken to answer live calls can have a dramatic impact on campaign efficiency. If the average value of this parameter varies a lot over the course of a day, then you may want to take account of it in rostering your shifts.

The examples in this folder increase and decrease both the time to answer and the time to no answer, relative to the values used in the base examples.

Note that Oceanic® will expect any time to no answer you set to be at least 50% greater than corresponding time to answer.

These examples add 15 and then 30 seconds wrap (following talk) to the Base examples and give you an idea of the reduced talk time per hour you will get as a consequence. And wait times will go up as well, since the rate at which agents are taking new calls has gone down, suppressing the pace of the overdial algorithm.