Uncertainty
Just as a predictive dialer has to cope with uncertainty in the real world, so too does Oceanic® generate uncertainty, in order to mimic the conditions encountered in a live situation.
Let's consider how Oceanic® does this by looking at a simple example:
Does this mean that we are guaranteed that for every sequence of 10 calls made, there will always be five live calls? Let's examine this notion by using a random number generator to make a sequence of calls.
Show example
In the table below, the random number values fluctuate between 0 and 1:
- Values < 0.5 are treated as no answers ('n')
- Values >= 0.5 are classified as live calls ('y')
| Random | Outcome | Random | Outcome | |||
|---|---|---|---|---|---|---|
| 1 | 0.55 | y | 26 | 0.66 | y | |
| 2 | 0.62 | y | 27 | 0.63 | y | |
| 3 | 0.81 | y | 28 | 0.05 | n | |
| 4 | 0.98 | y | 29 | 0.97 | y | |
| 5 | 0.61 | y | 30 | 0.68 | y | |
| 6 | 0.50 | y | 31 | 0.61 | y | |
| 7 | 0.73 | y | 32 | 0.81 | y | |
| 8 | 0.28 | n | 33 | 0.57 | y | |
| 9 | 0.98 | y | 34 | 0.36 | n | |
| 10 | 0.76 | y | 35 | 0.16 | n | |
| 11 | 0.20 | n | 36 | 0.29 | n | |
| 12 | 0.61 | y | 37 | 0.92 | y | |
| 13 | 0.88 | y | 38 | 0.91 | y | |
| 14 | 0.73 | y | 39 | 0.35 | n | |
| 15 | 0.49 | n | 40 | 0.40 | n | |
| 16 | 0.20 | n | 41 | 0.18 | n | |
| 17 | 0.89 | y | 42 | 0.73 | y | |
| 18 | 0.52 | y | 43 | 0.39 | n | |
| 19 | 0.13 | n | 44 | 0.77 | y | |
| 20 | 0.80 | y | 45 | 0.07 | n | |
| 21 | 0.41 | n | 46 | 0.60 | y | |
| 22 | 0.37 | n | 47 | 0.73 | y | |
| 23 | 0.44 | n | 48 | 0.25 | n | |
| 24 | 0.91 | y | 49 | 0.49 | n | |
| 25 | 0.66 | y | 50 | 0.28 | n |
So much for the idea that we are guaranteed that there will always be exactly five live calls every 10 calls. No problem for the call sequence from 19 to 28, but consider the calling sequences of 1 to 14 and 34 to 50. The live call rate is 85.7% in the first case, and 35.3% in the second case. The average live call rate for the whole table is closer to the expected average of 50%, but still adrift at only 40% overall.
Conclusion
Even if we do define stable live call and no answer ratios for a campaign to be run by Oceanic®, this still means that the actual call sequences being generated by it, at any point in time, may vary widely from such ratios.
For purposes of determining the optimum dialing rate, Oceanic® will know what the average level of no answers is, over recent calls, but it doesn't know how the randomizer is going to allocate call outcomes amongst the next sequence of calls. So in order to calculate an optimum dialing rate in the search process, Oceanic® has to explicitly allow for the impact of such swings
Changes in trends
The swings that a predictive dialer has to deal with in real life may be more difficult to manage, and the two predictive dialing controls we've included in Oceanic® will help you dampen performance down to reflect the actual conditions you may face, or are expecting to face.
Getting the best performance
As a general rule, the very best dialing performance is obtained when those responsible for compiling calling lists for individual campaigns ensure that the data is reasonably uniform. When data is not reasonably uniform, and when sudden and marked changes keep occurring (occasionally is tolerable), especially in live call/ no answer rates, the dialer may have to dampen the dialing rate, while it seeks to understand the changes, and wait times may rise.
Note
If you are not sure about the impact of uncertainty on call sequences, you can try running up your own set of random numbers to replicate the above example. For example, if you are using MS Excel, do the following:
- insert the RAND function in a cell, e.g. A1
- compute the event type by entering =if(a1<0.5,'n','y') in cell A2.
- copy A1 and A2 downwards to B1 and B2, C1 and C2 etc.