I am a computer programmer. If you ask me a technical question, be prepared for a very complicated answer. Occasionally, I get requests for a white paper about our predictive algorithms. While working on this post, I attempted to explain some finer points of the predictive algorithm calculations to one of our marketing team members.
The resulting blank stare was not very encouraging. I think I’ll put the white paper idea on hold and stick with a (hopefully) less complicated post.
Predictive Algorithms are what keep your agents talking. Besides being complicated formulas and calculations, predictive algorithms are essential to your agents’ productivity because they keep your agents busy talking without putting many live contacts on hold.
We’ve done research on how long the typical live contact will remain on hold. The average hold time was 11 seconds, but most people hung up much sooner.
At IAT, we have three goals when developing our predictive algorithms:
1. Never make a nuisance call
2.Never put a person on hold
3. Always keep the agent ready times as short as possible.
We have been tweaking our predictive algorithms for the past 25 years To date, the shortest average ready time reported was 10 seconds (using a regular bad debt list with low hit rates). That means agents are only waiting an average of 10 seconds between wrap up and the next call – hardly enough time to catch your breath! Meanwhile, the dialer is wading through all the answering machines, no answers and bad numbers.
IAT’s Predictive Algorithm is complex, but very effective. IAT’s predictive algorithms aren’t just a single calculation, but are comprised of three components:
1. Habits of the agent(s)
When an agent signs on and begins taking calls, the dialer collects a history of their habits (talk time, wrap up time, etc). This information is used to determine when the agent will become available for the next call.
If multiple people sign onto the same list, the dialer will calculate needed calls for the whole group, but it still bases those calculations on the habits of each individual agent.
2. Manager-established dialer pacing
A manager can decide how aggressively he wants the dialer to make calls. The more the pacing is increased, the shorter agent wait times are, but the more people are likely to be placed on hold.
3. Hit Rate/Transfer Rate
The dialer keeps track of both the Hit Rate (the amount of calls the dialer believes will be transferred to the agents) and the Transfer Rate (how many calls are actually transferred to the agents).
The dialer regularly adjusts the algorithm for the most recent call results/experiences. Sometimes I wonder if we have the dialer make too frequent adjustments, but without the repetition, the dialer isn’t as fine-tuned for the specific list, time of day and group of agents.
And that also means it wouldn’t be as successful at keeping agents busy collecting money.
Predictive algorithms: Sometimes difficult to understand, but they keep your agents talking and collecting, instead of waiting.
This article was written by Randy