Utilizing predictive models to improve predictive dialer pacing capabilities
Various embodiments of the invention provide methods, systems, and computer-program products for pacing outbound calls placed by a predictive dialer in a contact center. Specifically, an ensemble made up of a global predictive model and a local predictive model is applied to each dialing record found in a plurality of dialing records to provide a probability of an outbound call placed to the dialing record resulting in a live connect. Accordingly, a call pacing hit ratio can then be calculated based on the probability for each of the dialing records and this call pacing hit ratio can be used by a predictive dialer in various embodiments to more accurately pace the placing of outbound calls then by using conventionally derived call pacing hit ratios.
1. A method for pacing outbound calls placed by a predictive dialer in a contact center, comprising the steps of:
calculating a call pacing hit ratio based on a probability for each dialing record in a plurality of dialing records of an outbound call placed to the dialing record resulting in a live connect, wherein the call pacing hit ratio represents a percentage of outbound calls expected to result in a live connect and the probability for each dialing record is determined by applying an ensemble to the dialing record comprising a global predictive model representing comprehensive dialing trends across at least one of multiple industries, multiple purposes, multiple locations, and multiple predictive dialers and a local predictive model representing a specific dialing history associated with the contact center;
determining a number of outbound calls to place by the predictive dialer based on the call pacing hit ratio and at least one of a number of agents associated with the contact center that are currently available to handle calls and a number of agents associated with the contact center that are expected to become available to handle calls within a time period;
selecting a set of dialing records from the plurality of dialing records by the predictive dialer based on the number of outbound calls to place; and
placing virtually simultaneous outbound calls by the predictive dialer for each of the dialing records found in the set of dialing records.
2. The method of claim 1 , wherein determining the number of outbound calls to place by the predictive dialer is also based on a set target comprising at least one of: maintaining an abandonment rate; maintaining a time limit on how long a party can remain on hold before being connected with an agent; and maintaining a maximum wait time an agent can wait between calls.
3. The method of claim 1 , wherein the ensemble comprising the global predictive model and the local predictive model is generated using one of a technique of bagging, boosting, or stacking.
4. The method of claim 3 , wherein each of the global predictive model and the local predictive model is one of a decision tree, a support vector machine, a Bayesian network, clustering, reinforcement learning, or a neural network.
5. The method of claim 1 , wherein the predictive dialer calculates the call pacing hit ratio by calculating an average for the probabilities across the plurality of dialing records and multiplying the average by a number of dialing records found in the plurality of dialing records.
6. The method of claim 1 , wherein the plurality of dialing records are sorted based on a probability of making a right party contact by placing an outbound call to each of the dialing records found in the plurality of dialing records.
7. A non-transitory, computer-readable storage medium comprising computer-executable instructions for pacing outbound calls placed by a predictive dialer in a contact center that when executed by the predictive dialer are configured to cause the predictive dialer to:
calculate a call pacing hit ratio based on a probability for each dialing record in a plurality of dialing records of an outbound call placed to the dialing record resulting in a live connect, wherein the call pacing hit ratio represents a percentage of outbound calls expected to result in a live connect and the probability for each dialing record is determined by applying an ensemble to the dialing record comprising a global predictive model representing comprehensive dialing trends across at least one of multiple industries, multiple purposes, multiple locations, and multiple predictive dialers and a local predictive model representing a specific dialing history associated with the contact center;
determine a number of outbound calls to place based on the call pacing hit ratio and at least one of a number of agents associated with the contact center that are currently available to handle calls and a number of agents associated with the contact center that are expected to become available to handle calls within a time period;
select a set of dialing records from the plurality of dialing records based on the number of outbound calls to place; and
place virtually simultaneous outbound calls for each of the dialing records found in the set of dialing records.
8. The non-transitory, computer-readable storage medium of claim 7 , wherein the computer-executable instructions cause the predictive dialer to determine the number of outbound calls to place based also on a set target comprising at least one of: maintaining an abandonment rate; maintaining a time limit on how long a party can remain on hold before being connected with an agent; and maintaining a maximum wait time an agent can wait between calls.
9. The non-transitory, computer-readable storage medium of claim 7 , wherein the ensemble comprising the global predictive model and the local predictive model is generated using one of a technique of bagging, boosting, or stacking.
10. The non-transitory, computer-readable storage medium of claim 9 , wherein each of the global predictive model and the local predictive model is one of a decision tree, a support vector machine, a Bayesian network, clustering, reinforcement learning, or a neural network.
11. The non-transitory, computer-readable storage medium of claim 7 , wherein the computer-executable instructions cause the predictive dialer to calculate the call pacing hit ratio by calculating an average for the probabilities across the plurality of dialing records and multiplying the average by a number of dialing records found in the plurality of dialing records.
12. The non-transitory, computer-readable storage medium of claim 7 , wherein the plurality of dialing records are sorted based on a probability of making a right party contact by placing an outbound call to each of the dialing records found in the plurality of dialing records.
13. A system for pacing outbound calls in a contact center, the system comprising:
a predictive dialer configured to:
calculate a call pacing hit ratio based on a probability for each dialing record in a plurality of dialing records of an outbound call placed to the dialing record resulting in a live connect, wherein the call pacing hit ratio represents a percentage of outbound calls expected to result in a live connect and the probability for each dialing record is determined by applying an ensemble to the dialing record comprising a global predictive model representing comprehensive dialing trends across at least one of multiple industries, multiple purposes, multiple locations, and multiple predictive dialers and a local predictive model representing a specific dialing history associated with the contact center;
determine a number of outbound calls to place based on the call pacing hit ratio and at least one of a number of agents associated with the contact center that are currently available to handle calls and a number of agents associated with the contact center that are expected to become available to handle calls within a time period;
select a set of dialing records from the plurality of dialing records based on the number of outbound calls to place; and
place virtually simultaneous outbound calls for each of the dialing records found in the set of dialing records.
14. The system of claim 13 , wherein the predictive dialer is configured to determine the number of outbound calls to place based also on a set target comprising at least one of: maintaining an abandonment rate; maintaining a time limit on how long a party can remain on hold before being connected with an agent; and maintaining a maximum wait time an agent can wait between calls.
15. The system of claim 13 , wherein the ensemble comprising the global predictive model and the local predictive model is generated using one of a technique of bagging, boosting, or stacking.
16. The system of claim 15 , wherein each of the global predictive model and the local predictive model is one of a decision tree, a support vector machine, a Bayesian network, clustering, reinforcement learning, or a neural network.
17. The system of claim 13 , wherein the predictive dialer is configured to calculate the call pacing hit ratio by calculating an average for the probabilities across the plurality of dialing records and multiplying the average by a number of dialing records found in the plurality of dialing records.
18. The system of claim 13 , wherein the plurality of dialing records are sorted based on a probability of making a right party contact by placing an outbound call to each of the dialing records found in the plurality of dialing records.