System and method for automated population splitting to assist task management through analytics
View Patent ↗Task management supported by analytics has enhanced effectiveness through automated population splitting for generating models that predict task associated outcomes. Subpopulations are automatically analyzed for adequate sample size and selectively recombined based on characteristics of the subpopulations, such as to combine homogeneous subpopulations or to optimize a subpopulation characteristic. As an example, outbound calling campaign models are generated for time periods by splitting a sample population into time period subpopulations and combining subpopulation time periods having inadequate sample size so that combined subpopulations have a minimized total time period or greater similarity relative to other subpopulations.
1. A method for placing outbound calls to accounts during a campaign period, each account having account information, the method comprising:
dividing the campaign into plural time periods;
dividing a sample population of accounts having call results into the plural subpopulations based upon a characteristic of a variable of the sample population;
analyzing the subpopulations to identify at least one subpopulation having an inadequate sample size;
selectively combining the identified subpopulation with one or more contiguous of the other subpopulations so that all subpopulations remaining after the selectively combining have an adequate sample size;
generating a model for each subpopulation remaining after the selectively combining, the model predicting call results for outbound calls;
applying the account information to the models with a computer system to predict outbound call results for the accounts during the plural time periods; and
placing outbound calls with a telephone dialer to the accounts based at least in part on the predicted outbound call results.
2. The method of claim 1 wherein the campaign period comprises sixteen one hour interval subpopulation time periods.
3. The method of claim 1 wherein placing outbound calls further comprises placing outbound calls with a predictive dialer to optimize a predetermined factor.
4. The method of claim 3 wherein the predetermined factor comprises right party contacts.
5. The method of claim 3 wherein the predetermined factor comprises revenue associated with the account information.
6. A system for managing a task based on predicted outcomes associated with the task, the system comprising:
a variable splitter to split a population having a variable into plural subpopulations based upon a characteristic of the variable;
a subpopulation validator to identify one or more subpopulations that are inadequate to support a model;
a subpopulation pruner to combine the one or more identified subpopulations with another of the subpopulations so that plural subpopulations remain after the combining, each subpopulation that remains being adequate to support a model, the subpopulation pruner selecting at least a first and second consecutive time period to combine to minimize the maximum length of the time periods associated with the combined subpopulations;
a modeling engine to generate a model for each of the subpopulations that remain; and
a task manager to apply the models to predict outcomes and to manage the task based upon the predicted outcomes.
7. The system of claim 6 wherein the task manager comprises a predictive dialer and the predicted outcomes comprise the likelihood that an outbound call from the predictive dialer will result in a right party contact.
8. The system of claim 7 wherein the population comprises accounts, each account having at least one call attempt outcome at a time of day, and the subpopulations comprise time periods, each account assigned to subpopulations having a time period corresponding to the call attempt outcome time of day of the account.
9. The system of claim 7 wherein the task manager comprises a system for sending mail and the predicted outcome comprises a response to the mail by a recipient of the mail.
10. The system of claim 7 wherein the subpopulation pruner is further operable to combine the subpopulations to optimize a predetermined factor.
11. The system of claim 10 wherein the predetermined factor comprises a measure of model validation for the models generated from the subpopulations.
12. The system of claim 10 wherein the predetermined factor comprises the size of the subpopulations.
13. A method for managing a task, the method comprising:
splitting a population into at least two or more subpopulations;
determining with a computer system that at least one of the subpopulations provides a sample inadequate to generate a valid model;
combining the inadequate subpopulation with a selected of the other subpopulations that are contiguous with the inadequate subpopulation so that each subpopulation remaining after the combining provides a sample adequate to generate a valid model;
generating a model with a computer system for each of the subpopulations;
applying the models to predict outcomes associated with the task; and
managing the task according to the predicted outcomes.
14. The method of claim 13 wherein combining the inadequate subpopulation further comprises:
analyzing the subpopulations to determine the similarity of at least one characteristic between the subpopulations; and
selecting the subpopulation to combine with the inadequate subpopulation that has the greatest similarity to the inadequate subpopulation.
15. The method of claim 13 wherein combining the inadequate subpopulation further comprises:
analyzing the subpopulations for a desired characteristic; and
selecting the subpopulation to combine with the inadequate subpopulation that optimizes the desired characteristic.
16. The method of claim 15 wherein the desired characteristic comprises a length of time associated with the subpopulation.
17. The method of claim 16 wherein selecting the subpopulation to combine further comprises minimizing the length of time.