CALL MAPPING SYSTEMS AND METHODS USING BAYESIAN MEAN REGRESSION (BMR)

• US 20130251138A1
• Filed: 03/15/2013
• Published: 09/26/2013
• Est. Priority Date: 03/26/2012
• Status: Active Grant

• Current Assignee: Afiniti Ltd
• Original Assignee: The Resource Group International Ltd
• Inventors: SPOTTISWOODE, S. James P., CHISHTI, Zia
• Granted Patent: US 9,025,757 B2
• Time in Patent Office: Days
• US Class Current: 379/265.12
• CPC Class Codes: G06Q 30/00 Commerce
• CPC Class Codes: H04M 3/5158 in combination with automated outdialling systems devices for signalling identity of wanted subscriber H04M1/26
• CPC Class Codes: H04M 3/5232 Call distribution algorithms
• CPC Class Codes: H04M 3/5233 Operator skill based call distribution
• CPC Class Codes: H04M 3/5238 with waiting time or load prediction arrangements

Abstract

A method, system and program product, the method comprising: determining a distribution of real agent performance from previous real agent performance data; determining a set of hypothetical agents with respective hypothetical agent performances AP_i ranging from a worst performance to a best performance; calculating for each of the set of hypothetical agents a posterior distribution taking into account actual results of a respective actual agent in multiple skills, using the distribution of real agent performance and the set of hypothetical agents with respective hypothetical agent performances AP_i, to obtain a total probability for each hypothetical agent of the set of the hypothetical agents; repeating calculating the posterior distribution steps for multiple of the hypothetical agents to obtain the respective total probabilities for the respective hypothetical agents; determining one hypothetical agent with a better value of total probability as the actual agent'"'"'s most probable global performance. This method may also be applied to obtain caller global propensity.

Claims

1. A method, comprising:

determining or obtaining or receiving, by one or more computers, a distribution of real agent performance from previous real agent performance data for a respective skill k in a set of skills;

determining, by the one or more computers, a set of hypothetical agents with respective hypothetical agent performances AP_i ranging from a worst performance to a best performance for the respective skill k;

calculating for each of the set of hypothetical agents, by the one or more computers, a posterior distribution taking into account actual results of a respective actual agent in each of the set of skills, using the distribution of real agent performance and the set of hypothetical agents with respective hypothetical agent performances AP_i, to obtain a total probability for each hypothetical agent of the set of the hypothetical agents;

repeating, by the one or more computers, the calculating the posterior distribution steps for multiple of the hypothetical agents in the set of hypothetical agents to obtain the respective total probabilities for the respective hypothetical agents; and

determining, by the one or more computers, one of the hypothetical agents with a better value of total probability TP as the actual agent'"'"'s most probable global performance.

2. The method as defined in claim 1, wherein the agent performance is one selected from the group of sale or no sale, revenue per call, revenue generating units (RGU) per call, handle time and customer satisfaction.

3. The method as defined in claim 1, wherein the agent skills k comprise two or more selected from the group of sales of product or service A, sales of product or service B, and providing service advice for a product C.

4. The method as defined in claim 1, wherein the most probable global performance determining step comprises selecting one of the hypothetical agents with a best value of total probability TP as the actual agent'"'"'s most probable global performance.

5. The method as defined in claim 1, wherein the set of hypothetical agents comprises at least 10 hypothetical agents.

6. The method as defined in claim 1, wherein the set of hypothetical agents comprises at least 50 hypothetical agents.

7. The method as defined in claim 1, wherein the set of hypothetical agents comprises at least 100 hypothetical agents.

8. The method as defined in claim 1, wherein the real agent performance is binomial and distribution of real agent performance is truncated at least at one end thereof.

9. The method as defined in claim 1, wherein the calculating the posterior distribution comprises:

calculating for each hypothetical agent, i, in the set of hypothetical agents, by the one or more computers, for a first skill k and the hypothetical agent performance AP_i for the respective hypothetical agent, i, a probability of the evidence POE_ik that the respective hypothetical agent i would obtain S sales on N calls, that the respective actual agent in that skill k obtained; and

calculating, by the one or more computers, a total probability TP_i for the hypothetical agent i, comprising multiplying AP_i for the hypothetical agent by the POE_ik for each skill k for the hypothetical agent i.

10. The method of claim 1, further comprising:

using, by the one or more computers, demographic data or psychographic data of the agents and demographic data or psychographic data of the callers in a multi-element pattern matching algorithm in a pair-wise fashion for a desired outcome to obtain a valuation for each of multiple of agent-caller pairs, and

combining, by the one or more computers, results of the pattern matching algorithm and the respective most probable global performances of the respective agents to select one of the agent-caller pairs.

11. A method, comprising:

determining or obtaining or receiving, by one or more computers, a distribution of real caller propensity from previous real caller propensity data for a respective caller partition in a set of caller partitions;

determining, by the one or more computers, a set of hypothetical callers with respective hypothetical caller propensities CP_i ranging from a worst propensity to a best propensity;

calculating for each of the set of hypothetical callers, by the one or more computers, a posterior distribution taking into account actual results of a respective actual caller in multiple of the caller partitions, using the distribution of real caller propensity and the set of hypothetical callers with respective hypothetical caller propensities CP_i, to obtain a total probability for each hypothetical caller of the set of the respective hypothetical callers;

repeating, by the one or more computers, the calculating the posterior distribution steps for multiple of the hypothetical callers in the set of hypothetical callers to obtain the respective total probabilities for the respective hypothetical callers; and

determining, by the one or more computers, one of the hypothetical callers with a better value of total probability TP as the actual callers'"'"'s most probable global propensity.

12. The method as defined in claim 11, wherein the caller propensity is one selected from the group of purchase of product or service A or no purchase, purchase of product or service B or no purchase, purchase of product or service C or no purchase, revenue per purchase, revenue generating units (RGU) per call, and handle time.

13. The method as defined in claim 11, wherein the partition is based at least in part on one or more selected from the group of demographic data, area code, zip code, NPANXX, VTN, geographic area, 800 number, and transfer number.

14. The method as defined in claim 11, wherein the calculating the posterior distribution comprises:

calculating for each hypothetical caller, i, in the set of hypothetical callers, by the one or more computers, for a first partition and the hypothetical caller propensity CP_i for the respective hypothetical caller, i, a probability of the evidence POE_ik that the respective hypothetical caller i would have S sales, that the respective actual caller in that partition k had; and

calculating, by the one or more computers, a total probability TP_i for the hypothetical caller i, comprising multiplying CP_i for the hypothetical caller by the POE_ik for each partition k for the hypothetical caller i.

15. A system, comprising:

one or more computers configured with program code that, when executed, causes performance of the following steps;

determining or obtaining or receiving, by the one or more computers, a distribution of real agent performance from previous real agent performance data for a respective skill k in a set of skills;

determining, by the one or more computers, a set of hypothetical agents with respective hypothetical agent performances AP_i ranging from a worst performance to a best performance for the respective skill k;

calculating for each of the set of hypothetical agents, by the one or more computers, a posterior distribution taking into account actual results of a respective actual agent in the set of skills, using the distribution of real agent performance and the set of hypothetical agents with respective hypothetical agent performances AP_i, to obtain a total probability for each hypothetical agent of the set of the hypothetical agents;

repeating, by the one or more computers, the calculating the posterior distribution steps for multiple of the hypothetical agents in the set of hypothetical agents to obtain the respective total probabilities for the respective hypothetical agents; and

determining, by the one or more computers, one of the hypothetical agents with a better value of total probability TP as the actual agent'"'"'s most probable global performance.

16. The system as defined in claim 15, wherein the agent performance is one selected from the group of sale or no sale, revenue per call, revenue generating units (RGU) per call, and handle time.

17. The system as defined in claim 15, wherein the agent skills k comprise two or more selected from the group of sales of product or service A, sales of product or service B, and providing service advice for a product C.

18. The system as defined in claim 15, wherein the most probable global performance determining step comprises selecting one of the hypothetical agents with a best value of total probability TP as the actual agent'"'"'s most probable global performance.

19. The system as defined in claim 15, wherein the set of hypothetical agents comprises at least 10 hypothetical agents.

20. The system as defined in claim 15, wherein the set of hypothetical agents comprises at least 50 hypothetical agents.

21. The system as defined in claim 15, wherein the set of hypothetical agents comprises at least 100 hypothetical agents.

22. The system as defined in claim 15, wherein the real agent performance is binomial and distribution of real agent performance is truncated at least at one end thereof.

23. The system as defined in claim 15, wherein the calculating the posterior distribution step comprises:

calculating for each hypothetical agent, i, in the set of hypothetical agents, by the one or more computers, for a first skill k and the hypothetical agent performance AP_i for the respective hypothetical agent, i, a probability of the evidence POE_ik that the respective hypothetical agent i would obtain S sales on N calls, that the respective actual agent in that skill k obtained; and

calculating, by the one or more computers, a total probability TP_i for the hypothetical agent i, comprising multiplying AP_i for the hypothetical agent by the POE_ik for each skill k for the hypothetical agent i.

24. The system of claim 15, further comprising the one or more computers configured with program code for performing the steps:

using, by the one or more computers, demographic data or psychographic data of the agents and demographic data or psychographic data of the callers in a multi-element pattern matching algorithm in a pair-wise fashion for a desired outcome to obtain a valuation for each of multiple of agent-caller pairs, and

combining, by the one or more computers, results of the pattern matching algorithm and the respective most probable global performances of the respective agents to select one of the agent-caller pairs.

25. A system, comprising:

one or more computers configured with program code that, when executed, causes performance of the following steps;

determining or obtaining or receiving, by the one or more computers, a distribution of real caller propensity from previous real caller propensity data for a respective caller partition in a set of caller partitions;

determining, by the one or more computers, a set of hypothetical callers with respective hypothetical caller propensities CP_i ranging from a worst propensity to a best propensity;

calculating for each of the set of hypothetical callers, by the one or more computers, a posterior distribution taking into account actual results of a respective actual caller in multiple of the caller partitions, using the distribution of real caller propensity and the set of hypothetical callers with respective hypothetical caller propensities CP_i, to obtain a total probability for each hypothetical caller of the set of the respective hypothetical callers;

repeating, by the one or more computers, the calculating the posterior distribution steps for multiple of the hypothetical callers in the set of hypothetical callers to obtain the respective total probabilities for the respective hypothetical callers; and

determining, by the one or more computers, one of the hypothetical callers with a better value of total probability TP as the actual callers'"'"'s most probable global propensity.

26. The system as defined in claim 25, wherein the caller propensity is one selected from the group of product or service A or no purchase, purchase of product or service B or no purchase, purchase of product or service C or no purchase, revenue per purchase, revenue generating units (RGU) per call, and handle time.

27. The system as defined in claim 25, wherein the partition is based at least in part on one or more selected from the group of demographic data, area code, zip code, NPANXX, VTN, geographic area, 800 number, and transfer number.

28. The system as defined in claim 25, wherein the calculating the posterior distribution step comprises:

calculating for each hypothetical caller, i, in the set of hypothetical callers, by the one or more computers, for a first partition and the hypothetical caller propensity CP_i for the respective hypothetical caller, i, a probability of the evidence POE_ik that the respective hypothetical caller i would have S sales, that the respective actual caller in that partition k had; and

calculating, by the one or more computers, a total probability TP_i for the hypothetical caller i, comprising multiplying CP_i for the hypothetical caller by the POE_ik for each partition k for the hypothetical caller i.