Churn analysis system

• US 8,630,892 B2
• Filed: 03/30/2012
• Issued: 01/14/2014
• Est. Priority Date: 08/31/2011
• Status: Active Grant

• Current Assignee: Accenture Global Services Limited (Accenture PLC)
• Original Assignee: Accenture Global Services Limited (Accenture PLC)
• Inventors: Bhalla, Anuj, Amonju, Vukieali, Walls, Terry Lynn, Honts, Robert Wayne, Logering, Matthew Dean
• Primary Examiner(s): Sterrett, Jonathan G
• Assistant Examiner(s): ROSS, SCOTT M
• Application Number:
  US13/436,482
• Publication Number:
  US 20130054306A1
• Time in Patent Office: 655 Days
• Field of Search: 705/7.29, 705/7.31
• US Class Current: 705/7.31
• CPC Class Codes: G06Q 30/02 Marketing; Price estimation or determination; Fundraising
• CPC Class Codes: G06Q 30/0201 Market modelling; Market analysis; Collecting market data
• CPC Class Codes: G06Q 30/0202 Market predictions or forecasting for commercial activities

Abstract

A churn analysis system helps a business analyze, predict, and reduce customer churn. The system analyzes customer experiences by using an insightful block level approach to correlate customer experience with customer churn. Through the block level approach, the system is able to more accurately predict and effectively reduce future customer churn. As a result, businesses are able to reduce customer acquisition costs and improve customer retention rates.

Claims

1. A method for analyzing customer propensity to churn, comprising:

defining a maximum contact intensity parameter that provides a decision threshold for determining how to categorize customer interactions;

accessing a database of customer interaction data that represents interactions of a customer with a service provider;

analyzing, by a computer processor in communication with the database, the customer interaction data to create a customer experience block for the customer, the customer experience block capturing, from the customer interaction data, a first contact to resolved contact interaction sequence of the customer with the service provider, where the processor creates the customer experience block by;

sorting the customer interaction data chronologically for the customer;

identifying a first specific contact interaction in the customer interaction data as the first contact interaction when a previous contact interaction in the customer interaction data exceeds the maximum contact intensity with respect to the first specific contact interaction;

assigning a start row index to the first contact interaction;

identifying a second specific contact interaction in the customer interaction data as the resolved contact interaction when a subsequent contact interaction in the customer interaction data exceeds the maximum contact intensity, the customer experience block thus created such that any selected contact interaction in the first contact to resolved contact interaction sequence is within the maximum contact intensity of an immediately preceding contact interaction in the interaction sequence, if any;

assigning a common customer identifier of the customer to each contact interaction in the first contact to resolved contact interaction sequence;

assigning an end row index to the resolved contact interaction; and

assigning a block index to identify the customer experience block and to associate the first contact to resolved contact interaction sequence with the start row index and the end row index;

saving the customer experience block in a unified service analytic record where the block index distinguishes between multiple different customer experience blocks for different customers in the unified service analytic record;

determining an interaction metric that is specific to a particular customer contact interaction in the first contact to resolved contact interaction sequence;

determining a block metric derived from all customer contact interactions in the first contact to resolved contact interaction sequence;

submitting the unified service analytic record, interaction metric and block metric to a churn prediction model; and

receiving a customer churn analysis result from the churn prediction model.

2. The method of claim 1, further comprising:

identifying a plurality of unresolved contact interactions in the first contact to resolved contact interaction sequence, where the unresolved contact interactions represent interactions of the customer occurring after the first contact interaction and before the resolved contact interaction, each unresolved contact interaction being within the maximum contact intensity of any immediately preceding contact interaction and any immediately subsequent contact interaction in the interaction sequence;

assigning intermediate row indices to the unresolved contact interactions, where the start row index, intermediate row indices, and end row index collectively represent a chronological order of all customer contact interactions in the interaction sequence; and

associating the block index with the start row index, intermediate row indices, and end row index and any block metrics derived from all customer contacts in the interaction sequence.

3. The method of claim 1, where the maximum contact intensity comprises a time period.

4. The method of claim 1, where the customer interaction data spans multiple customer interaction channels;

and where the block metric comprises a block path giving a chronological customer interaction channel sequence for the first contact to resolved contact interaction sequence.

5. The method of claim 1, where the interaction metric comprises a churn flag indicating customer churn has occurred at a predetermined period of time from a start time of the particular customer contact interaction.

6. The method of claim 1, further comprising creating the churn prediction model by:

extracting from the database of customer interaction data a set of sample customer data;

determining a plurality of sample interaction metrics and a plurality of sample block metrics from the set of sample customer data; and

creating a best fit equation by performing statistical regression analysis on the set of sample customer data based on the plurality of sample interaction metrics and the plurality of sample block metrics.

7. The method of claim 4, where the customer interaction channels comprises a retail location of the service provider.

8. The method of claim 4, where an additional block metric comprises an agent count giving a total number of unique agents who interacted with the customer in the first contact to resolved contact interaction sequence.

9. The method of claim 6, where the statistical regression analysis is Cox regression analysis.

10. The method of claim 6, where the customer churn analysis result predicts a customer propensity to churn for the customer.

11. A system for analyzing customer propensity to churn, comprising:

a computer processor; and

a memory in communication with the computer processor, the memory comprising churn analysis logic, which when executed by the computer processor causes the computer processor to;

define a maximum contact intensity parameter that provides a decision threshold for determining how to categorize customer interactions;

access a database of customer interaction data that represent interactions of a customer with a service provider;

create, from the customer interaction data, and store in the memory a customer experience block for the customer, the customer experience block capturing, from the customer interaction data, a first contact to resolved contact interaction sequence of the customer with the service provider,

where the processor creates the customer experience block by;

sorting the customer interaction data chronologically for the Customer;

identifying a first specific contact interaction in the customer interaction data as the first contact interaction when a previous contact interaction in the customer interaction data exceeds the maximum contact intensity with respect to the first specific contact interaction;

assigning a start row index to the first contact interaction;

identifying a second specific contact interaction in the customer interaction data as the resolved contact interaction when a subsequent contact interaction in the customer interaction data exceeds the maximum contact intensity, the customer experience block thus created such that any selected contact interaction in the first contact to resolved contact interaction sequence is within the maximum contact intensity of an immediately preceding contact interaction in the interaction sequence, if any;

assigning a common customer identifier of the customer to each contact interaction in the first contact to resolved contact interaction sequence;

assigning an end row index to the resolved contact interaction; and

assigning a block index to identify the customer experience block and to associate the first contact to resolved contact interaction sequence with the start row index and the end row index;

save the customer experience block in a unified service analytic record stored in the memory, where the block index distinguishes between multiple different customer experience blocks in the unified service analytic record;

determine an interaction metric that is specific to a particular customer contact interaction in the first contact to resolved contact interaction sequence;

determine a block metric derived from all customer contact interactions in the first contact to resolved contact interaction sequence;

submit the interaction metric and block metric to a churn prediction model; and

receive a customer churn analysis result from the churn prediction model.

12. The system of claim 11, where the maximum contact intensity comprises a time period.

13. The system of claim 11, where the customer interaction data spans multiple customer interaction channels;

and where the block metric comprises a block path giving a chronological customer interaction channel sequence for the first contact to resolved contact interaction sequence.

14. The system of claim 11, where the interaction metric comprises a churn flag indicating customer churn has occurred at a predetermined period of time from a start time of the particular customer contact interaction.

15. The system of claim 11, where the churn analysis logic further causes the computer processor to create the churn prediction model by:

extracting from the database of customer interaction data a set of sample customer data;

determining a plurality of sample interaction metrics and a plurality of sample block metrics from the set of sample customer data; and

creating a best fit equation by performing statistical regression analysis on the set of sample customer data based on the plurality of sample interaction metrics and the plurality of sample block metrics.

16. The system of claim 15, where the statistical regression analysis is Cox regression analysis.

17. The system of claim 15, where the customer churn analysis result predicts a customer propensity to churn for the customer.

18. A method for analyzing customer propensity to churn, comprising:

defining a maximum contact intensity parameter that provides a decision threshold for determining how to categorize customer interactions;

accessing a database of customer interaction data that represents interactions of a plurality of customers with a service provider;

analyzing, by a computer processor in communication with the database, the customer interaction data to create a plurality of customer experience blocks for the plurality of customers, the plurality of customer experience blocks capturing, from the customer interaction data, a plurality of first contact to resolved contact interaction sequences of the customers with the service provider, where the processor creates the customer experience blocks by;

sorting the customer interaction data chronologically for each customer;

identifying, for each customer experience block, a first specific contact interaction in the customer interaction data as the first contact interaction when a previous contact interaction in the customer interaction data exceeds the maximum contact intensity with respect to the first specific contact interaction;

assigning a start row index to the first contact interaction;

identifying, for each customer experience block, a second specific contact interaction in the customer interaction data as the resolved contact interaction when a subsequent contact interaction in the customer interaction data exceeds the maximum contact intensity, each customer experience block thus created such that any selected contact interaction in a specific first contact to resolved contact interaction sequence is within the maximum contact intensity of an immediately preceding contact interaction in the specific first contact to resolved contact interaction sequence, if any;

assigning a common customer identifier of the customer to each contact interaction in the specific first contact to resolved contact interaction sequence;

assigning an end row index to the resolved contact interaction; and

assigning a block index to identify the customer experience block and to associate the first contact to resolved contact interaction sequence with the start row index and the end row index;

saving the plurality of customer experience blocks in a unified service analytic record comprising block indices of the plurality of customer experience blocks, the block indices distinguishing between the plurality of customer experience blocks;

determining, from the unified service analytic record, interaction metrics specific to particular customer contact interactions in the plurality of first contact to resolved contact interaction sequences;

determining, from the unified service analytic record, block metrics derived from all customer contact interactions in the plurality of first contact to resolved contact interaction sequences;

submitting the unified service analytic record, interaction metrics and block metrics to a churn prediction model; and

receiving a customer churn analysis result from the churn prediction model.