Predicting likelihood of customer attrition and retention measures

• US 9,165,270 B2
• Filed: 04/25/2009
• Issued: 10/20/2015
• Est. Priority Date: 12/20/2000
• Status: Active Grant

• Current Assignee: Acoustic, L.P.
• Original Assignee: International Business Machines Corporation
• Inventors: Dickinson, Jonathan, Desai, Paritosh
• Primary Examiner(s): BROCKINGTON III, WILLIAM S
• Application Number:
  US12/430,080
• Publication Number:
  US 20090276289A1
• Time in Patent Office: 2,369 Days
• US Class Current: 1/1
• CPC Class Codes: G06Q 10/04 Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem" market predictions or forecasting for commercial activities G06Q30/0202
• CPC Class Codes: G06Q 10/063 Operations research, analysis or management
• CPC Class Codes: G06Q 10/067 Enterprise or organisation modelling
• 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

The present invention relates to a system and method for customer retention. Historical transaction and customer data may be received from stores. Likewise, recent customer transaction data may be received from the stores. The transactions are linked to each customer. Attriters, historical customers who discontinued shopping, are identified. Next, risk factors for attrition may be identified by examining the attriters'"'"' transaction history for commonalities. From the risk factors a loss model may be generated. The loss model may be used, in conjunction with current transaction data, to generate the likelihood of loss for each of the current customers, which may then be reported. Retention measures may be generated for each customer by comparing the customer'"'"'s transactions to the loss model and the risk factors. The retention measures may be outputted to the stores, and a price optimization system. Likewise, the retention measures may be validated by comparing actual customer loss to the loss model.

Claims

1. A method for predicting likelihood of customer attrition, useful in association with at least one customer, the method comprising:

receiving historical data from at least one store, wherein the historical data includes historical transactions, and wherein the historical transactions are associated with historical customers;

receiving customer data for a plurality of customers from the at least one store, wherein the customer data includes transactions;

identifying at least one customer that provides economic value to at least one store from the plurality of customers;

linking each corresponding transaction of the customer data with one of the at least one customer and producing a reduced set of data limited to the at least one customer for faster processing;

identifying attriters, wherein the attriters are the historical customers who engage in an attrition behavior during the historical data;

identifying risk factors for attrition from the historical data, wherein the risk factors are defined by the historical transactions;

generating a loss model, using a computer, utilizing the identified risk factors, wherein generating the loss model comprises;

tuning the loss model in response to a difference between previous actual data and data generated by the loss model exceeding an expected discrepancy, wherein tuning the loss model includes correcting the loss model for biases produced by lost customers and to filter customers that are beyond being retained by retention measures;

generating likelihood of loss for each of the at least one customer, using the computer, by comparing the linked transactions to the loss model;

generating and validating at least one retention measure for the at least one customer using the computer, wherein validating includes;

calculating actual customer loss from applying the at least one retention measure at the at least one store;

modeling an expected customer loss via the loss model; and

determining a difference between the actual customer loss and the expected customer loss generated by the loss model over a time period to indicate an effectiveness of the at least one retention measure over that time period; and

controlling processing of the computer to process the reduced set of data limited to the at least one customer to increase performance speed.

2. The method recited by claim 1, wherein the attrition behavior is defined by configurable criteria.

3. The method recited by claim 2, wherein the attrition behavior is defined as at least one of customers who discontinue shopping, customers who stop shopping in a given time frame, and decline in average purchase size from a historical basis.

4. The method recited by claim 1, wherein the at least one retention measure is generated by comparing the linked transactions to the loss model and the risk factors.

5. The method recited by claim 1, further comprising calibrating the loss model by discounting a frequency risk factor of the attriters after date of attrition.

6. The method recited by claim 1, further comprising identifying high value customers from the plurality of customers.

7. The method recited by claim 1, wherein the risk factors include at least one of behavioral categories, tenure, basket value, number of adults, available customer information, number of stores shopped at, store quality, competitor type, frequency, and gaps in shopping frequency.

8. The method recited by claim 1, wherein the linking each corresponding transaction utilizes at least one of payment identifiers, loyalty program identifiers, registry identifiers, and biometric identifiers.

9. The method recited by claim 8, wherein conflicts between identifiers utilized for linking each corresponding transaction are resolved in favor of a most accurate identifier.

10. The method recited by claim 1, further comprising cleansing the received customer data from at least one store.

11. The method recited by claim 4, further comprising outputting the at least one retention measure to the at least one store.

12. The method recited by claim 1, further comprising outputting the at least one retention measure to a price optimization system, whereby the price optimization system uses the at least one retention measure to generate a preferred set of prices for one or more products, and wherein the preferred set of prices is configured to limit customer loss by a set threshold.

13. The method recited by claim 1, further comprising generating a loss value for each of the at least one customer by comparing the likelihood of loss for said customer with an economic value said customer provides the at least one store.

14. An apparatus comprising a likelihood of customer attrition predictor, useful in association with at least one customer, the likelihood of customer attrition predictor comprising:

a data analyzer including a processor configured to receive historical data from at least one store, wherein the historical data includes historical transactions, and wherein the historical transactions are associated with historical customers, and wherein the data analyzer further is configured to receive customer data for a plurality of customers from the at least one store and to identify at least one customer that provides economic value to at least one store from the plurality of customers, wherein the customer data includes transactions, and wherein the data analyzer further is configured to identify attriters, wherein the attriters are the historical customers who engage in an attrition behavior during the historical data;

an identity determiner including a processor configured to link each corresponding transaction of the customer data with one of the at least one customer and produce a reduced set of data limited to the at least one customer for faster processing;

a loss model engine including a processor configured to;

identify risk factors for attrition from the historical data, wherein the risk factors are defined by the historical transactions;

generate a loss model utilizing the identified risk factors, wherein generating the loss model comprises;

tuning the loss model in response to a difference between previous actual data and data generated by the loss model exceeding an expected discrepancy, wherein tuning the loss model includes correcting the loss model for biases produced by lost customers and to filter customers that are beyond being retained by retention measures;

a likelihood of loss determiner including a processor configured to generate likelihood of loss for each of the at least one customer by comparing the linked transactions to the loss model; and

a stratagem generator configured to;

generate at least one retention measure for the at least one customer; and

validate the at least one retention measure by;

calculating actual customer loss from applying the at least one retention measure at the at least one store;

modeling an expected customer loss via the loss model; and

determining a difference between the actual customer loss and the expected customer loss generated by the loss model over a time period to indicate an effectiveness of the at least one retention measure over that time period;

wherein the processing of the loss model engine, likelihood of loss determiner, and stratagem generator performed by the corresponding processors are controlled to process the reduced set of data limited to the at least one customer to increase performance speed.

15. The apparatus of claim 14, wherein the data analyzer is enabled to define the attrition behavior by configurable criteria.

16. The apparatus of claim 15, wherein the attrition behavior is defined as at least one of customers who discontinue shopping, customers who stop shopping in a given time frame, and decline in average purchase size from a historical basis.

17. The apparatus of claim 14, wherein the at least one retention measure is generated by comparing the linked transactions to the loss model and the risk factors.

18. The apparatus of claim 14, wherein the loss model engine is further configured to calibrate the loss model by discounting a frequency risk factor of the attriters after date of attrition.

19. The apparatus of claim 14, further comprising a valuable customer identifier configured to identify high value customers from the plurality of customers.

20. The apparatus of claim 14, wherein the risk factors include at least one of behavioral categories, tenure, basket value, number of adults, available customer information, number of stores shopped at, store quality, competitor type, frequency, and gaps in shopping frequency.

21. The apparatus of claim 14, wherein the linking each corresponding transaction utilizes at least one of payment identifiers, loyalty program identifiers, registry identifiers, and biometric identifiers.

22. The apparatus of claim 21, wherein conflicts between identifiers utilized for linking each corresponding transaction are resolved in favor of a most accurate identifier.

23. The apparatus of claim 14, further comprising a data cleanser configured to cleanse the received customer data from at least one store.

24. The apparatus of claim 17, further comprising an outputter configured to output the at least one retention measure to the at least one store.

25. The apparatus of claim 14, further comprising an outputter configured to output the at least one retention measure to a price optimization system, whereby the price optimization system uses the at least one retention measure to generate a preferred set of prices for one or more products, and wherein the preferred set of prices is configured to limit customer loss by a set threshold.

26. The apparatus of claim 14, further comprising a loss value generator configured to generate a loss value for each of the at least one customer by comparing the likelihood of loss for said customer with an economic value said customer provides the at least one store.

27. A computer storage product for predicting likelihood of customer attrition, useful in association with at least one customer, comprising:

a non-transitory computer readable medium having computer readable program code embodied therewith for execution on one or more processors, the computer readable program code comprising computer readable program code configured to;

receive historical data from at least one store, wherein the historical data includes historical transactions, and wherein the historical transactions are associated with historical customers;

receive customer data for a plurality of customers from the at least one store, wherein the customer data includes transactions;

identify at least one customer that provides economic value to at least one store from the plurality of customers;

link each corresponding transaction of the customer data with one of the at least one customer and produce a reduced set of data limited to the at least one customer for faster processing;

identify attriters, wherein the attriters are the historical customers who engage in an attrition behavior during the historical data;

identify risk factors for attrition from the historical data, wherein the risk factors are defined by the historical transactions;

generate a loss model utilizing the identified risk factors, wherein generating the loss model comprises;

tuning the loss model in response to a difference between previous actual data and data generated by the loss model exceeding an expected discrepancy, wherein tuning the loss model includes correcting the loss model for biases produced by lost customers and to filter customers that are beyond being retained by retention measures;

generate likelihood of loss for each of the at least one customer by comparing the linked transactions to the loss model;

generate and validate at least one retention measure for the at least one customer, wherein validating includes;

calculating actual customer loss from applying the at least one retention measure at the at least one store;

modeling an expected customer loss via the loss model; and

determining a difference between the actual customer loss and the expected customer loss generated by the loss model over a time period to indicate an effectiveness of the at least one retention measure over that time period; and

control processing of the one or more processors to process the reduced set of data limited to the at least one customer to increase performance speed.

28. The computer storage product recited by claim 27, wherein the attrition behavior is defined by configurable criteria.

29. The computer storage product recited by claim 28, wherein the attrition behavior is defined as at least one of customers who discontinue shopping, customers who stop shopping in a given time frame, and decline in average purchase size from a historical basis.

30. The computer storage product recited by claim 27, wherein the at least one retention measure is generated by comparing the linked transactions to the loss model and the risk factors.

31. The computer storage product recited by claim 27, wherein the computer readable program code further comprises computer readable program code configured to calibrate the loss model by discounting a frequency risk factor of the attriters after date of attrition.

32. The computer storage product recited by claim 27, wherein the computer readable program code further comprises computer readable program code configured to identify high value customers from the plurality of customers.

33. The computer storage product recited by claim 27, wherein the risk factors include at least one of behavioral categories, tenure, basket value, number of adults, available customer information, number of stores shopped at, store quality, competitor type, frequency, and gaps in shopping frequency.

34. The computer storage product recited by claim 27, wherein the linking each corresponding transaction utilizes at least one of payment identifiers, loyalty program identifiers, registry identifiers, and biometric identifiers.

35. The computer storage product recited by claim 34, wherein conflicts between identifiers utilized for linking each corresponding transaction are resolved in favor of a most accurate identifier.

36. The computer storage product recited by claim 27, wherein the computer readable program code further comprises computer readable program code configured to cleanse the received customer data from at least one store.

37. The computer storage product recited by claim 30, wherein the computer readable program code further comprises computer readable program code configured to output the at least one retention measure to the at least one store.

38. The computer storage product recited by claim 27, wherein the computer readable program code further comprises computer readable program code configured to output the at least one retention measure to a price optimization system, whereby the price optimization system uses the at least one retention measure to generate a preferred set of prices for one or more products, and wherein the preferred set of prices is configured to limit customer loss by a set threshold.

39. The computer storage product recited by claim 27, wherein the computer readable program code further comprises computer readable program code configured to generate a loss value for each of the at least one customer by comparing the likelihood of loss for said customer with an economic value said customer provides the at least one store.