Apparatus and method for predicting customer behavior

• US 9,129,290 B2
• Filed: 02/24/2009
• Issued: 09/08/2015
• Est. Priority Date: 02/22/2006
• Status: Active Grant

• Current Assignee: [24]7.ai, Inc.
• Original Assignee: 24/7 Customer Incorporated
• Inventors: Kannan, Pallipuram V., Jain, Mohit, Vijayaraghavan, Ravi
• Primary Examiner(s): Miller, Alan S
• Application Number:
  US12/392,058
• Publication Number:
  US 20090222313A1
• Time in Patent Office: 2,387 Days
• Field of Search: 705 711- 742, 707600-606, 703/22, 379/265.06
• US Class Current: 1/1
• CPC Class Codes: G06N 7/01 Probabilistic graphical models, e.g. probabilistic networks
• CPC Class Codes: G06Q 10/067 Enterprise or organisation modelling
• CPC Class Codes: G06Q 30/016 After-sales
• CPC Class Codes: G06Q 30/02 Marketing; Price estimation or determination; Fundraising
• CPC Class Codes: G06Q 30/0202 Market predictions or forecasting for commercial activities

Abstract

A predictive model generator that enhances customer experience, reduces the cost of servicing a customer, and prevents customer attrition by predicting the appropriate interaction channel through analysis of different types of data and filtering of irrelevant data. The model includes a customer interaction data engine for transforming data into a proper format for storage, data warehouse for receiving data from a variety of sources, and a predictive engine for analyzing the data and building models.

Claims

1. An apparatus for generating a prediction of customer behavior and for selecting and assigning an interaction channel to said customer from among a plurality of interaction channels, comprising:

a memory;

a customer interaction data engine processor communicatively coupled to said memory and configured for executing in said memory;

transforming data received from a plurality of sources into a format acceptable for storage, said data comprising problem dimension data relating to a customer interaction arising from an issue associated with a product or service, product dimension data relating to a product or service purchased by a customer, structured and unstructured customer dimension data about a plurality of customers, and structured and unstructured agent dimension data about a plurality of agents;

a data warehouse coupled to said customer interaction data engine, said data warehouse configured for storing said transformed data; and

a predictive engine coupled to said data warehouse and communicatively coupled to said memory, said predictive engine processor configured for executing in said memory;

receiving said stored data from said data warehouse;

compiling said data and determining contributing variables, wherein said compiling and determining comprises computing contributing variables according to whether the variable is for a numerical or categorical prediction and wherein contributing variables for a numerical prediction or a categorical prediction are computed using one or more statistical or predictive algorithms comprising any of;

linear regression, logistic regression, Naï

ve Bayes, neural networks, and support vector machines;

using said contributing variables to generate predictive models;

identifying predictive trends in customer behavior as a function of particular data from said stored data, said particular data comprising;

a product or service purchased by customers, time that has elapsed from a time said product or service was purchased, customer location, a problem associated with said product or service, and customer impact associated with said problem associated with said product or service, wherein customer impact is a function of type of customer interaction and customer lifecycle;

predicting any of;

probability of a customer to face a particular problem or issue based on an engagement stage of the customer with the company, wherein the engagement stage is product or service specific and measured by time after purchase or is the stage of a life cycle of the customer;

a customer'"'"'s preference of a particular channel based on type of concern and reduction of resolution time through the particular channel; and

a probable impact of a particular problem on a customer'"'"'s loyalty, growth, and profitability score; and

selecting and assigning an interaction channel to said customer from among a plurality of interaction channels based upon said predicting.

2. The apparatus of claim 1, wherein said customer interaction data engine processor receives unstructured data from conversations between an agent and a customer and said predictive engine processor generates variables for predicting customer behavior based on said text.

3. The apparatus of claim 1, wherein said customer interaction data engine processor receives unstructured data from conversations between an agent and a customer and merges said unstructured data into said format acceptable for storage.

4. The apparatus of claim 1, said customer interaction data engine processor configured for extracting keywords from said received data, said received data further comprising interactions between an agent and a customer, said keywords indicative of at least agent cognitive capabilities, agent emotions, and customer emotions, said customer interaction data engine processor configured for calculating a net experience score using said keywords.

5. The apparatus of claim 1, wherein said problem data comprises at least a unique identification number associated with a particular problem, a description of said problem, a category, a sub-category, and customer impact.

6. The apparatus of claim 1, wherein said product data comprises at least a unique identification number associated with each product, a product, a category, a sub-category if applicable, information about launch of said product or service, and notes.

7. The apparatus of claim 1, wherein said customer data comprises at least a unique identification number associated with each customer, a company name, a location, a phone number, a role contact, an account start date, a first shipment date, a last shipment date, a number of products or services purchased by each customer in a last month, a number of products or services purchased in a last year, a statistics code, a credit limit, and a type of industry associated with a customer.

8. The apparatus of claim 1, wherein said agent data comprises at least a unique identification number associated with each agent, each agent'"'"'s name, each agent'"'"'s skill sets, each agent'"'"'s relevant experience, interactions per day, customer satisfaction scores, gender, average handle times, holding time, talk time, call outcome, call satisfaction, and experience in a process.

9. The apparatus of claim 1, further comprising a user interface for allowing a user to generate models.

10. The apparatus of claim 1, further comprising a user interface for receiving information from a customer for said predictive engine processor to use in comparing said customer to a model of said generated predictive models.

11. A computer-implemented method for generating a model for predicting customer behavior and for selecting and assigning an interaction channel to said customer from among a plurality of interaction channels, the method comprising the steps of:

receiving at a customer interaction data engine problem data comprising data relating to any customer interaction arising from a purchase of a product or service;

receiving at said customer interaction data engine product data comprising data relating to a product or service purchased by a customer;

receiving at said customer interaction data engine customer interaction data comprising information about a customer;

receiving at said customer interaction data engine agent data comprising information about an agent;

transforming with said customer interaction data engine said problem data, product data, customer interaction data, and agent data into a storage format;

storing said data on a computer readable storage medium in a data warehouse;

determining with a predictive engine contributing variables using said data stored in said data warehouse, wherein determining comprises computing contributing variables according to whether the variable is for a numerical or categorical prediction and wherein contributing variables for a numerical prediction or a categorical prediction are computed using one or more statistical or predictive algorithms comprising any of;

linear regression, logistic regression, Naï

ve Bayes, neural networks, and support vector machines;

building with said predictive engine a plurality of predictive models using said contributing variables, wherein any of said plurality of predictive models is configured to predict any of;

probability of a customer to face a particular problem or issue based on an engagement stage of the customer with the company, wherein the engagement stage is product or service specific and measured by time after purchase or is the stage of a life cycle of the customer;

a customer'"'"'s preference of a particular channel based on type of concern and reduction of resolution time through the particular channel; and

a probable impact of a particular problem on a customer'"'"'s loyalty, growth, and profitability score;

testing and validating said plurality of predictive models;

receiving a request to generate a predictive model for a particular customer using, among other things, said contributing variables;

generating said customer predictive model in real time; and

using said customer predictive model to select and assign an interaction channel to said customer from among a plurality of interaction channels.

12. The method of claim 11, wherein said contributing variables are determined by applying the Naï

ve Bayes algorithm to said data.

13. The method of claim 11, further comprising identifying with a processor trends in customer behavior as a function of said data stored in said data warehouse, wherein said stored data comprises:

a product or service purchased by customers, time that has elapsed from a time said product or service was purchased, customer location, a problem associated with said product or service, and customer impact associated with said problem associated with said product or service; and

generating a model that predicts a mode for interacting with a customer to minimize cost associated with said interaction based on said identified trends in customer behavior.

14. The method of claim 11, further comprising the steps of:

presenting a customer with a user interface for asking said customer to identify at least one of said customer'"'"'s name, a product or service purchased by said customer, and a time of purchase;

comparing said information identified by said customer to said model; and

predicting a mode of interaction with said customer that minimizes a cost of interacting with said customer, said predicting based in part on said comparing said information to said model.

15. The method of claim 14, further comprising the steps of:

responsive to identifying said mode of interaction as a phone call with an agent;

identifying a specific agent based on said agent'"'"'s experience level; and

providing said customer with a number of said specific agent.

16. The method of claim 11, further comprising the steps of:

receiving with said customer data interaction engine text files from conversations between an agent and a customer;

extracting keywords from said text files that are indicative of agent cognitive capabilities, agent emotions, and customer emotions;

weighting said keywords; and

calculating a net experience score of the customer based on said weighted keywords.

17. The method of claim 13 or 14, wherein said mode for interacting with said customer comprises telephone, internet chat, email, and web.

18. The method of claim 13, wherein said trends in customer behavior further comprise a function of an agent, an agent'"'"'s gender, and an agent'"'"'s experience.

19. The method of claim 11, wherein said predictive engine is configured to generate a model comprising at least one of a text table, aligned bars, stacked bars, discrete lines, scatter plots, Gantt charts, heat maps, side-by-side bars, measure bars, circles, scatter matrices, and histograms.

20. An apparatus for generating a prediction of customer behavior and for selecting and assigning an interaction channel to said customer from among a plurality of interactions channels, comprising:

a customer interaction data engine processor configured for transforming data received from a plurality of sources into a format acceptable for storage, said data comprising problem data relating to a customer interaction arising from a purchase of a product or service, product data relating to a product or service purchased by a customer, customer data about a plurality of customers, and agent data about a plurality of agents;

a data warehouse coupled to said customer interaction data engine, said data warehouse configured for storing said transformed data;

a predictive engine processor coupled to said data warehouse, said predictive engine configured for;

receiving said stored data from said data warehouse,

compiling said data and determining contributing variables, wherein said compiling and determining comprises computing contributing variables according to whether the variable is for a numerical or categorical prediction and wherein contributing variables for a numerical prediction or a categorical prediction are computed using one or more statistical or predictive algorithms comprising any of;

linear regression, logistic regression, Naï

ve Bayes, neural networks, and support vector machines,

using said contributing variables to generate predictive models,

receiving a request for a predictive model,

generating said requested predictive model in real time in response to said request wherein said requested predictive model predicts any of;

probability of a customer to face a particular problem or issue based on an engagement stage of the customer with the company, wherein the engagement stage is product or service specific and measured by time after purchase or is the stage of a life cycle of the customer;

a customer'"'"'s preference of a particular channel based on type of concern and reduction of resolution time through the particular channel;

a probable impact of a particular problem on a customer'"'"'s loyalty, growth, and profitability score, and

using said predictive model to select and assign an interaction channel to said customer from among a plurality of interaction channels based upon said predicting.