Browse Prior Art Database

System and Method to Identify Customer Dissatisfaction for Service Providers

IP.com Disclosure Number: IPCOM000239629D
Publication Date: 2014-Nov-20
Document File: 6 page(s) / 245K

Publishing Venue

The IP.com Prior Art Database

Abstract

Customer satisfaction is essential for maintaining customers, a key part for Customer Relationship Management (CRM) System. Traditionally, service providers try to identify customer satisfaction about their services through conducting surveys and defining metrics to measure their services performance, and the main objective for survey is to determine customer satisfaction. However customer dissatisfaction is a key factor for losing a customer, in order to maintain existing customers, service providers need to know customer dissatisfaction. But there exist the following problems and challenges for service providers to identify customer dissatisfaction: 1. High-cost and impractical for conducting survey for all services l Survey is time-consuming, l Huge number of services, (including some customers do NOT want to be surveyed). 2. Subjective survey objects selection and limited survey result l Survey object are RANDOMLY selected or selected by experts with their insights, l Satisfied and unsatisfied services are imbalanced (for example, 8:1, even more). In this invention, an approach for identifying customer dissatisfaction will be disclosed.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 53% of the total text.

Page 01 of 6

System and Method to Identify Customer Dissatisfaction for Service Providers

   Our invention aims to identify customer dissatisfaction by using statistical analysis techniques. The core idea is to classify the service requests into two groups (one is satisfied group and the other is unsatisfied group) through,


a collector to collect customer satisfaction information in historical and recent survey request records, as well as the metrics records for all service requests;


a satisfaction features selector to determine the important features for customer satisfaction prediction;


a customer satisfaction prediction trainer to build customer satisfaction models based on the collected historical services requests, metrics, and related survey results;


a model assemble to ensemble the models by model weights


a customer satisfaction predictor to predict recent service requests and classify them into two groups: satisfied groups and dissatisfied groups;


a survey object selector to help determine the survey service request objects;


a model trigger to calculate the models weights based on predicted results and surveyed results


and a data processor & model trigger to collect surveyed results and trigger customer satisfaction prediction trainer to update customer satisfaction prediction models.

  By using the above approach, user can find more dissatisfied customers than traditional survey approaches, then help user to maintain those identified dissatisfied customers.

The following is the process of our disclosure:


Step 1: Retrieve historical service requests and related surveys.

Step 2: Select service request features which are highly corrected with customer satisfaction.

Step 3: Train customer satisfaction prediction model set based on service requests and related surveys.

Step 4: Ensemble the prediction models based on their precision and recall through weighting models
Step 5: Retrieve recent service requests, predict their satisfaction, and group them by predicted satisfied and dissatisfied records.

Step 6: Select the survey candidates based on the predicted satisfaction results and other features.

Step 7: Conduct survey based on the satisfied and unsatisfied service request group
Step 8: Analyzed recent survey service requests impact on the trained model set. If the impact is significant, then go to Step 9, otherwise go to Step 2 to continuously train model set.

Step 9: Update the weights of the prediction models by comparing predicted results and surveyed results, and go to Step 4.

1


Page 02 of 6

More details about the above steps go as follows,

Step 2: Feature Selection


Objective:Select features from the collected metrics...