System and method of analyzing social media to predict the churn propensity of an individual or community of customers

Abstract

A system and method for mining social media signals and cues i) created by a user (for example, a customer) and/or ii) to which the user is exposed (the “data”), and for processing that data as it relates to a service (including a fee or subscription-based service), in order to predict the user's predisposition or likelihood to either leave the subscription or the service or reduce his/her engagement with the subscription or the service.

We claim: 
     
         1 . A computer implemented method of collecting/mining data relating to social media influence around a customer, and analyzing said data to predict a customer's predisposition to either leave a subscription or a service or reduce his/her engagement with a subscription or a service which comprises: a) receiving a plurality of social media inputs associated with the customer; b) determining a churn probability for the customer; and c) performing an action based on the determined churn probability. 
     
     
         2 . A computer-implemented method to characterise social influence and to predict behavior of a user, said user being part of a social network which comprises a) creating a dynamically updatable social influence profile of the user, b) predicting future behavior of the user based on influence given by the user and received by the user from his social circles, and thereafter c) predicting the user's predisposition to either leave a subscription or a service or reduce his/her engagement with a subscription or a service. 
     
     
         3 . A computer implemented method of collecting/mining data relating to social media influence around a customer, and analyzing said data to predict a customer's predisposition to either leave a subscription or a service or reduce his/her engagement with a subscription or a service comprises:
 a) identifying a social media profile of the customer;   b) comparing customer and his social media profile to clusters of customers, based upon similar social media profiles (“cohorts”); and   c) calculating predicted churn behavior of the customer, based upon known churn behavior of cohorts.   
     
     
         4 . The method of  claim 3  wherein cohorts are identified by a) extracting a plurality of feature vectors of the customer; and b) computing cohorts from the feature vectors. 
     
     
         5 . The method of  claim 3  wherein feature vectors are social network inputs, cues and influences. 
     
     
         6 . A system, comprising: an information module that is configured to identify a user of a service; a probability module that is configured to determine a churn probability for the user of the service; and an action module that is configured to perform an action based on the determined churn probability. 
     
     
         7 . The system of  claim 6  wherein the probability module includes a churn calculator that is configured to analyze one or more behaviors associated with the user within a plurality of social networks and platforms (social media profile of the user), to compare user and his social media profile to clusters of other users, based upon similar social media profiles (“cohorts”); and to calculate predicted churn behavior of the user, based upon known churn behavior of cohorts. 
     
     
         8 . A computer implemented method of designing an efficient customer retention program for managing customer churn among customers of a business, the customer retention program including an analysis of the causes of customer churn and identifying customers who are most likely to churn in the future, so that appropriate steps may be taken to prevent customers who are likely to churn in the future from churning, the method comprising:
 a) identifying a social media profile of the customer;   b) comparing customer and his social media profile to clusters of customers, based upon similar social media profiles (“cohorts”);   c) calculating predicted churn behavior of the customer, based upon known churn behavior of cohorts; and   d) performing an action based on the predicted churn behavior of the customer.   
     
     
         9 . A computer-implementable method for predicting and delivery of churn signals for customers that are at risk of terminating their subscription and/or service to the customer retention units at the provider company, wherein the churn predictions are generated by analysis of full social media profiles of customers. 
     
     
         10 . The method of  claim 9  wherein customer loyal/disloyal characteristics towards services and subscriptions are used in the churn prediction. 
     
     
         11 . The method of  claim 9  wherein customer engagement with rival companies plays a factor in the prediction of churn signals. 
     
     
         12 . The method of  claim 9  wherein the influence of social networks on customers is incorporated in the prediction of churn signals. 
     
     
         13 . The method of  claim 9  wherein churn signals are utilized to prevent customers from canceling their contracts and/or subscriptions. 
     
     
         14 . The method of  claim 9  wherein a social media profile for a customer is comprised of all of their historical posts, blogs, status updated, communications, and general publicly available material on their social media accounts from the group consisting of, but not limited to, TWITTER, FACEBOOK, LINKEDIN, INSTAGRAM, WORDPRESS, and GOOGLE+. 
     
     
         15 . The method of  claim 9  wherein a social media profile may include private material on their social media accounts that is available to the social media accounts of the companies they have subscribed to but not the general public. 
     
     
         16 . The method of  claim 9  wherein life events are utilized in the prediction of churn signals. 
     
     
         17 . A machine implemented system that predicts and delivers churn signals to customer relationship management (CRM) software of service-based or subscription businesses for customers who are at risk of cancelling their services which comprises:
 a) a processor system that lives on the CRM software as a plug in;   b) a second processor that continuously monitors and processes SMPs of that company's customers to find new churn signals; and   c) a third processor with live communication between the first and the second processor and which delivers new signals from the first processor to the second as soon as new churn signals are predicted.   
     
     
         18 . The system of  claim 17  wherein the second processor and the third processor include a means to communicate with the first processor and automatically submits new churn signals to processor one in real time. 
     
     
         19 . The system of  claim 17  wherein an interface on the first processor shows new churn signals for the customers and includes reasons as to why such customers might churn and the probability of the churn. 
     
     
         20 . The system of  claim 17  wherein the second and third processor are the same. 
     
     
         21 . The system of  claim 17  wherein the second processor includes a churn signal application management and interface. 
     
     
         22 . The system of  claim 17  wherein the social media profiles are rendered from social media outlets selected from the group consisting, but not limited to, FACEBOOK, TWITTER, INSTAGRAM, LINKEDIN, and online blogs. 
     
     
         23 . A non-transitory, tangible computer-readable medium storing instructions adapted to be executed by a computer processor to perform a method for generating a customer churn prediction, for an entity in need of such prediction, said method comprising the steps of: extracting and receiving, by a churn prediction program executing on the computer processor, a variety of social media inputs; pre-processing the social media inputs to identify relevant social media posts, data trends and social network structures (pre-processed data); extracting and engineering features of the pre-processed data, such features comprising at least one of i) assessed social media postings, ii) assessed life events, iii) assessed engagement with the entity and competitors of said entity iv) assessed trend predisposition of customers to the entity based upon their prior churns, v) assessed one or more communities of customers to the entity and predisposition of the customers to the entity to churn based upon churn risk of the one or more communities; create feature vectors based at least upon i) to v); aggregating feature vectors into a database and creating churn model in the processor (churn model of aggregated features); determining, by the churn prediction program executing on the computer processor, predicted churn behavior of any one customer to the entity based upon, the comparison of at least one feature vector of the any one customer to the churn model of aggregated features.