Call Us @

(212) 608 6307

Case Study - Customer Behavior Prediction

Business Problem

The ROI - Reduced I.T. Costs, Increased Security, Performance and Mobility

In Insurance industry, executives calling prospective customers face a problem of how to gauge customer sentiment even if qualitatively that will lead to the customer buying an insurance product.

The executive needs to make sure that the customer remains in a positive frame of mind for the buying to happen. If the executive can somehow get to know the sentiment of the customer as the customer is speaking, the executive may be able to handle the customer well if the customer is going into negative frame of mind by making the customer special customized offers. The executive can put the customer’s mind at ease by replying to the customer’s doubts properly.

Possible Solution

The problem of gauging the customer sentiment can be solved by creating a tool which allows us to know the sentiment of the customer. The sentiment has to be captured from the customer’s voice in real time so that it may be useful to the insurance executive.

Sentimeter1

There can be two possible solutions – one approach can be that the voice of the customer is converted to text on a real time basis so that we may capture the sentiment from the text and the other approach is to analyse the features of the customer’s voice to capture the sentiment directly from his voice.

There can be challenges to both the approaches. Voice to text conversion may face the challenge of being accurate enough and fast enough for it to be real time basis. Direct voice to sentiment may face the challenge of being able to distinguish between the various emotions employed by a human being when speaking.

So, the person’s voice has to be the input and the output may be sound features or the text of his speech.

Outcome

Card image cap

We have created Sentimeter which is a tool to know the sentiment in terms of positive or negative scores (-1 to 1) in real time as one speaks in the tool. The tool takes a recording of voice at every 10 second interval and converts it to text.

For instance, if the customer response sentiment is showing a negative score, the service agent can change his talk to pacify the customer and turn the customer sentiment to positive through certain offers.

This text is then converted to sentiment polarity i.e. sentiment score from -1 to 1 as the sentiment may be negative or positive. This tool has been specifically created to help customer service or marketing agents in any industry to gauge the sentiment of the customer they are talking to and adapt their response to the sentiment score.

Case Study - Service Quality Control

Business Problem

In any customer care centre or call centre, a supervisor always faces the problem of how to gauge performance of an executive and adapt a training programme to suit the needs of the executive to better the executive’s performance while dealing with prospective or existing customers.

Problem Description

The problem is how to design a training programme to assess the training needs of an call centre or customer care executive to allow him to better be able to judge the customer sentiment and adapt his responses to the customer sentiment.

First it boils down to how to assess the customer’s sentiment or emotions from his voice in real time so that appropriate responses can be generated. One approach to the problem is to capture the customer’s voice and convert it to text to gauge the sentiment.

This conversion will allow the executive to better understand the sentiment variation in customer’s voice to handle it properly. The customer’s voice text should be recorded for training purposes. Next, the executive voice or responses should also be converted to text to record it for examining it later on. Both the texts can be analysed to judge which words were repeated frequently or which words were conveying the intent of the speaker.

Sentimeter

Text analytics can provide the basis for a training programme based on those words which when spoken by the customer led to lead generation or buying of services. The customer care or call centre executive may be trained to speak certain words frequently which have led to customer satisfaction in the past. Text analytics of the customer’s voice as well as executive’s voice will provide plenty of data to fish for answers to the training design problem.

Possible Solution

The solution to the question of gauging the performance of an executive in service calls and designing a training programme to specifically address the needs of the executive has to be answered by an approach to have texts of the customers voices so as to identify which words lead to more positive sentiment and hence better customer satisfaction. We have to capture the text of executive’s voice as well to assess the gaps between the present performance and ideal performance of service calls.

Outcome

We have designed a tool called Sentimeter which allows us to capture and record the text of customer’s voice as well as executive’s voice. Sentimeter uses text analytics to provide the list of frequently used words by the customer pertaining to both positive as well as negative sentiments.

Sentimeter can convert the voice to text and record it for examination later on. It can do text analytics on the text to give you 25 most frequently words from the text. It can give you the list of positive and negative sentiment words by comparing the text with the recorded graph of customer’s sentiment variation in qualitative terms. This can go a long way in helping the supervisor to assess the training needs of an executive and address it specifically.

Past records of successful calls may be used to know such words of customers which led to a satisfied customer or call resulting in lead generation. Same can be done for executives call records to get the words most frequently used by successful executives.