Seeing the future before it happens
A wireless carrier in Europe asked Repustate to help build a text analytics component to integrate with their call center software
in order to track and analyze all interactions with customer
service representatives (CSRs). The idea was to predict when a customer,
after having a certain amount of negative interactions with the
carrier, would decide to leave for the competition.
The new process would be as follows:
The voice data of each call would be converted to text using speech-to-text software.
This text is then sent to Repustate for analysis. Repustate
analyzes each stored call for sentiment and semantic meaning,
looking for mention of specific products and services. An
overall sentiment score was tracked for each customer.
At regular intervals, a separate process would query the
customer database to see if any customers have sustained
negative sentiment below a certain threshold. If the
sentiment remained below a threshold for too long or the
overall sentiment score for a customer was too low, a text
message was sent to the customer apologizing for the recent
problems and offering a special promotion e.g. a discount on
next month's service.
Subsequent phone calls by customers would display historical
information on the CSR's screen which outlined the sentiment of
past calls, the issues discussed (which were determined based on the
semantic analysis of each call) and prompted the CSR to offer
promotions or to ask how the service has been recently.
The results were impressive. The chart below shows the carrier's
monthly churn rate compared to the industry standard.
Industry churn (%)
Carrier churn (%)