Getting accurate, actionable signals from unstructured datasets like reviews and contact centres is very challenging. Using key word analysis to discern topics and sentiment only works well at a high level or where accuracy is not important. PrediCX makes machine learning easy, providing accurate insight across actionable topics. Labels such as ‘Hotel -Dirty’ carry inherent bad sentiment, and highlight a clear action to take to increase customer satisfaction. Furthermore, these comments could be buried in an otherwise positive review, limiting the use of traditional sentiment analysis.
Below are the topics discovered and labelled from live, publically available Expedia reviews. You can explore the graph below. Use the buttons above the graph to de-select labels with positive and negative connotations, respectively. The scale below can be used to shorten the time period, and individual labels can be selected and de-selected:
In the graph below, we can see when customers wrote in a review that they were leaving Expedia or not using again. These records can be drilled into in PrediCX, allowing for the reviewer to be contacted. Guest recovery, over the long term, is far cheaper than attracting new customers, and should be of the utmost priority regarding customer churn:
PrediCX works in near-real time. In the graph below, we can see what customers are talking about right now. This allows for the implementation of an early warning system, along with triggers and alerts to allow users to actively combat emerging issues before they grow organically:
© 2018 Warwick Analytics. All rights reserved. Registered in England & Wales. Number 07724630. Registered address 35 Kingsland Road, London, E2 8AA. VAT 120435168.