An economic lifeline: Text mining customer experience

The gap between what customers expect and what consumers experience with a firm is often expressed in the text fields of feedback mechanisms - which is where text mining can tap into this valuable insight.

By Tony Lopresti, Clarabridge

A company's success changes on the basis of consumer opinion on their products, service delivery and innovation, which is why text mining is increasingly being adopted by companies to analyse customer feedback and drive improvements.

Consumer feedback is expressed in surveys, call-centers, CRM systems, on the web and in other touchpoints. The gap between what they expect and what consumers experience is often expressed in the text fields of those feedback mechanisms, which if mined can provide great insight for companies wishing to secure their customer base in a tight economy.

Technology in action

In the past, companies made great advances using data mining and business intelligence against their structured data in CRM systems, transactional databases, etc. To manage the unstructured data (opinions captured in text), companies either resorted to lengthy and expensive manual processes on a sampling of the data or did not include the comments at all. Analysis on customer churn, service ratings and cost drivers was made without the context provided by the unstructured comments or were so delayed by their manual process to make the decision ineffective.

Text mining or text analytics is the high-powered engine that replaces the manual process. One financial services corporation reduced their 67 hour process of reading through only 1500 surveys to a 30 minute analysis of tens of thousands of surveys.

A text mining platform incorporates several key technologies that fall into three high level groups:

1. Collect and connect: Enterprises should establish listening posts across their business units and points where they interact with customers. A text mining platform's source connectors can harvest content from one, many or all of these listening posts. Analysts can have a single combined view of their surveys, CRM notes, call centre files, or can slice and dice to view a single source.

2. Mine and refine: Once the unstructured data is sourced, the next stage processes the text to extract meaning. Millions of verbatims get processed and stored in a data warehouse that links up to the accompanying structured data. Certain underlying technology engines come into play at this stage.

• Natural language processing converts the text into a format understandable by the software by extracting parts of speech, linguistic relationships, and other grammatical information.
• Categorisation automatically classifies the text verbatim into an unlimited number of categories based on the topics being discussed.
• Sentiment extraction reveals how customers think and fee by identifying the attitudes, perceptions and feelings as expressed in the feedback.
• Clustering bubbles up interesting themes to help analysts explore feedback beyond their category model.
• Entity and fact extraction identifies the named entities and associated facts (e.g. people, things, events).

3. Analyse and discover: Users of text mining platforms need to have interfaces that allow them to analyse and report on their findings; that provide a view and means to play with the stored text; and that allow the administrative configuring of the system. Within an enterprise, you'll find multiple sources of data and multiple consumers of the insight, so the analytical tool needs to be sophisticated and sufficiently robust. Text mining tools should integrate with business intelligence tools to take advantage of those underlying systems.

In the age of the consumer, every one has a story about good, bad or indifferent experiences with products and services and can easily call, fill out a survey or blog about it. A text mining platform empowers companies to use those stories to their advantage and turn consumer opinion into product, services and innovation improvements.

Tony Lopresti is vice president of marketing at Clarabridge