Use Text Analytics To Listen to Customers on Their Terms


Share on LinkedIn

It’s never been more important to really hear the voice of your customers. Surveys can help, but when was the last time you filled out a long survey asking you to rate your experiences with a product or service? Right, me neither.

Loyalty guru Frederick Reichheld contends that surveys should be as short as possible to make it easy for customers to participate. In fact, what not ask just one question: “Would you recommend us to your friends and colleagues?”

This approach, part of Reichheld’s Net Promoter Score (NPS) methodology, has come under fire from the research community, contending that Reichheld’s “one question” may not be valid in all situations, and doesn’t provide enough diagnostic capability. So you need to ask more questions to figure out why the customer was a “promoter” or “detractor.”

Or do you?

Intuit, an NPS advocate (promoter, even), uses a text analytics solution from Clarabridge to mine the written comments by TurboTax users, which come in via email, chat and the phone. Even if just a small percentage of TurboTax’s 10 million users complain, it creates an enormous volume of data. Without text analytics, Intuit missed the complete picture.

Sid Banerjee, founder/CEO of Clarabridge, launched the company in 2005 to glean insights from unstructured data typically found in documents. Natural language processing techniques are used to tag words and sentences, then the software tries to figure out the meaning of clusters of words, and does frequency counts for reports.

Text analytics has applications in most any industry. Clarabridge’s early customers have been in retail and hospitality businesses that want to understand customer feelings and desires, and dig out key issues driving changes in customer loyalty.

Survey data, such as captured by Enterprise Feedback Management (EFM) systems, is of course a good source for write-in comments to analyze. Segment customers by their NPS scores or other loyalty metrics, and you’ve got powerful diagnostics for text feedback.

But you could also analyze the notes from call center agents, and then see how different segments compare. According to Banerjee, Marriott slices customer’s text feedback data by their rewards program status.

One financial services firm learned that high and low net worth customers have very different issues and needs, which resulted in web site design changes. (Gee, I wonder if they have a segment for: “I wish I hadn’t bought that house at the market peak with a subprime loan.”)

And there’s more. What about all those posts in the blogosphere and online forums? They are a rich source of insight of what customers think and feel about products and services, and their underlying wants, needs and desires. Banerjee told me that one major CPG company is currently trying to do just that.

Text analytics seemed targeted to large enterprises, at least for now. Clarabridge’s solution starts around $100K and can go up substantially from there, depending on the data volume and number of users. A hosted solution is also available.

In a sign that the emerging text analytics market is maturing, BI giant Business Objects bought Inxight Software earlier this year. Analytics leaders SAS and SPSS offer text mining solutions too.

What do you think? Is text analytics the answer to really understanding customers on their terms, rather than just relying on structured feedback data? If you agree, press 1. Disagree, press 2. To add your comments, just type them in below!


  1. This has been shaping up as an exciting topic this year. As you mentioned, organizations today capture volumes of unstructured data – words in the form of customer comments, survey or claim forms, service records, technician notes, emails, public records and various other documents. Corporate strategy is all about collecting information from many different sources, evaluating the probabilities of potential outcomes and then making intelligent decisions concerning the direction of the business. Yet, many organizations have remained for the most part, indifferent to their unstructured data sources (considered by many to represent over 85% of an organization’s data) that often reflect the true voice and sentiment of their customers, suppliers and employees.
    In the past, little was done to use this data to better understand the business insight it contained because it fell outside of the well-established structured data filters. I was recently involved with Aberdeen research that shows organizations are now starting to prioritize the use of unstructured data. AberdeenGroup conducted an in-depth research survey that benchmarked companies’ current and planned direction toward the integration of unstructured data with traditional structured information. The findings from Aberdeen’s July benchmark report – “Data Management 2.0: Making Sense of Unstructured Data” show that the top 20% of companies who are achieving increased performance (the “Best-In-Class”, or “BIC”) have several characteristics in common, and have found added value from previously untapped information sources.
    In fact, 24% of those classified as best-in-class (BIC) rated the convergence of unstructured and structured information as one of their “top priorities” while 56% of BIC ranked it as a “high priority”. Best-in-Class companies are 55% more likely to have text analytics in place than the industry average. In addition, BIC companies are 33% more likely to have federated search technology in place than the industry average. By incorporating solutions for both search and analysis of unstructured information, the BIC have improved employee productivity and customer insight, as well as reduced risk. Additional findings from the research support the following improvements sited by BIC companies:

    • 68% reported greater than 10% improvement in time spent searching for knowledge objects (employee productivity).
    • 94% reported greater than 10% improvement in response time to customer demands.
    • 66% reported greater than 10% improvement in the ability to reduce risk by preventing harmful events before they occur.
    • 65% reported greater than 10% improvement in customer insight (“view of the customer”).


Please use comments to add value to the discussion. Maximum one link to an educational blog post or article. We will NOT PUBLISH brief comments like "good post," comments that mainly promote links, or comments with links to companies, products, or services.

Please enter your comment!
Please enter your name here