Share this content

Customer feedback breakthrough: Time for text analytics?

1st Jul 2011
Share this content

Dan Lee explores how text analytics is allowing customer experience management to view and act on structured and unstructured feedback together.

Today’s successful hospitality companies are typically well versed in guest satisfaction and customer loyalty programs. Many frontline staff members are experts when it comes to closing the loop with customers, while managers know just how to tie quantitative performance metrics to ROI.
But consider this situation. A hotel manager notices that her branch has been receiving low quantitative scores for food quality on the standard guest satisfaction follow-up survey. Does a 2 on a scale of 1 to 10 actually tell her anything about the root cause of the issue?
The 2 out of 10 could signify:
  • Bad-tasting food
  • Unhealthy food
  • Cold food
  • Food that's inappropriate for kids
  • Not enough food
  • Limited menu choices
… and many other qualitative attributes of the food service.
Some businesses choose to address this issue by adding several follow-up questions or a “check all that apply” question, allowing customers to select from a list of possible causes of a negative rating. But what if the goal is to discover new sources of customer frustration? A “check all that apply” question does not serve this need.
If the original quantitative question on the survey were followed with an open-ended question asking customers to elaborate on the chosen score, the hotel manager would have a much richer data set for evaluating the problem and choosing how to address it - though she would still have to find a way to sift through all the responses. More on that later.
In short, only with verbatim comments can businesses find out what's really driving scores up or down.
Quantitative scores are undeniably useful. They provide a quick overview of a problem and are instrumental in tracking progress on particular metrics over time. The key is to view and act on structured and unstructured feedback together.
Text analytics: how it works
That’s much easier said than done. There are many reasons hospitality companies - and many businesses in other industries - started out using CEM to track only quantitative metrics. Open-ended feedback requires a much more concerted effort before it yields any value.
The old-school method is to hire workers to manually tag each response in a spreadsheet. There are several issues with this approach. It’s:
  • Not scalable
  • Not consistent
  • Not thorough
Luckily, new technology automates and simplifies the process of reading and analysing each and every response. Sophisticated text analytics engines do so using natural language processors, which are able to parse millions of sentences to determine which parts of speech relate to each other and thus tie positive or negative scores to specific triggers. For example, a good text analytics platform should be able to tell that a person who writes, “The service, while I was expecting it to be good, was not wonderful,” was in fact complaining about service quality.
Many text analytics systems have built-in software that can correctly interpret misspellings and identify common synonyms as well. For example, if a person searches for comment relating to “poor service,” the platform should provide the option to select common synonyms like “bad” or “lousy” and input other case-by-case synonyms less obvious to a machine, such as “slow.”
The best solutions go even further. Look for advanced technology for analysing the impact, positive or negative, of each comment (and even each word) on overall satisfaction scores. Some issues may be mentioned frequently but have little to no impact on guest satisfaction. Finally, the solution of your choice should not only be capable of processing an enormous amount of unstructured data, but also present the results quickly in an intuitive, user-friendly way.
Case study
A leisure hospitality customer used text analytics to fill two needs, one at the corporate level and one at the frontline.
The standard guest survey for this company includes the following open-ended questions:
  • Tell us about a positive surprise you experienced on your trip.
  • Tell us about a negative surprise you experienced on your trip.
Program managers were very surprised at the volume and detail of responses for the negative surprise question. The amount of information that came in far exceeded the staff’s ability to handle it manually.
A text analytics solution helps this company take action on the data in two ways:
1)      Taking advantage of the intuitive interface, frontline staff members are able to quickly delve into responses looking for specific situations they could address for the next trip. If several people complained about a shrimp dish, for instance, the company would simply remove it.


2)      Corporate executives are able to discover issues across multiple trips and properties to find patterns and systematic root causes and solve these issues using process changes or staff retraining. For example, if a particular stop on a tour is less than satisfying, the executives and their teams can look into finding a replacement that is more exciting, kid-friendly, picturesque—or whichever quality survey respondents found most lacking.
The business benefits of this two-pronged approach have been evident for this hospitality company. By looking to eliminate negative surprises, the staff has effectively improved guest satisfaction and increased guests’ likelihood to rebook.
Key takeaways
  1. You don’t have to be a data wonk to use text analytics if you select the right solution. Text analytics solutions exist that can run complex algorithms on the back end with simple interfaces and reporting for anyone to use on the front end.
  2. Put your structured and unstructured feedback in one place. Quantifiable data is imperative for noticing and tracking operational trends and impacts. But the verbatims will uncover emerging themes that can provide greater insight into the root causes than the numbers can.
  3. When choosing a text analytics platform, look for scalability and an intuitive user interface.
  4. Text analytics software’s potential is virtually limitless. Aside from survey verbatims, you can use text analytics to analyse other sources, such as social media feedback, emails, chat sessions, contact centre notes, and more. Stay tuned.
Dan Lee is the senior director of product solutions for Medallia, Inc., a CEM vendor headquartered in Palo Alto, CA. He has over 20 years of product management and engineering experience.

Replies (0)

Please login or register to join the discussion.

There are currently no replies, be the first to post a reply.