Have you thought of using text analytics for surveys? If you ever had to read hundreds of customer responses to open-ended questions, you probably did!
We get it! Coding open-ended questions is a tedious task. Word clouds are an easy but poor alternative. And outsourcing coding can be a hit and miss. In fact, market researchers have been trying to avoid asking open-ended responses for years! Simply because dealing with customer responses is just too painful!
At Thematic, we’ve been helping brands to apply text analytics to surveys for a while. Here are the 5 key reasons why it is time to choose text analytics to find meaningful insights in customer responses.
1. Surveys are more common than ever
Surveys are the easiest way of getting customer feedback. Sign up for SurveyMonkey, fire up a Google Form, and you are underway. You can collect customer feedback quickly and easily. Net Promoter Score (NPS) surveys, in particular, are becoming a standard. Have a look at just how quickly NPS popularity increased over time:
2. Open-ended questions rule the surveys
Traditional multi-page surveys with scale questions lead to survey fatigue. Today companies keep surveys short, focusing them on a single metric. Open-ended questions are used to collect free-form answers from customers.
The advent of NPS surveys made open-ended questions a requirement. Two questions, “What impressed you?” and “What can we improve?”, offer insights into what drives people to become promoters or detractors, and are often more valuable than the scores themselves.
3. Manual coding isn’t practical
Customer responses, also called verbatims, traditionally have been analyzed by using human coders who create code frames from an initial sample, and then extend it with new codes, as new responses come in. Because running a survey used to take 6 to 12 weeks, it has not been an issue. But when thousands of responses come in overnight, coding them weeks later is impractical.
Also, code frames evolve as the business and the market changes. Re-applying them to the old customer responses is costly. Consistency becomes virtually impossible because it is such a subjective and bias-prone task. You would need to find a group of subject matter experts dedicated to coding responses over time. It’s challenging!
4. Text analytics for surveys is more accurate than ever
With advances in Deep Learning, algorithms get better and better at understanding the meaning of words and phrases. In fact, here at Thematic, we’ve shown that automated coding can be as accurate as human coding.
5. Text analytics is more affordable than ever
In the past, experts needed to train algorithms on labelled data. Companies like Medallia, Maritz CX, ClaraBridge still employ people to supervise this training process by hand. Of course, such approach makes automated coding expensive. But today, state-of-the-art algorithms just need raw customer responses. The find similarities, define common themes and group them hierarchically. Text analytics for surveys today is as easy as providing the data. No complex configuration or wait time required.
The best thing about the new text analytics approaches is that the themes emerge from customer responses, rather than being supplied by the company. You end up learning what customers care about and how they describe it in their words.
