How to derive customer insights through text analytics
As business people, we’re often trying to figure out: what do customers want? How do they perceive what we deliver them? We typically answer these questions by collecting customer feedback. But collecting the feedback is just the first step towards finding the answers. The second is to analyse the data. In this article, you will learn how to choose customer insight services or a Text Analytics solution to turn free-form customer comments into meaningful customer experiences.
What is Text Analytics?
If you’ve ever read hundreds of customer comments, you will know that a proper analysis is not that simple. The traditional approach is to use manual coding, where a person reads each comment and tags it with categories or themes. Text analytics automates this process with algorithms that interpret the meaning of words and phrases, by using statistics and linguistics.
Text analytics helps companies make sense of written customer communication at scale, whether via email or web surveys, SMS, social media, web chat, online forums and review sites. The main goal of this technology is to transform spreadsheets of raw customer comments into customer experience insight.
For example, the image to the right shows Google’s summary of customer reviews by theme and sentiment. The online store Dick’s Sporting Goods has great selection and communication. But they struggle with customer service and customers don’t like their return policy. The company knows straight away what should be the focus.
Unfortunately, many customer experience platforms that collect customer feedback still only visualise free-text data as word clouds. The following example shows why you should instead use proper Text Analytics.
How the right Text Analytics solution could have saved millions
Some of you may remember how, in 2015, the New Zealand government attempted to change the national flag. They asked New Zealanders: “What do you stand for?” After analysing this data, the agency running the survey presented this:
But after reading the actual comments on the StandFor website (now offline), this word cloud didn’t add up. By applying Text Analytics, this is what we discovered:
Turns out, most New Zealanders who responded did not talk about equality and freedom. Instead they talked about “keeping the current flag”. Following this initial survey, millions were spent on campaigns and design competitions, followed by two referendums which resulted in New Zealanders keeping their flag.
Collecting data is not enough. If you don’t pay attention to the analysis of free-text customer comments, you can easily end up wasting time and money.
Three Simple Questions for deciding on the best approach
Here are three simple questions to ask a Text Analytics vendor for your customer experience programs:
1. Is it easy to set up?
What is required from you to get started? Do you have access to the right resources? Some solutions will require manually tagged customer feedback to train and test the algorithms. If all you have is a spreadsheet with customer comments, it will complicate the set-up of such solutions. Others may require creating rules for categories you care about. Again, this is a time-consuming process. Recent leaps in technology mean that modern solutions can offer much easier set-up and maintenance.
2. Is it flexible?
How well will the solution respond to changes in customer feedback? Can it capture new themes as they emerge? For example, customers may be responding to a new feature the company released or they may start mentioning a new competitor. The market never stays still, and customer expectations are also always on the rise. An effective Text Analytics solution needs to account for this.
3. Is it accurate?
Finally, as the New Zealand flag example shows, the most important thing about the analysis is the accuracy. After all, the goal of customer feedback analysis is to make decisions based on what customers are saying. Fully automated solutions tend to make several types of mistakes. Here is what you need to watch out for, and how you can do control quality:
| Common mistake in automated analysis | How to ensure the accuracy of a solution |
| Only one of the themes a customer mentioned is found | Avoid solutions that only provide one theme per comment. |
| Customer comment is tagged with an incorrect theme | Make sure to read a good sample of comments assigned to several themes. |
| No themes are found in a customer comment | Make sure to find out how many comments were assigned to the Other category or not given any category and expect to see around 20%. If it is 30% or more, read a sample to verify why. |
For a deep dive into the accuracy of automated solutions, check out Maurice Fitzgerald’s book Net Promoter – Implement the System, with several sections dedicated to Text Analytics.
Finally, make sure that the results of the Text Analytics are provided as simple and easy-to-interpret reports. You will have additional questions about slices of your data, such as different customer segments, or time periods. So, the report should be interactive and have a user-friendly interface.
If you collect hundreds or thousands of customer comments per week, Text Analytics tools will help you make sense of this data in a scalable way. Now you know how to cut through the technology hype and choose a solution based on your needs. Remember: Setup, Flexibility, and Accuracy!
Get a free demo on your data to see what insights Text Analytics can find for you
Related content:
How to automate coding of Net Promoter Score customer comments
How small call centres can achieve big results with speech analytics-as-a-service
Last updated on: October 26, 2022
