How Text Analytics Adds Value to NPS February 20, 2019 - Blogs on Text Analytics

 

 

Companies are always trying to find easy and accurate ways to measure customer satisfaction or customer sentiment.  In the last decade Net Promoter Score (NPS) has emerged as a favorite.  Many companies have adopted it as a standard. The one-question survey developed by Fred Reicheld, Bain and Company, “How likely are you to recommend this product/brand/service to your friends or colleagues,” is straightforward, as is the 0-10 scoring system that accompanies it. Companies can ask the question and over time see how their score moves up or down. Thousands of companies have adopted the NPS system. There are detractors to NPS to be sure. No system is perfect. But many studies have shown there is a correlation between NPS and customer sentiment which in turn can influence revenue and other growth indicators. However, even the creators of NPS agree the score isn’t complete by itself. You need a verbatim to go with it. “Why did you give us this score?”This is what allows you to make NPS actionable. This is what gives you the information to make changes to further delight promoters or to help move neutrals or detractors towards becoming promoters of your brand.

More than a score, with Text Analytics NPS is a learning tool to improve your business

Let’s look at an example of how you can learn from your customers by analyzing an open-ended question attached to the NPS survey. Let’s say Jane, a 35-year-old mother of two from Chicago is asked how likely she would be to recommend your energy bars to a friend or colleague and she gives you a 9 score. This is excellent and your marketing people are thrilled the NPS has increased, job well done. When asked in an attached question why she says the bars are well priced and use tasty, natural ingredients which are two factors that are very important to her but she was a bit disappointed the packaging didn’t include recycled paper like your competitor and the bars were a bit hard to find because they were placed so low on the supermarket shelf at her neighborhood store.  She raises two issues that deserve consideration and follow up. You can send people out to investigate what is happening with your shelf placement and how to improve it and you can consider doing additional research to see if recycled packaging is something that makes sense for your business and could increase sales. It is also possible that Jane has alerted you to a possible problem. What would the reaction be if the recycled packaging comment went on social media? Might it cause you some PR problems and override the otherwise positive comment? One of the people you could contact is Jane and ask her about this as a follow-up. The extended benefit is it closes the loop with the customer and likely increases Jane’s loyalty to your brand.

How Text Analytics adds value

How can you make sure you capture comments like Jane’s and everyone else who responds to the NPS survey? If you have thousands or tens of thousands of NPS responses it is tempting and happens all too often, that you just take the NPS score and ignore the comments or perhaps just look at one aspect such as detractor or take a sampling of the open-ended responses. If you use text analytics you can analyze all the comments; detractor, neutral, and promoter. You can find actionable items, test your pre-determined hypothesis, measure change over time, and discover new issues or even potential innovations you hadn’t thought about.

Qualitative Coding of your NPS verbatims

Using a qualitative analysis tool like QDA Miner you can create themes (a codebook) and code/tag the comments of the open-ended questions associated with your NPS. To establish the themes you can start with your own experience as to what you expect to be the drivers of your business. In the example of the power bars these might be price, freshness, taste, ingredients, etc. Other products or services would have their own set of themes. You can then code or tag all the responses based on those themes and create new ones that are revealed by the customer comments. Environmental impact could be a new theme of which packaging would be a code or tag and another might be the source of the ingredients. You can get as granular as makes sense for what you are analyzing. You can then compare the codes, to see the most frequent drivers for your customers. For example, what are the top 5 detractors and the top five promoters and how do they relate? You can look at how the codes relate to each other and how they relate to your NPS data. People who rate you 0-6 tend to be talking about these five issues how does that compare with your net promoters. You can reuse and tweek your codebook with each survey as you get more data and see how the comments are changing in relation to your NPS data and the actions you have taken to make improvements, introduce new products or services, redesign your website, etc. You now have a learning tool, not just a score.

Exploratory Text Mining and Automated Categorization of your NPS verbatims

The example above works well if you are processing hundreds or a few thousand comments.  But what if you are dealing with tens of thousands or maybe millions of comments? You need a more automated process because while computer-assisted tagging will help speed up the process it isn’t designed to process extremely large volumes of responses. You will have difficulty discovering what you don’t expect to find or learn what you don’t know. This is where an automated text-mining tool like WordStat can be used. You can take a few approaches. The first is, if you have started as described above, you can take the themes and codes you built-in QDA Miner and use them as the basis of a content analysis dictionary in WordStat.  You can enlarge your dictionary from there, test the coverage etc. If you are starting from scratch with large volumes of open-ended questions you can do some exploratory text mining of the text. Find out what are the most common words and phrases in the comments. Use the topic modeling feature of WordStat to identify the top 30, 40, 50, topics your customers are talking about. This is also where you might be able to discover issues you didn’t expect or innovative ideas. From here you can build a categorization dictionary around the themes or topics. This dictionary can be used each time you do an NPS survey on a similar product or service so you can compare how the themes relate to previous comments and to your NPS score over time. You can also build predictive models by determining what promoters say when they like or dislike something. This will allow you to get a sense of what might be happening to your NPS if you are using the dictionary to analyze other formats such as longer customer surveys, social media comments, etc.

Tell Your Customer Experience Story

Numbers on a chart that move up or down over time are fine for showing snapshots but they don’t tell the whole story. Another feature of text analytics (QDA Miner & WordStat) is the ability to use a wide range of charts, graphs, comparative analysis formats to let you visualize your results. This allows you and your audience to see the data in context and see how the quantitative and qualitative data relate to each other, how it changes over time, and is impacted by different cultures, geographies, ages, and gender. You might be saying, this sounds like a lot more work. I need to hire more people or develop new skill sets in my organization. You are probably right but the ROI is there. It has been shown that higher NPS can lead to fewer returns, fewer repairs, lower customer churn, greater revenue but it is very hard for NPS to do it alone. To reach its full potential NPS needs you to ask the why and those comments need text analytics to deliver the answers.