Surveys make Text Analysis Better

A few months back I wrote a post titled: "Text Analysis makes Surveys Better" (click here to read the post). The genesis for the post was my perception that organizations needed a better way to collect customer feedback than just long and involved surveys. Among others, I made three main points in the post: 

• Functional customer surveys (those short enough to have high response rates) rely upon open answer comments for key insights into the customer experience. 
• Comment categorization and analysis therefore is critical to a successful process. 
• That for low volume processes (under 1000 comments / month) analysis can be a manual process. But, in higher volume surveys automated verbatim analysis adds a lot of value.

As businesses increasingly employ transaction based feedback processes they are coming to rely on customer comments almost entirely for insights. Social media is a key driver for this phenomenon as it promotes a "quick hit" type of process (i.e. select a star and make a comment). Some businesses have implemented single question, transaction based surveys using Net Promoter. Not surprisingly text analysis tools are being used to gain insight on all these comment streams.

As a result, many businesses have moved away from "research" oriented customer surveys, choosing instead to use single question Net Promoter / Customer Satisfaction surveys with open answer comment fields. In effect, these businesses have chosen to rely on verbatim feedback analysis, almost exclusively, for generating insights about their customers. 

This kind of feedback management approach has the advantage of being simple to implement and can be effective for insight generation when feedback volumes are small. The NPS metric, or satisfaction metric for that matter, provides base level context for feedback interpretation and analysis. For instance, topics with negative sentiment in comments coming from detractors are generally assumed to have some impact on NPS scores. Where feedback volumes are small, time can be taken to validate the "truth" of that assumption. Without going into lots of depth on this, in my experience the things people talk about in their comments (topics) are often same across NPS categories (i.e. Promoters often experience many of the same issues that detractors experience). So, validating "truth" associated with comments is quite important to building improved processes. NPS or CSAT by themselves are typically not enough to ensure this, as they don't by themselves provide enough context to the feedback.  

However, when feedback volumes expand in different ways the need for additional context to customer comments also expands. Some examples:  

• Differences in regional or country specific comments 
• Operational differences about how customers are handled (i.e. which call center handled the customer) 
• Does the same NPS or Satisfaction scale even apply across regions or countries?

Its easy to see that a simple sort of feedback process could be problematic when comment volumes rise and interpretation complexity increases.  Some things ameliorate these challenges, at least to a degree. Automated text analysis solutions, for instance.  Text analysis tools (www.etuma.com) deal quite effectively with high volumes of comments. And, if there is background data behind the surveys (for region or country for example), these tools can use the background data to provide additional context and better analyses. 

But, even in a scenario where automated text analysis is applied to single-question NPS surveys, and background data is available, there is often a need for additional context in order to understand how to best take action on feedback.  Some types of additional context include: 

• Expectations - What is reasonable vs. unreasonable in the customer's mind for any given challenge highlighted in their comments?
• Alternatives - Are alternatives available to customers either from the business itself or competitors?  Are alternatives reasonable if available?
• Costs - Are customers willing to absorb higher costs for improved processes
• Business opportunities - Would more customers actually recommend if problems or issues are better dealt with? Would they buy more? Or more often?

These are the types of contextual "truths" that must be learned via an interactive process with customers. Customer surveys (www.questback.com) are by far the easiest and lowest cost way of getting this type of data.  

The value add of driving customer insight generation from customer feedback, in my view, is substantial. First, a lot of data becomes available to the insight generation process because of the feedback process. This enables insight generation to be a short easy follow-up survey to the initial feedback survey (which was itself short and easy). With the automation available today via APIs filtered data can emerge from the feedback process and be used to trigger insight generation.  

Of course, a process of automated feedback, automated text analysis and automated insight generation requires a single, or group of, integrated system(s). The system(s) would of course require some kind of Analytic "back end" to help make sense of all the data. I am currently working with customers who are putting together this kind of optimized feedback gathering / data analysis / insight generation process. The platforms my customers are using are relatively low cost and are easy to use. So, businesses that want to improve their processes by using more automation for feedback, analysis and insight can do so without breaking the bank, or disrupting their operations.

At the end of the day I find it fascinating how businesses are changing the way they gather and analyze customer feedback and generate insights from it based on technology.

Stewart Nash

LinkedIn: https://www.linkedin.com/in/stewartnash