Like most kids, I cared a great deal about my school grades when I was growing up. At the time, they felt like the ultimate measure of success or failure. In retrospect, I know that grading is somewhat subjective. One teacher’s “A” may be another’s “B-minus.” But without qualifiers, there was no way for me to understand why I received a specific grade or how to improve my performance—or replicate what I’d done well.
Understanding the ‘Why’ Behind Scores
As a data scientist specializing in customer experience, I know that customers are a lot like teachers. Some are tough graders—others are a little more forgiving and liberal with high marks. Call them what you will—scores, structured data, or metrics—but without context, their value is limited. For that, you need unstructured, open-ended comments: “human” data. And in addition to giving you the “why” behind a particular score, human data in the form of survey comments, social reviews, voice and video transcripts, etc., provides a treasure trove of intelligence that can inform decisions across your entire enterprise.
In the past, data science skewed heavily toward quantitative data, driven by both technology innovation and limitations. We’ve had the tools to collect increasingly more information, but at the same time, lacked the chops to analyze unstructured data.
Today, things are different. Analytics tools for unstructured data, when properly tuned, can reach accuracy rates upwards of 90 percent. I see some brands really pushing the envelope and doing predictive—and even prescriptive—analysis on this data, blowing accuracy and actionability rates out of the water. What that means for CX analysts: if you stake your career on scores alone, then you’ll find yourself at a strong disadvantage on the CX battlefield.
Fifteen years of collecting and analyzing customer feedback have given us a huge body of data. So, based on the words customers use (both the order and how they’re modified), we can now run an in-depth sentiment analysis to understand just how frustrated (or pleased) customers are. We can quantify their emotions—a factor once seen as “soft” in the data science world. However, thanks to more sophisticated analytics techniques on this human data, we can now actually transform soft factors into hard, quantifiable intelligence.
For example, let’s look at a rental car client. By exploring its experience data, we found that 1,000 customers mentioned “dirty” while another 200 mentioned “slow.” With five times the amount of mentions, we could assume that car cleanliness should be a priority for the company. However, by understanding the sentiment in the larger context of the comment (paired with the quantitative data), we find that the customers who mentioned “dirty” were actually not very angry, while those who mentioned “slow” were ready to sever the relationship forever.
Analysis of unstructured data allows companies to prioritize efforts likely to have the greatest impact on customer satisfaction and, ultimately, the bottom line. When we fixate on scores, we have to make assumptions and focus on the drivers we think matter to our customers. Just like grades on a report card, scores are only part of the story and don’t promise sustained success. To achieve context—to understand how to improve deficiencies and replicate what’s already working—we must take advantage of the technology and capitalize on the value of human data.