Monkey Loves You: The Peril of Oversimplified Customer Feedback

What can a Black Mirror episode teach us about CX and NPS? Let’s dive into the troubling nature of minimizing the range of customer emotion to one basic metric.

If you’re not a fan of Black Mirror or haven’t been inclined to pay for a Netflix subscription, please bear with me, especially if you’re interested in customer feedback management.

The episode that stuck with me was the Black Museum episode in season 4, and in particular, one of the sub-stories: Monkey loves you. It lays out the unsettling scenario in which the protagonist’s late wife’s consciousness is transferred into a teddy bear-like mechanical monkey, through which she can only express herself using two basic phrases: “Monkey loves you” or “Monkey needs a hug.”

Through this concept, the episode explores how human expressions are limited and distorted through technology, reducing human connection to something mechanical and absent of any real depth. 

The intention of this episode was certainly not to address how we collect and analyze survey feedback, but it made me wonder: to what extent are we reducing our customer’s experiences in our analysis?

The NPS trap, promoter blindspot, and other survey analysis pitfalls

Net Promoter Score is a standard benchmark used by companies worldwide, making it a good way for businesses to gauge their performance compared to their competitors. However, the downside of this oversimplification is that it creates tunnel vision, also referred to as the NPS trap. Essentially, it warns us of the dangers of oversimplifying the complex landscape of customer experiences.

Too often, survey analysis centers around detractor feedback alone, overlooking the subtle frustrations hidden in otherwise positive metrics. This can lead to what we might refer to as promoter blindspot, where valuable insights from loyal customers go unnoticed simply because their scores are high. Often, these customers still describe friction, disappointment, or missed opportunities, but their voices don’t get the same attention. 

It’s important to recognize that dissatisfaction doesn’t always come with a low score, and loyalty doesn’t mean silence on what needs to improve.

We face similar challenges with our contact center data. Post-call surveys are often our main lens into the customer experience: CSAT scores, resolution rates, or even open-ended comments. But all of these come after the moment has passed. They offer only a partial view of what actually happened and even less insight into why.

What we tend to overlook is the most human part of customer service: the conversation itself. Contact centers are not just support operations. They are full of real-time questions, confusion, frustration, needs, and moments that matter. But how often do we truly listen? A customer might give a positive survey score because the agent was kind, while the real issue that caused the call is buried in the words of the conversation, not in the survey box.

How to listen to customers more deeply

Now that we’ve explored some of the blindspots and limitations inherent in customer experience data analytics when focusing solely on metrics, let’s delve into how we can overcome these challenges with Text Analytics. 

Text Analytics helps with structuring and prioritizing your unstructured data, the conversations themselves or the open-ended comments in surveys since they hold the richest insights beyond traditional metrics, really allowing us to listen to what the customer expressed. 

Topics 

Let’s start with the basics: the topics you apply to your unstructured data. A topic is essentially a set of keyword combinations, designed to capture specific comments within customer feedback. For example, a topic might flag phrases such as “the person that helped me was impatient with me”. This approach allows you to organize unstructured text into meaningful categories that we call Topics.

The foundation of effective text analytics lies in building a topic list that reflects your core business operations. Industry-specific topics help ensure that the insights you gather are relevant and actionable within the context of your products, services, and customer journeys.

Once a proper topic list is in place, the goal isn’t that you can still read it comment by comment. Instead, we organize it so we can help you surface quickly which topics require your attention. We highlight topic volumes, NPS averages, and metrics based on sentiment (which we’ll get into later).

Beyond these essential industry specific topics, topics with a unique angle to the data offer an additional layer of insight. For example, our Emotion Topics, Mental Crisis topics, and Customer Suggestion topics. These unique topics can help you tremendously trying to find valuable layers into your existing topics. Let me explain how!

Topic Co-occurence

The Topic Co-occurrence module in Medallia Text Analytics helps reveal how different topics or issues surface together in customer feedback. Rather than looking at topics in isolation, this tool identifies patterns where two or more topics appear within the same piece of feedback, either across the full comment or even within the same phrase. This level of detail helps uncover connections that may not be obvious through standard topic analysis.

For example, pairing your core, industry-specific topic list with emotional signal topics allows you to see which common problems, like product setup, billing, or agent support, are frequently associated with strong emotional responses. Your “Billing Ease of Understanding” topic might often co-occur with the Emotion topic “Anxiety.” You can also filter co-occurrence by sentiment, or even by speaker (agent vs. customer) when analyzing conversation data. This allows you to pinpoint whether a specific issue is being raised by the customer, addressed by the agent, or both, when analyzing your contact center conversations.

This kind of layered insight is key to identifying root causes, tracking emerging issues, and understanding how different experiences affect the customer’s overall perception. By using topic co-occurrence, businesses can make more targeted, informed improvements to products, services, and communication.

Sentiment Analysis 

Another layer of how unstructured data can be untangled is by using the sentiment model at the topic level.

Medallia’s sentiment engine evaluates the tone of each phrase in feedback, whether from survey comments, chats, or transcribed speech, and classifies them from strongly negative to strongly positive, with categories in between.

In Text Analytics reporting, this phrase-level sentiment is rolled up to show how each topic is performing on a sentiment level. For example, you might quickly discover that the topic “Find Products Online” has a negative percentage of 60%. 

But sentiment does more than highlight negative experiences. It plays a key role in helping you prioritize. One key metric is the Net Sentiment Score (NSS), which is simply the percent of positive sentiments minus the percent of negative sentiments. This score helps you quickly understand whether a topic leans more positive or negative.

The NSS is an extremely valuable and easy-to-use metric, but it doesn’t account for how often a topic comes up. To help you prioritize, especially when comparing a highly negative topic with thousands of comments to one with only a few, we use the NSS Impact Score. This metric combines the sentiment score with the volume of that specific topic, to calculate its overall influence on your Net Sentiment Score. In short, it tells you how strongly a topic is pulling your customer sentiment up or down — so you can focus on the issues that matter most.

Beyond the NPS 

Just as the Monkey loves you or Monkey needs a hug phrase in Black Mirror represents a limited expression of human sentiment, we must ensure that our analysis of customer feedback doesn’t fall into the same trap. 

It’s not enough to just understand that a customer “loves” us. We need to dig deeper to understand why they feel that way. Using the multiple text analytics analysis, we gain a richer, more complete picture of both our strengths and the areas where we can improve. 

The world of text analytics is incredibly rich, offering countless ways to listen more deeply and more meaningfully. What I’ve shared here only scratches the surface; with new AI capabilities like summarization and theme detection, we have more tools than ever to go beyond sentiment and uncover the real stories behind the scores. 

But the bigger picture here is that it’s not just about hearing that Monkey loves you; it’s about learning how to keep that love growing and evolving.

Psst…have you noticed that survey rates are rapidly declining? Here’s our guide on how to navigate this new reality.