The short answer
High NPS, satisfaction and likelihood-to-recommend scores can hide relationship risk because they capture how someone rated an experience, not the complete emotional experience behind it. Across more than 350,000 comments analyzed by Emtherical, nearly 3 out of 10 people giving the highest ratings also expressed a negative emotion. What appeared to be success sometimes required intervention.
When filling out her patient survey, she gave the hospital the highest possible likelihood-to-recommend rating.
By every traditional measure, she was a success story. Her response would lift the score reported to leaders, reassure the care team and contribute to the conclusion that the experience was working as intended.
But she also wrote about feeling anxious.
Her gratitude was real. So was her anxiety. She was not contradicting herself, and she had not failed to tell the organization how she felt. She gave them both parts of her experience.
The score simply made one part much easier to see than the other.
When the score becomes the story
Organizations rely on NPS, Overall Satisfaction, likelihood-to-recommend ratings and star ratings for good reason. These measures make complex experiences easier to track. They allow leaders to see change over time, compare performance and focus attention.
But over time, the score can become more than an indicator. It can become the story.
A Promoter becomes a loyal customer. A five-star review becomes an unqualified success. A top-box patient becomes evidence that nothing needs attention. The person, with all the texture and complexity of their experience, disappears behind the classification.
That is a risky trade. The simplicity that makes a score useful is also what makes it incomplete.
What 3 out of 10 reveals
At Emtherical, we found this pattern across very different kinds of experiences. Thirty-four percent of patients giving the highest likelihood-to-recommend rating also expressed a negative emotion. So did 27 percent of financial services NPS Promoters and 26 percent of consumers leaving five-star Amazon reviews.
Across more than 350,000 comments, nearly 3 out of 10 top ratings contained something the rating alone could not reveal.
What looked like success sometimes required intervention.
For the patient, anxiety might reflect uncertainty about what happens next. For the financial services customers, frustration might point to a recurring obstacle that loyalty has not erased. For the five-star reviewer, disappointment might reveal an expectation that went unmet despite genuine enthusiasm for the product.
These people were not hidden because they withheld their experiences. They were hidden because organizations had learned to treat the score as the clearest—and sometimes the only—signal that mattered.
The negative emotion did not automatically mean the relationship would end. It meant the score alone could not tell the organization whether the relationship was strengthening, under strain or beginning to break.
Emotion changes what the score means
A positive score tells us how someone rated an experience.
Emotion helps us understand what the experience meant to them.
That distinction matters because experiences do not end when a survey is submitted. Experiences accumulate. Each interaction can make a deposit into a relationship or create a withdrawal from it. Gratitude and caring can deepen connection. Frustration and anxiety can place the relationship under strain. Disappointment can signal that an expectation or promise was broken. Aversion or loathing may reveal a rupture that is far more difficult to repair.
Over time, those emotional experiences influence behavior. They shape whether people return, recommend, forgive a future mistake, reduce their involvement or quietly walk away. A score offers a snapshot. Emotion helps us see the relationship taking shape beneath it.
That is why two people giving the same rating may be telling us very different things. One may feel genuinely attached. Another may be staying despite repeated frustration. A third may be satisfied today but carrying a disappointment that changes what they do tomorrow.
If every positive score is treated as the same kind of success, those differences remain invisible until they appear later as churn, declining trust, a lost customer or an experience that suddenly seems to deteriorate.
Emotion guides the response
Understanding emotion does more than reveal hidden risk. It helps organizations respond appropriately.
An anxious person may need reassurance and clarity. A frustrated person may need an obstacle removed. Someone who is disappointed may need the organization to acknowledge what went wrong and correct it. A person expressing gratitude may deserve recognition of the people and practices that created that feeling.
The score alone cannot guide those choices. A positive rating accompanied by gratitude may be an experience to celebrate and replicate. The same rating accompanied by frustration may require attention. Accompanied by disappointment or aversion, it may call for intervention before the relationship deteriorates further.
This is the difference between measuring an outcome and understanding the human experience that produced it.
The risk of teaching AI the same blind spot
This concern becomes more urgent as AI takes a larger role in organizational decisions. If we build intelligent systems around the metrics we have always prioritized, we risk automating the same incomplete view of people—only faster and with greater confidence.
A system that sees a Promoter may conclude that no action is necessary.
A system that recognizes the frustration inside that Promoter’s experience may reach a very different conclusion.
The opportunity is not simply to process more data. It is to carry more of the human experience into the decision: the emotion, the context, the strength of the relationship and the response the moment requires.
Finding our humanity
Emtherical began with a purpose: thoughtfully using AI to understand and empathize with people. Our starting point was not the technology. It was the growing sense that the business world had become so full of data that people were disappearing into the mist.
That concern feels more urgent now. AI will become part of more conversations, more experiences and more decisions. It will know more about us, speak more fluently to us and increasingly act around us. The challenge is to ensure that, as the technology becomes more capable, the human being does not become less visible.
Human understanding must become part of how decision intelligence is built.
That means carrying the emotions, context, relationships and lived meaning of an experience forward into the choices an organization makes. Not because empathy sounds good, but because decisions made without people in full view are decisions made with only part of the truth.
The future should not be about making people more legible to machines. It should be about using machines to help us remain more attentive to people.
If we get that right, AI will not make our world less human. It may help us rediscover what being human asks of us: to listen more closely, to resist easy conclusions and to see one another more fully.
Related questions
Can an NPS Promoter still be at risk
Yes. A Promoter classification reflects a numeric recommendation score. It does not establish that the person’s experience was free from frustration, anxiety, disappointment or another negative emotion.
Why is sentiment analysis not enough
Sentiment generally classifies language as positive, negative or neutral. Emotion provides greater specificity. Frustration may suggest removable friction, anxiety may require reassurance, and disappointment may indicate a broken expectation.
How can the same experience contain positive and negative emotion
People regularly experience more than one emotion at the same time. A patient may feel grateful for compassionate care while also feeling anxious about what happens next. One emotion does not invalidate the other.
How should organizations respond to negative emotion inside a positive score
The response should depend on the emotion, context and strength of the relationship. Some experiences call for reassurance or friction removal. Others require acknowledgment, correction or more urgent intervention.
Research and methodology note
Emtherical analyzed more than 350,000 comments across healthcare, financial services and consumer product reviews. A top rating was defined according to the metric used in each dataset: the highest likelihood-to-recommend rating in healthcare, NPS Promoter status in financial services and a five-star rating in consumer reviews. Emtherical’s EM10 framework identifies the nuanced emotions expressed in unstructured feedback, including the possibility that a single comment contains more than one emotion.
About the author
Kirsten Zapiec is CEO and Co-Founder of Emtherical. A senior insights, brand and customer-experience leader, she has spent her career helping organizations understand people more deeply and make better decisions from human experience data.