Every quarter, thousands of companies send a single question to their customers: “How likely are you to recommend us?” The answers come back as a number between 0 and 10. That number gets subtracted, divided, and turned into a score that lands on an executive dashboard. Then nothing happens.
The score goes up. The team celebrates. The score goes down. The team panics. Neither reaction produces a specific action, because the score itself carries no information about what to do next. It is a thermometer with no diagnosis.
The NPS funnel · where signal disappears
The score goes to the dashboard. The “why” goes nowhere.
The origin story everybody skipped
Fred Reichheld published “The One Number You Need to Grow” in the Harvard Business Review in December 2003. The article made a bold claim: the Net Promoter Score was the single best predictor of organic growth. One number. One question. Simple, memorable, and easy to benchmark.
The business world adopted it almost overnight. By 2010, two-thirds of Fortune 1000 companies were running NPS programs. But the research that followed told a more complicated story.
In 2007, Timothy Keiningham and colleagues published a longitudinal study in the Journal of Marketing that tested Reichheld's core claim directly. Their finding: NPS was no better than the American Customer Satisfaction Index at predicting revenue growth. It wasn't worse, either. It was just one metric among many — not the breakthrough its creator advertised.
A year earlier, Neil Morgan and Luciano Rego examined multiple satisfaction and loyalty metrics in Marketing Science. Their conclusion was sharper: average satisfaction scores and top-two-box measures outperformed NPS for predicting business performance across most categories. The “one number” wasn't even the best number.
The dashboard trap
None of this stopped NPS from spreading. And the reason it spread reveals the real problem.
NPS is easy to collect, easy to display, and easy to compare. It fits on a slide. It fits in a quarterly review. It fits in a conversation between a CEO and a board member who has four minutes for customer sentiment.
So the entire NPS industry — the survey tools, the benchmarking services, the consultancies — optimized for the score. They built dashboards to track it over time. They built segmentation to slice it by cohort. They built alerts for when it drops below a threshold. All of this treats the number as the output.
The number is not the output. The number is a signal that something changed. The “why” question — the open-ended follow-up that most NPS tools tack on as an afterthought — is where you learn what actually changed. Did the onboarding break? Did a competitor launch something better? Did your last release introduce a bug that three people have filed tickets about?
The “why” answer tells you. The score does not.
The “why” field nobody designed for
Here is the uncomfortable part. Most NPS programs include a “why” follow-up. But nobody designed for it. The field sits below the score selector, usually with a label like “Tell us why you gave this score” and a blank textarea. Most respondents skip it entirely. The ones who don't write two words: “It's fine.” “Like it.” “Too slow.”
This is not a data problem. It is a design problem. The “why” field is a recall task — you are asking someone to generate an explanation from memory for a feeling they may not fully understand themselves. That is cognitively expensive. So people satisfice. They write the minimum and close the tab.
The tools that collect NPS don't help. They invest in the score dashboard, the trend chart, the segment comparison. The “why” responses get dumped into a table or piped into a word cloud. Nobody reads them because there is nothing to read — a thousand variations of “good product” and “needs work” are not actionable.
The “why” field · before and after probes
What changes when the form asks back
The fix is not to abandon NPS. The score has value as a trend line — Keiningham's research showed it correlates with growth, just not uniquely. The fix is to take the “why” field seriously. Treat it as the primary output, not the appendix.
Follow-up probes turn a passive textarea into a conversation. When someone writes “it's slow,” the form asks back: “Which part feels slow — loading, navigation, or something else?” When someone writes “love it,” the form asks: “What specifically would you miss if it went away?”
Two things happen. First, the response gets specific. “It's slow” becomes “the reporting page takes 12 seconds to load when I filter by date range.” That is a ticket. You can assign it on Monday morning. Second, more people answer, because a direct question is easier to respond to than a blank textarea. The probe does the cognitive work of narrowing the scope.
AI pre-fill goes further. When the respondent hands the form to their AI assistant, the draft draws on real context — support tickets they filed, features they use, conversations they had. The “why” answer arrives with specifics the respondent would have forgotten. They review the draft, correct what is wrong, and submit.
The score still ends up on the dashboard. But now the “why” column is worth reading.
The metric is not the measurement
Reichheld was right about one thing: the follow-up question matters more than the score. In his original article, he argued that the real value of NPS was the conversation it started — the qualitative feedback loop between company and customer. The industry heard “one number” and built dashboards. It missed “conversation” entirely.
If you are running an NPS program and your “why” responses average fewer than ten words, the problem is not your customers. The problem is that nobody designed the “why” field to produce real answers. A blank textarea and a generic prompt will always produce generic responses.
Your score is a number. The “why” is where the work starts.