Every NPS survey asks two things: a number from zero to ten, and an open text box underneath it that says “why?” The number gets extracted into a dashboard, trended quarter over quarter, and presented in a board deck. The sentence sitting right below it, the one place a customer explains their actual reasoning in their own words, gets read once by whoever ran the survey and then filed away.
That is the wrong half to automate past.
A score is a customer rounded down to one digit
A seven out of ten could mean “this mostly works but your support team never replies” or “I like the product but I’m evaluating a competitor next quarter.” Both round to the same number and trigger the same dashboard color. The distance between those two customers is enormous, and it is the first thing lost the moment the score replaces the sentence as the unit of analysis.
The comment box is the only unprompted research a company gets for free
Most customer research has to be scheduled, incentivized, or extracted through a call where the customer is managing the relationship as much as answering honestly. The NPS comment box is different - the customer has already decided to respond, is already in a reflective mode, and is writing without anyone in the room to perform for. It is closer to a focus group than a survey metric, and almost no company treats it that way.
Why it gets skipped anyway
Reading fifty scores takes five minutes. Reading fifty comments takes an hour, and most of them say something unglamorous and specific - a workflow that broke, a feature a competitor shipped first, a support reply that took too long. None of that compresses into a slide the way a trend line does, so it gets summarized into “sentiment is mixed” and the specifics die in a spreadsheet nobody reopens.
What changes if someone actually reads them
The comments that matter most aren’t the five-star raves or the one-star complaints - both are already visible elsewhere. It is the sixes and sevens, the customers who are still there but have quietly started hedging, who hand a company its best early warning for free. A founder who reads a hundred of those comments a quarter will know more about where the product is actually losing trust than any dashboard built on the number above them.