How to read NPS comments next to the score

The score itself is simple. People answer "How likely are you to recommend us to a friend or colleague?" on a 0 to 10 scale. Those who pick 9 or 10 are promoters, 7 or 8 are passives, and 0 to 6 are detractors. The score is the percentage of promoters minus the percentage of detractors (Bain).

The score was never meant to stand alone. In Bain's own introduction to the system, the 0 to 10 question is followed by "What is the primary reason for your score?" and "What could we do better?" The same paper says "the other questions are even more important than the score itself" (Bain, Markey and Reichheld).

A number between -100 and 100 can't tell you what to fix. The written reasons can.

"What is the main reason for your score?" works for anyone, so it's the safe default. If your tool can branch on the score, a question shaped to each group gets more specific answers.

QuestionPro's guide suggests questions like these for each group (QuestionPro):

• Detractors (0 to 6): "What was missing or disappointing about your experience?"
• Passives (7 or 8): "What would make you more likely to recommend us?"
• Promoters (9 or 10): "What is the main reason you would recommend us to others?"

The same guide warns against the mismatch: asking a detractor what they loved most wastes the moment, and asking a promoter what went wrong feels tone deaf.

Keep it to one open box per person. On Pew Research Center's panel, open questions go unanswered about 18% of the time on average, against 1% to 2% for closed questions. Questions that ask for several sentences get skipped more than ones that ask for a word or a phrase (Pew). Ask for one reason, not an essay.

Don't pool every comment. A detractor's "support took four days to reply" and a promoter's "support is quick" are about the same thing, but they mean different work. Read and label each group separately, then compare.

• Detractors tell you what broke. Look for one concrete event per reply.
• Passives tell you what's missing. Nothing went wrong, it just wasn't enough.
• Promoters tell you what to protect. If you change the thing they praise, the score can drop for reasons nobody complained about.

Sometimes a 7 comes with a glowing comment, or a 9 comes with a complaint. Colin Shaw describes a company whose people left positive comments but gave middling scores. His read: "There is a significant difference between what customers say and what they mean," and politeness is a strong motivator (CustomerThink).

Don't move people between groups based on their words. Keep the score they gave and label the comment as written. A pile of polite 7s with no complaint is its own finding: nothing is wrong, and nothing stands out enough to talk about.

Reading is not enough once you have more than a few dozen comments. Give each comment a short label that says what happened, count the labels per group, and reread the comments under the top counts before you decide anything. Our guide on how to analyze open-ended survey responses walks through the labelling step by step.

Jeff Sauro at MeasuringU did this with a small NPS study of iTunes (MeasuringU). Of 56 people, 31 were promoters and 12 were detractors, an NPS of 34%. He pulled 16 points out of the detractors' comments and grouped them. "Difficult to use" and "proprietary" each had 4. Two cautions from the same article apply to any small batch. With so few comments, the likely share of all detractors who'd say "difficult to use" ran anywhere from about 10% to 50%. And "there can be a high amount of variability between evaluators grouping items," so he suggests two people label the comments separately and compare.

Then track the counts next to the score, every time you run the survey.

The numbers below are invented to show the method. They are not from a real survey.

Say a software company runs the survey every quarter. In the first quarter, 200 people answer: 90 promoters, 64 passives and 46 detractors, an NPS of 22. In the second, 210 answer: 97 promoters, 67 passives and 46 detractors, an NPS of 24. On its own, that looks like a quiet quarter. Here are the detractors' reasons, counted the same way both times:

The score barely moved, but the reasons swapped places. The support fix worked: 17 down to 6. The price change added 12 detractors, about as many as the support fix took away. A flat score hid both. (Scores worked out as percent promoters minus percent detractors, rounded.)

The counting step is the part Enform does for you. The tour below shows it reading a batch of replies.

• Reread before you act. Open the 18 "price went up" comments. Some may be about the increase itself, some about paying for a feature they never use. Those are two different fixes.
• Pass comments on, and reply. Bain calls this closing the loop: share each comment, as soon as possible, with the people who shaped that experience, and contact the people whose feedback needs a follow-up. A short note saying what you changed tells people their answer was read.
• Keep the labels steady. Reuse last quarter's list and add new labels only when a new reason shows up. Renaming labels every time makes the counts impossible to compare.
• Watch the "no reason" row. If it grows, the follow-up may be too broad.

Three ways to do the counting:

• By hand in a spreadsheet. Export the replies to Google Sheets or Excel, filter by score group, add a label column and count with COUNTIF. Free, and slow past a few hundred comments.
• A general AI chat. ChatGPT or Claude can sort a pasted column of comments quickly. Check its counts against the replies before you trust them.
• A survey tool with an NPS question. SurveyMonkey's NPS template works out the score and sorts people into promoters, passives and detractors for you (SurveyMonkey). See Enform vs SurveyMonkey.

Enform writes the survey from a description, you share one link, and it reads every reply and counts what people said. Every count opens to the replies behind it, so you can reread the 18 before deciding what to fix. It fits when the written reasons matter more to you than dashboards. Start from the NPS survey template, compare NPS survey tools, or see pricing.