Net Promoter Score

Companies measured satisfaction long before this metric existed. The trouble started after the measuring. Into the early 2000s the standard instrument was a thirty-question survey: was the support agent polite, was the screen easy to read, was the price fair, each rated one to five and averaged together. The result came back as 4.1. Next quarter it came back as 4.2.

Numbers that don't move produce meetings that don't move. Someone would report that satisfaction was up a tenth of a point, someone else would ask what to do about it, nobody could answer, and the meeting went to the next item. With thirty questions mixed into one average, there was no way to tell which part of the experience had anything to do with renewals.

Net Promoter Score (NPS) sidesteps that by cutting the survey to one question. Instead of asking whether customers are satisfied, it asks whether they would recommend the product to someone else. A recommendation puts the customer's own reputation on the line, which makes it a stronger signal than mild approval. And rather than averaging, it subtracts the share of very low answers from the share of very high ones — so a pile of lukewarm responses can't flatten the result toward the middle.

Ask for likelihood to recommend on an eleven-point scale from 0 to 10, sort the answers into three groups, and subtract the bottom group's share from the top group's share.

The groups split like this.

1. Promoters — people who answered 9 or 10. They bring the product up unprompted, and renewals and referrals mostly come from here.
2. Passives — people who answered 7 or 8. Nothing is wrong, but a better option will move them. They drop out of the arithmetic entirely.
3. Detractors — people who answered anywhere from 0 to 6. They may say negative things about the product to other people.

The math is one subtraction. Out of 200 responses, 90 answered 9 or 10, 60 answered 7 or 8, and 50 answered 0 to 6: that's 45 percent minus 25 percent, or 20. The scale runs from minus 100 to 100, and the result is read as a score rather than a percentage. Not 20 percent — just 20.

Counting a 6 as a detractor looks harsh, and it stays that way on purpose. Someone who gives a 6 out of 10 usually gives a lower number the next time you ask. Moving the cutoff to 7 or 8 raises the score, but the number stops being comparable to any other period or any other study.

Three facts travel with the number: when you asked, who you asked, and how many answered. A score collected on signup day is a different number from one collected after six months of use. A score from paying customers only is a different number from one that includes free users.

The benchmark comes from inside the company, not from outside it. The most common misread is seeing an industry average of 31, comparing it to your own 24, and concluding that you're behind — without knowing when that 31 was collected, from whom, or with what wording.

Three comparisons are worth making: your own number from last quarter using identical wording, the gap between customer segments, and the link between the score and actual behavior. The third one earns its keep. At Pagedeck, teams that answered 0 to 6 renewed the following year at 41 percent, while teams that answered 9 or 10 renewed at 88 percent. Once that link was established, the score stopped being a number that needed interpreting and became an early warning about renewals.

Rules of thumb do circulate. Subscription software is generally considered healthy above 30. Use that to get a rough sense of where you sit, not as a passing grade.

Response rate deserves a look of its own. The people who answer surveys are usually the delighted and the furious; everyone in the middle skips it. If you sent the survey to 30,000 users and 400 answered, compare the makeup of those 400 against the makeup of all users first. If paying customers are 12 percent of the base but 40 percent of the respondents, the score describes paying customers, not the user base.

Put a free-text box at the end of the question. A score of 24 doesn't tell you what to fix tomorrow, but reading thirty sentences from the people who scored low will show the same word repeating. Collecting that feedback and turning it into product decisions is the practice known as voice of the customer (VOC).

"We're at 24 this quarter. That's up 6 points, but half as many people responded."

Not a report that the score went up — a report that it can't be called an increase. When the respondents change, so does the number.

"Enterprise accounts on their own come to 41. Add the free plan and it's 9."

One average is hiding two groups that have nothing in common.

"Send the list of accounts that scored 0 to 6 over to sales."

Treating the score as a renewal-risk list rather than a reporting figure.

Pagedeck surveys the admins of its paying teams every quarter. When results land, Mia writes up the score alongside the response rate, and the accounts that scored 0 to 6 go to Sofia. Sofia starts with the teams whose renewal falls within three months. Last quarter, eleven of the eighteen accounts she called said the same thing: permission settings were too complicated.

The score itself rarely moves much in a single quarter. What the meeting actually spends time on isn't the change in the number but the sentence that low scorers kept writing.

The most common mistake is turning it into a target. It looks like any other metric, so "let's hit 30 this quarter" comes out naturally. But this number swings by 10 points or more just from changing who gets asked and when. Once a target is attached, the team starts fixing the survey instead of the product — sending it only to customers who just completed a purchase, or excluding accounts that recently filed a complaint. Six months later the score reads 30 and the renewal rate hasn't budged.

Reading intent as action is the other one. A wave of 9s and 10s feels like word of mouth is about to follow. The question measures willingness, and willingness needs one more thing before it becomes behavior: an occasion. If the score is high but new signups show no referral path, the problem isn't the product — there's no convenient way to pass it along, and the thing to fix is the invite screen.