Skip to content
Open Screeb

Improving your NPS

Go beyond the NPS figure: read its distribution, find why users score low, compare cohorts, and act on detractors, passives and promoters.

Concept4 min read, updated Oct 15, 2026

An NPS is a starting point, not a verdict. The same score can hide very different situations, and it only moves when you act on what users tell you. This page explains how to read your NPS in depth, and how to turn each group of respondents into a lever. For the formula itself, see How Screeb calculates NPS.

Two products with an NPS of +20 can have nothing in common: one has many promoters and many detractors, the other mostly passives. Before you act, look at what is behind the figure.

  • The three groups. Compare the Promoters, Passives and Detractors (see Answer groups). A large passive group is the easiest to move.
  • The distribution. The Result distribution shows where answers pile up. A peak at 10 means a solid base of fans. A cluster at 5 and 6 often means a frustration that users do not voice yet.
  • The trend. The score and the group counts compare with the previous period. A score that drops after a release tells you more than a score on its own, see Track your releases. Where each figure shows is listed in Where the NPS shows.

The score tells you how much, the open answer tells you why.

An overall NPS mixes users who live very different experiences. Filter the report by plan, country or segment to see who pulls the score down.

The NPS report also includes an NPS per cohort table: one row per cohort of users, grouped by their first month of activity, and one column per month. If the users who joined in May score far above those of April, look at what changed in between: onboarding, pricing, a feature. Cohorts are explained in Create user segments.

To know where your score stands in your market, see Compare your scores with your industry.

A detractor who took the time to explain a low score is giving you one of the most useful signals you can get. Treat it as a trigger, not a statistic.

  • Answer fast. Aim for a human reply within one or two days. Past that, the user has often made up their mind.
  • Route it automatically. On All responses, use Share responses with Filtered responses on the scores from 0 to 6: each new detractor is posted to the channel of the team that owns the relationship, see Share responses and reports in Slack.
  • Fix the real issue. Reply with something concrete (a fix, a workaround, a date) rather than an apology. A calendar question lets detractors book a call with you, see Add a calendar question.
  • Close the loop. Tell the user once it is fixed. It is the step most teams skip, and the one that changes their next answer.
  • Track repeat detractors. A user who stays a detractor survey after survey is a churn risk, not one more low score.

Passives are not unhappy: they have not found enough value yet. Filter on their answers, then Create segment from responses to reach them, or target a message with the Gave a score rule. Show them the features they do not use yet, see Choose who sees your message and when.

Promoters (9 or 10): turn them into advocates

Section titled “Promoters (9 or 10): turn them into advocates”

Ask promoters for a testimonial, a review or a case study while their answer is fresh, and invite them to a referral program.

Week Goal Actions
1 Understand Read the distribution, filter by group and cohort, tag the open answers.
2 Detractors Route new detractors to their owners, reply, fix the top blocking issues.
3 Passives Build a segment of passives and target them with a message on unused value.
4 Promoters and team Collect testimonials, share the results and the next actions with the team.

Then measure again on a regular rhythm, see Choosing between NPS, CSAT, CES and NES.

  • The bands of the admin (How the admin judges the score) are a general reference. Your own trend, and your industry, matter more than a fixed target.
  • The NPS measures the relationship with your product as a whole. To measure one interaction or one task, use a CSAT, CES or NES question instead.