How to Improve NPS With Feedback Data: A 90-Day Plan

• NPS goes up when you fix what drives it, not when you chase the score itself.
• In 90 days you can segment the score, read every comment, fix two or three drivers and close the loop with every detractor.
• Set targets against your own trend. Report the change in NPS points, together with the actions behind it.

NPS is easy to calculate and hard to move. Many teams report it every month, watch it drift by a point or two and can't say why.

The score isn't the problem. A single number for the whole company hides the places, journeys and moments that are pulling it down. This plan breaks the work into three phases you can finish in a quarter. If you need a refresher on the measure itself, start with our NPS guide.

• Know which NPS you are measuring. Relationship NPS tells you how customers feel about you overall. Touchpoint NPS tells you how one moment went. Both are useful, but don't mix them in one trend line. Our relational vs transactional NPS guide explains the difference.
• Keep the question identical across channels and over time.
• Always ask why. Add one open question: "What is the main reason for your score?"
• Check your volumes. Before comparing segments, make sure each has enough responses to be stable.

Split NPS by location, channel, journey and customer group. The company average usually hides a wide spread.

For example, a retailer with an overall NPS of +28 might find one region at +41 and another at +9. Within that second region, the gap may sit almost entirely in click-and-collect orders. (Illustrative figures.)

Then read what detractors and passives wrote, all of it, not a sample. An AI engine can code every comment into topics and sentiment in minutes. In two weeks you should have a clear list of the issues behind the low scores in each segment.

Rank the issues by their impact on NPS, not by how often they are mentioned. A topic that appears in a fifth of comments but barely moves the score matters less than a rarer one that reliably turns customers into detractors.

Pick two or three issues, no more. For each one, agree on:

• an owner;
• a deadline;
• what "fixed" will look like in the data.

Small fixes count. Telling customers clearly when an order will arrive is often worth more than a large project that ships next year.

While the fixes are under way, make sure no unhappy customer is left without a reply:

1. 1A low score triggers an alert and opens a case for the right owner.
2. 2The owner contacts the customer within a target you set. Forty-eight hours is a good starting point.
3. 3They listen, resolve what they can and record what they learned.
4. 4The case is closed only when the customer has heard back.

Track how many loops are closed on time; that number tells you as much as NPS does. Passives deserve attention too, because they are the customers most easily won over by a competitor. Promoters can be invited to share their experience in a public review.

Compare each segment you worked on with the period before the fixes. Look at three things:

• did the share of comments about each issue fall?
• did NPS in that segment rise?
• did the change hold for more than a few weeks?

Report the change in NPS points, never as a percentage. Show the actions behind it, so leadership can see the score moved because something was fixed.

Then start the next cycle. The issues ranked fourth and fifth last quarter may now be at the top.

• Use your own trend as the benchmark. Industry averages differ by country, channel and survey method.
• Set targets by segment. A region at +9 has more room to improve than one at +41.
• Never ask customers for a high score. "Please give us a 10" makes the number look better and the data worse.
• Be careful with bonuses tied only to NPS. They reward the score rather than the fixes. Pair the score with loop-closure rates and completed actions.

Three steps of this plan take the most time when done by hand. AI shortens each of them:

• Reading every comment: each one is coded into topics and sentiment.
• Ranking issues: topics are ranked by their impact on NPS, and you can estimate the gain from fixing each one.
• Spotting risk early: by connecting feedback with transaction data, you can see which passives and promoters are at rising risk of leaving before their next score arrives.

Behind the score are customers telling you, in their own words, what would make them stay. Ninety days is enough to show that you heard them.

Our research, analytics and CX team writes about what we learn running feedback programmes across 15+ industries.