leverboard
Try it on your data

How to improve NPS

NPS goes up when fewer customers give 0 to 6 and more give 9 or 10. That happens when you fix the reasons behind the low scores, not when you ask differently. Find which reason costs the most points, fix it, close the loop with the customers who raised it, and measure with enough responses to see the change.

Four steps

Each step is explained below.

  1. Find what costs points

    Rank the reasons in your comments by how many points each costs the score, not by how often it comes up.

  2. Fix the top reason

    One reason at a time, with an owner and a date. Read its comments to see what exactly goes wrong.

  3. Close the loop

    Reply to the customers who raised it. Tell them what you changed.

  4. Measure the change

    Same question, channel and timing, with enough responses to tell a real change from noise.

1. Find what costs points

Most teams count the complaints in their comments and start with the most frequent one. Frequent is not the same as expensive. What a reason costs the score is how much it lowers the score when it comes up, times how often it comes up.

In 4,000 public clothing reviews, fit or size was mentioned in 32 % of reviews. “Doesn’t look flattering” came up in only 15 %, but each mention went with a much lower rating, so it cost the average rating more: −0.14 stars against −0.09.

Complaints customers mention most

  1. 1Fit or size32 %
  2. 2Doesn’t look flattering15 %
  3. 3Fabric10 %

What actually costs points

  1. 1Doesn’t look flatteringup 1−0.14
  2. 2Fit or sizedown 1−0.09
  3. 3Fabric−0.06

Women’s E-Commerce Clothing Reviews, public domain (CC0), 1 to 5 stars. Share of reviews on the left; on the right, stars each theme costs the average rating.

Read the numbers before you act

Three checks keep you from fixing the wrong thing.

Separate the reasonsUnhappy customers often raise several problems in one comment. If you compare the score of comments that mention a theme with the rest, each theme takes the blame for the others. In our restaurant test, raw differences like that came out about twice the separate effect of each complaint.
Check how much data a reason rests onA reason mentioned in 12 comments can look dramatic and be noise. Leverboard marks reasons with fewer than 30 comments as low data: a lead to read, not a number to report.
An effect is not proof of causeA reason that goes with lower scores is where to look first. The next survey shows whether fixing it moved the score.

2 and 3. Fix one reason, then close the loop

Pick the reason at the top of the cost ranking and open its comments. They usually narrow a broad theme like “delivery” down to one step that fails: late dispatch, a carrier, missing tracking.

Bain recommends following up with customers whose feedback merits it, so that employees learn and improve, and calls it closing the loop. For detractors that means a personal reply that fixes their case. For the reason as a whole, it means telling the customers who raised it what changed.

Give it an ownerOne person, one date, one change. A list of ten priorities is no priority.
Fix the case and the causeA refund settles one complaint. A changed process stops the next hundred.
Reply to detractors quicklyWithin days, from someone who can act. Ask what happened and what would put it right.
Ask passives what is missing“What would have made it a 9 or 10?” gives you the next item on the list. See promoters, passives and detractors.

4. Measure the change

An NPS from 100 responses is likely within about 15 points of the true value; from 400 responses, about 8. A gain of 5 points with 100 responses a month is not evidence that the fix worked. Wait for enough responses, or compare quarters instead of months. What is a good NPS score? explains the ranges.

Keep the survey fixedSame question, same channel, same moment in the customer journey. Change any of them and the score moves on its own.
Do not over-askFor surveys sent after an interaction, Bain recommends asking the same customer at most once every three months.
Watch the reason, not only the scoreIf the fixed reason comes up less often, or costs fewer points, the fix worked even before the headline score settles.

What does not improve NPS

Some moves raise the number without changing what customers think. They make the score useless for the next decision.

Asking for a 9 or 10“Please rate us 10” raises the score and hides the reasons.
Surveying only happy momentsSending the survey after a successful delivery but not after a complaint measures your best days.
Changing the scale or wordingA 1 to 10 scale or a softer question breaks the comparison with every earlier score. Bain’s scale is 0 to 10.
Acting on the most frequent complaint by defaultIt may be the cheapest one. Check what it costs first.

Questions

What is the fastest way to improve NPS?

Fix the reason that costs the most points, not the one mentioned most often. Then reply to the detractors who raised it. Anything that raises the score without changing the experience, such as asking for a 10, makes the score useless.

How long does it take to improve NPS?

As long as the fix takes, plus enough new responses to see it. With 100 responses a month, each monthly score is uncertain by about 15 points either way, and a change between two months even more, so a real gain often shows only when you compare quarters.

Is 37 a good NPS score?

It is clearly positive: the share of promoters is 37 points higher than the share of detractors. Whether it is good for you depends on your earlier scores, asked the same way. See what is a good NPS score.

Should NPS be 0 to 10 or 1 to 10?

0 to 10. Bain, which created NPS, uses an 11-point scale from 0 to 10 because a zero is never read as a good score, while on a 1 to 10 scale people mix up the ends.

Do I need software to find what costs points?

You can read and tag comments by hand in a spreadsheet, which works for a few hundred. Telling frequent from expensive reasons needs each reason’s effect measured apart from the others, with a likely range. That is what Leverboard does.

Find the reason that costs you the most

Your score says how many customers are unhappy. Your comments say why. Leverboard measures which reason costs the most, how sure that is, and shows the comments behind every number. What the numbers mean.

Try it on your dataSee a sample report

First analysis free. No card.

Net Promoter, NPS and Net Promoter Score are trademarks of Bain & Company, Inc., NICE Systems, Inc. and Fred Reichheld. Sources: Bain, Measuring your Net Promoter Score and Three types of Net Promoter Scores.