04 April 2017 | by Tim Bock

The NPS Recoding Trick: The Smart Way to Compute the Net Promoter Score

 

npscomputation1

The Net Promoter Score is most people’s go-to measure for evaluating companies, brands, and business units. However, the the standard way of computing the NPS – subtract the promoters from the detractors – is a bit of a pain. And, in most apps, you cannot use it in stat tests, so you are never really sure when the NPS moves whether it reflects a change in performance, or just some random noise in the data.

 


 

The standard way of computing the NPS

likelihoodtorecommend

The table to the right shows data for Apple.  Fourteen (14.4%) said they were Not at all likely to recommend Apple, 2.6% gave a rating of 1 out of 10, 2.0% gave a rating of 2, etc. If we add up the first seven categories (0 through 6), 51% of people in this data set are Detractors. Adding up categories 7 and 8 gives us 31% Neutrals, and then the remaining 18% are Promoters. So, in this data set, Apple’s NPS is -33.3, which is not great. (Among Apple customers, the NPS is much higher.)

 

 


 

A smarter way

The table below shows the raw data for the 153 people in the data set. The actual ratings, out of 11, are shown in the Recommend: Apple column. The second column shows the recoded data, where the original values are replaced by new values. The trick is to replace values of 0 through 6 with -100, values of 7 or 8 become 0, and values of 9 or 10 get a new value of 100. The column on the right shows the recoded data.

Why is this recoding clever? Once you have recoded the data this way, you can compute the NPS by computing the average, and you get exactly the same answer as you do when using the standard way.


 

Why the smarter way is so much smarter

The genius of the smarter way is that we can now compute NPS using any analysis app that is able to compute an average. For example, I have used the multiway table feature in Displayr to compute the NPS for Apple by age and gender, by just selecting the three variable (age, gender, and the recoded NPS variable).

The multiway table is created using Insert > More > Tables > Multiway Table.

 


 

Doing this with your own data using Displayr

The fastest way to do do this is to start using Displayr, and then:

  1. Import a data set: Home > Data Set (Data). If you want to play around with a live example where the data is already in Displayr, click here.
  2. Drag the variable containing the likelihood to recommend data from the Data tree onto the page, so that it creates a table (like the first table in this post)
  3. Select the table, and select Data Manipulation > Utilities (far right) > Compute > Net Promoter Score. This will add the NPS to the bottom of the table.
  4. (Optional) Select the variable (in the Data tree, bottom-left), and change the Structure to Numeric (in the Object Inspector on the right). This will mean that you only ever see the NPS, rather than seeing both the NPS and the percentages in each category.

Alternatively, if you want to do the calculations “by hand”:

  1. Import a data set: Home > Data Set (Data). If you want to play around with a live example where the data is already in Displayr, click here.
  2. Drag the variable containing the likelihood to recommend data from the Data tree onto the page, so that it creates a table (like the first table in this post).
  3. Select the variable in the Data tree.NPSRecodedValues
  4. Select Data Manipulation > Values (Data Values). 
  5. Change the entries in the Value column so that they look like those to the right, and press OK.
  6. Change the Structure to Numeric (in the Object Inspector on the right). This will cause the table to show the Average, which gives us the NPS.

 

Author: Tim Bock

Tim Bock is the founder of Displayr. Tim is a data scientist, who has consulted, published academic papers, and won awards, for problems/techniques as diverse as neural networks, mixture models, data fusion, market segmentation, IPO pricing, small sample research, and data visualization. He has conducted data science projects for numerous companies, including Pfizer, Coca Cola, ACNielsen, KFC, Weight Watchers, Unilever, and Nestle. He is also the founder of Q www.qresearchsoftware.com, a data science product designed for survey research, which is used by all the world’s seven largest market research consultancies. He studied econometrics, maths, and marketing, and has a University Medal and PhD from the University of New South Wales (Australia’s leading research university), where he was an adjunct member of staff for 15 years.

No comment. Share your thoughts.

Leave a Reply

Your email address will not be published. Required fields are marked *

Human? *

Keep updated with the latest in data science.