Know-How Exchange : Research/Metrics

I don't know, but I've posted your question to Fred Reicheld through www.theultimatequestion.com/theultimatequestion/home.asp and invited him to reply, so here's hoping!

Mark,
If you are asking how large a sample to take for your NPS to be valid, then you have a couple parameters to worry about. First, you look at the acceptable error rate. In most situations, a 5% error rate is acceptable...so you'd say that the NPS is 65% +/- 3% (5% of 65%). In most business situations, this is adequate, since the rest of the business information is much less accurate. The second parameter is the degree of confidence. Normally, we use 95% confidence. We say, "with 95%, the NPS is between 62% and 68%. The third parameter is the population size. The largest sample required for 95% confidence of error no greater than 5% is 385 for populations at or above 468,000. An assumption here is that they population is homogeneous and the sample is drawn randomly. The last parameter is the probability (of a positive answer). Worst case is 50% (the toss of a coin. For probabilities lower or higher, the required sample size for a given confidence level and error is lower. For a population of 468,000, confidence of 95%, error of 5%, and probability of 50%, for instance, the sample size required is 385. For 65% probability (or 65% NPS), the sample size required is 350. For 35% NPS, the sample size is also 350. Since until we sample, generally we don't know the probability, we look at the worst case sample and use 50%. That way, if the answer is less or greater than 50%, our error will be less than 5% because we have more samples than needed.

To play with the numbers a little, see:

https://www.berrie.dds.nl/calcss.htm

I hope this helps.

Wayde

ChrisB - thanks for bringing this post to our attention.

Randles.mark - I'd like to point you to the "Who Is Using Net Promoter?" discussion forum and blogs for some related responses. The answer really does vary depending on your organization. I am assuming you are collecting NPS data for your own company.

One person posted a question related to industry standards, at the link below, which gets at some of what you're asking. I would like to call out the 1st response by Dr. Vince Nowinski, one of the methodologists at Satmetrix, the co-developer of Net Promoter.

https://netpromoter.groupee.net/eve/forums/a/tpc/f/2731073251/m/1801039112

I'd also like to point you to "The Rules of Measurement" as outlined in Chapter 6 of Fred Reichheld’s book, The Ultimate Question: Driving Good Profits and True Growth, in this forum link:

https://netpromoter.groupee.net/eve/forums/a/tpc/f/2731073251/m/3571009702

At the link above, there are a number of links to related discussion forum posts and Net Promoter blogs.

If you have further questions, indeed, please feel free to post your question to the Net Promoter discussion forum, and I will work with the develpers of Net Promoter for a response. (Note: I don't mean to take anyone away from this great forum, which I personally love!)

Regards,
Amy
Project Manager for Netpromoter.com

Hi Mark

In one interview I've read, Mr. Reicheld suggests that we throw away traditional notions of representative sampling (all very well explained above) and consider a customer census. His thesis is that the NPS question is the only one that you need to know to grow your business. By eliminating all the other questions you might ask your customers you can achieve a far greater number of interviews / responses than a traditional survey. This is a powerful, yet counter intuitive idea.

Imagine for a moment that you could have a view of all your customers and know if they are Promoters, Passives or Detractors. This goes beyond just knowing your statistically valid NPS score and goes to customer segmentation. By cross tabulating the NPS with customer data (say revenue) you could generate a very interesting matrix for your marketing and sales organization. For example, you would be able to point to those customers who generate the most revenue but have the lowest likelihood to do repeat business with your company (high revenue, low NPS).

Alternatively, you could also identify your best reference customers (high revenue, high NPS). These customers you might funnel into a reference program or put a higher priority to ensure that they continue to be highly satisfied with your organization.

So to answer your question... all your customers.