Posted by Richard Evensen on April 18, 2011
(Note: Due to concerns that the strong language of my original blog would overshadow the message, I have rewritten the first paragraph. The goal of this blog is to stimulate conversation – admittedly, through confrontation – in order to determine what correlation NPS really has and how to best use it, or not!)
As the Customer Satisfaction (CSAT) analyst in the Market Insights Team, I get a lot of questions about Net Promoter Score (NPS). As the “ultimate question,” NPS is being positioned as a cure to all business ills and a way to understand everything you need to know about your customers . . . with just one question. Most market insights professionals struggle with NPS as it goes against our training to accept data correlations which have not been proven. Presenting actionable insights tied to NPS when correlation has not been proven? It feels like selling snake oil and is likely one of the drivers for the view that “researchers hate this metric.”
Has NPS — like Britney Spears’ singing — been overhyped?
My professional colleagues — including Timothy L. Keiningham, Bruce Cooil, Tor Wallin Andreassen and Lerzan Aksoy (A Longitudinal Examination of Net Promoter and Firm Revenue Growth) and Jeffrey Henning (Net Promoter Score [NPS] Criticisms and Best Practices) — have already beautifully, and very analytically, dissected and disproven NPS. They are not alone in their views or criticisms. So, why do we still use NPS when there is no clear statistical conclusions as to its effectiveness?!
The reality is that, for some, NPS may provide some correlation with customer action and be a potential indicator of customer spend (TBD in my book as I have yet to see sufficient evidence where this is the case). For many, it’s an easy way for executives to evaluate their business’ health and motivate, if not compensate, for improvements. Granted, if there is weak correlation between NPS and customer action and you’re compensating for improvements in NPS, well, does this make strategic sense?
For most, I would have to say: Stop using NPS. Or, better said, start using it more properly.
The central issue for me with NPS is the formulation: Score = Promoters (% of customers scoring 9-10 on a 10 point scale) minus Detractors (% scoring 6 or less on a 10 point scale). Why is this an issue?
- It’s culturally insensitive. Case in point is a recent rating given to me by a European client. To preface, he told me “10 is for God. 9 is for el Maestro. So, 8 is as good as you can get.” He gave me between and 8 and a 9 and I was ecstatic! In NPS terms though, well, it just wasn’t good enough. There are other examples as well and research which demonstrates cultural skewing which needs to be factored into assessments (see Kristin Cavallaro’s article “Data Use: Are global scales as easy as 1-2-3 or A-B-C,” January 2011 — via Quirks or SSI).
- It’s not a good motivator. I’ll be honest that I have not quantified the psychological impact of NPS but do question how motivated employees are to improve a 20% score to a 30% score (when best-in-class is sometimes only 40% to 60%!). Seriously, is this something you can commit to and then will brag about with your friends? As a motivator, NPS psychologically causes us to reset our bar to a much lower level. Should we strive for only 50% of our customers to be satisfied with their support interactions?
- It has questionable correlation. The biggest concern with the formula is that there is not sufficient quantitative evidence to support a correlation between NPS and customer actions (renewals or even actual recommendations). Given its time in the field, SatMetrix should be able to show 100s if not 1000s of cases with high correlation coefficients (Pearson’s r=70%+). To date, they have only provided a totally unsupported position that “findings support the link between Net Promoter and financials.” Please provide us the proofs as I have several clients who disagree with your findings!
So, how can we potentially use NPS?
Once you get past the problematic formula and the dubious view that this “ultimate question” is the answer to everything, you end up with what is really a very good customer satisfaction question! Just realize that you are testing for perceived value, not intention and definitely not action! As such, treat it like any other customer satisfaction question and make sure you capture other input around your customers’ experience to ensure sufficiently robust Voice of the Customer (VoC) input. For some other good tips and input from your colleagues, check out the lengthy NPS discussion on the Market Insights Community page: http://community.forrester.com/thread/4449?tstart=0.
In upcoming Forrester research, we’ll dig a bit more into NPS (and look at some different ways to "fix the formula"), along with providing some input on how to make to make CSAT more actionable. As always, we would love to hear your views:
- Do you use NPS as-is or have you made any modifications to make it work better for you?
- What statistical correlation (please provide coefficient # and scale) do you find between NPS and your operational data?
- How much is NPS being used at your company to motivate and compensate employees . . . and is this effective?
- What do you think of the views expressed in this blog?
Please feel free to join in the discussion here or on the Market Insights Community page.