Does NPS Actually Predict Business Outcomes?

No metric since the inception of the Likert scale has become as ubiquitous as NPS. Its unchallenged stature as the North Star of customer affinity means that NPS needs no introduction or explanation. In a nod to that preeminence, many companies follow it as an act of faith because it provides a standardized benchmark that’s easy to track and compare. Whether the NPS report card also supplies meaningful context for brand performance assessment has been the subject of some debate – but it’s fair to say that any customer affinity metric, no matter how powerful, is just the starting point for an ambitious plan of action. Fred Reichheld, founding father of this “Bain-child,” has, in fact, devoted several books to the topic, arguing that NPS should be considered a way of thinking about what is owed to customers and an important step toward fulfilling that obligation. It should not be a vanity contest for brand leadership.

But for a while, even Reichheld puzzled over why so few people had made a serious empirical attempt to prove its worth, and although there are now more case studies and modeling efforts to justify use, the issue has not been put officially to bed. One reason may be that some of the predicted outcome measures for the purpose – such as corporate growth or profitability – are themselves the product of so many drivers that even important metrics of brand affinity like NPS can account for only a limited amount of the variance. It stands to reason that NPS would predict some things. The question is what and how well?

To that point, why have people swung from the fences in attempts to relate NPS to something as “big” and complex as overall corporate growth or profitability? Perhaps a metric that has earned so much notoriety implicitly invites that level of proof. But, in practical terms, there is another more obvious factor at play. Many companies simply lack the right sort of data to link NPS with concrete, measurable customer behaviors and value metrics. It is far easier to demonstrate the relationship in B2B or financial services sectors, where customer data are rich, than in consumer packaged goods – especially if the goal is to confirm that NPS is truly a leading indicator of something and not merely a loose correlate. If people who claim to buy a lot of Cheerios give NPS scores consistent with their self-reported behavior, we have learned only about halo effects, and we have not moved much beyond applause meters. If, by contrast, the correlation is weak, we might question whether the NPS is optimally framed for the topic of cereal, a category in which the notion of literal “recommendation” may not express the relationship between consumers and their brand.