Another article in a short series on de-coding commercial growth and performance metrics, this one's on the poster child of the 'Loyalty Advocates', Net Promoter Score or NPS. It's a metric developed by Frederick F. Reichheld, a Bain & Company Consultant and Partner. He was inspired by Andy Taylor, then the CEO of Enterprise Rent-A-Car, at a CEO conference a few years before Reichheld published his seminal article in the December 2003, Harvard Business Review called, "The One Number You Need To Grow".
Since publication of the HBR article, popularity and adoption of NPS has grown considerably. Some reports suggest that two-thirds of Fortune 1000 companies use NPS. Its advocates attest to its motivational benefits, its simplicity, correlation with revenue and growth and its value as an indicator of loyalty. However, despite widespread adoption of one form or another, questions are now being asked about its reliability as a predictor of growth, the empirical evidence that backs it up, its accuracy and even the value of it as a useful business metric, let alone an indicator of loyalty. Some respected academics have gone so far as to say it's 'fake science' and does not predict growth and research has not supported claims that it's capable of predicting future performance (Sharp, 2008; Keiningham et al.,2007)
What is NPS?
NPS is essentially a satisfaction metric, it 'tests' the propensity or likelihood of customers to recommend an organisation to a friend of colleague. The central question, "How likely is it that you would recommend [organisation X] to a friend or colleague?", is often accompanied by more open-ended or 'driver' questions about, why or what's driving the response in that way. It's expressed as an integer, calculated by dividing responses to the central question into promoters (score 9 or 10), passives (score 7 or 8) and detractors (score 6 or lower), based on a response score to the central question between 0 and 10. The NPS score is the number of promoter minus the number of detractor responses. There is no weighting by response score and 'passives' are not factored into the calculations.
Variability by industry
It's no surprise that NPS results vary by industry, just as other customer satisfaction indicators do. There are many reasons for this, but differences start with the different perceptions and the more positive or negative images that B2B and B2C customers have of the businesses in different industries or sectors. In B2C, department stores and banks are perceived differently, in B2B suppliers of cleaning and janitorial services are perceived differently from professional services, or infrastructure and utility suppliers.
Many benchmarks have been developed with different average industry quartile scores. But even within an industry, perception is still a factor, there are differences in sub-sectors and so it's not clear, what good really looks like. What's the value of inter-company comparison? Does an NPS of +63 mean that we can generate more growth and revenue than our closest competitor with a score of +52, even though they already generate two times our revenue? So you have a score of +63, so what? What does that actually mean, based on this score, what are the immediate next actions you should take? Maybe a different non-industry segmentation would be more useful?
Variability by respondent
Respondents perceptions and proclivity are important, but not factored into NPS scoring and the probability of them actually recommending your products and services is likely to be lower than you'd like. From a practical point of view, people just don't have a lot of time to go around recommending all the companies, products and services they like at any given time, even ones they score highly in an NPS questionnaire. Just because they say they will recommend, does not mean they will.
We all know that what people say and what they do are often quite different. Sometimes, on reflection, there's just not enough in it for them, they may be prevented or constrained in what they are allowed to recommend, there's no opportunity to recommend or no ones interested or listening. And then of course, things fade from memory, and without stimulus any promoter, even passives can disappear. All this means that intention to recommend, doesn't really mean a lot and it's actual behaviour and action we should focus on.
Good customers and promoters
In some quarters there's an assumption that promoters are good customers, they help drive growth and revenue, and detractors are bad ones and that actions should therefore be predicated on the 'good customers', the promoters or conversion of all customers into promoters. This is a dangerous assumption. Do promoters have a higher cost to serve, are they less sophisticated buyers, are they more promiscuous, in actual fact less loyal than they say. Are detractors, detractors for good reason, do they have legitimate concern, is there perception correct and something that should be acted upon? Do detractors or passives need to become promoters to be good customers, no they don't. There's no real evidence that promoters are good customers or that detractors are bad ones. Untested relative over-focusing on promoters and their perception could be a mistake.
Interpretation of the question
On a scale of 0-10, how would you respond to: How likely is it that you would recommend [organisation X] to a friend or colleague? What is it you are recommending? Who are you recommending to, which friend, what colleagues? Why are you recommending, what are you recommending? It's not clear from the question what is, or why a recommendation could or should be made. Understandably, many respondents won't think that much about this, they'll just take a stab at the answer. However, that's a problem for those who have to interpret and act on the results, especially if the business intends to act on the results.
Extent of, and manipulation of the sample
As with all sample based assessments the size and nature of the sample can have a profound affect on the results. Arguably for NPS to work as a predictor of future growth there has to be a 'Whole Customer', sample, a partial or random sample of the customer base isn't enough. Who identifies and selects participants becomes an issue if sales are involved in selection and are at the same time incentivised by NPS scores.
Behaviour vs attitude
It's long been assumed that attitude drives behaviour, but more recent marketing studies suggest this might not be the case and that behaviour drives attitude. In any event, what people say an what they do can be quite different. The correlation between attitude and behaviour in a survey is roughly 0.4 (Kraus 1995) and that's for 'very regular behaviour', NPS and voluntary recommendation is not regular behaviour. Promoters may not need or want to buy more, they may be able to influence colleagues to buy, but will they, can they influence other people or companies to buy, when will they have the opportunity to do so? There are too many variables to predict.
Repeat purchase vs loyalty vs recommendation
Repeat purchase is not a definitive indicator of loyalty and recommendation does not necessarily follow from loyalty. In consumer and corporate environments there is a natural loyalty, but it's not exclusive, it's polygamous. Even when given a wide choice, consumers and businesses select from a few alternatives. They're loyal to the extent that it simplifies their lives. This kind of natural loyalty is not a indicator of potential or actual recommendation. If loyalty is not a predictor of recommendation then the assumption that NPS provides insight into loyalty and consequently recommendation is wrong.
Alternatives
There are a few alternatives to consider, Customer Satisfaction Score (CSAT) and Customer Effort Score (CES). Other include analysis of Positive Word of Mouth (PWOM) and Negative Word of Mouth (NWOM), and its impact on intention to purchase.
What questions should you ask?
An improvement to the standard NPS question would include a range of tagged responses rather than just a scale, for example: "How likely is that you would recommend [organisation X] to a friend or colleague?" - answer very unlikely, somewhat unlikely, neutral, somewhat likely, very likely. Better still, as "Have you ever recommended [company X] to a friend or colleague?" - answer, yes in the last 3 months, yes but more than 3 months ago, no. A better test of loyalty would be "In the past 3 months have you considered leaving [company X] or switching to an alternative supplier?". Even a straightforward 'happiness' index might be better, two questions: "Are you happy with [company X]?", answer, yes very happy, happy enough, neutral, not really, not happy at all.
Summary
It's important to see NPS for what it is, it's an indicator of 'recommendation intent', perhaps a test of sentiment, at a point in time, from a sample of customers. It's not an indicator of behaviour, of loyalty, not a predictor of repeat purchase, retention or actual growth. Arguably NPS has some tenuous relationship to prospective growth through intended positive recommendation, but any correlation to actual revenue growth is much weaker than suggested, probably non-existent. It may be one of many metrics that a businesses uses to understand its commercial performance, but it should be treated with some caution as the assumptions around it and its evidence base as a predictor of loyalty, recommendation and growth are not proven, consequently its use or misuse, could be misleading and in the worst case damaging.
Like any other customer satisfaction metric, it's helpful in focusing attention on customers, if well managed it could be an indicator of growing or lessening propensity to recommend, that's something that might be good to know, but that's about it. NPS is certainly not 'the one number you need to grow'.