Moving beyond the Net Promoter Score


Share on LinkedIn

I read Khadeeja Safdar and Inti Pacheco’s recent story for The Wall Street Journal with great interest. “The Dubious Management Fad Sweeping Corporate America” highlighted the shortcomings of the Net Promoter Score, or NPS, as a measure of customer satisfaction for businesses. There is no need for me to revisit those shortcomings which Khadeeja and Inti covered so brilliantly, but if we accept that the end is near for NPS, what does the future hold? Is there a better way for businesses to measure customer satisfaction?

Let’s look at the differences between measuring customer engagement (CE), and customer success (CS). CE is the softer side of a customer’s relationship with a brand, how valuable and important it is to each individual customer based on their unique set of experiences over time. CS, on the other hand, is about commercial results ─ upselling, cross-selling and acquiring new customers, all expressed in monetary terms. Importantly, superior CE makes it easier for businesses to meet revenue targets.

CS is easy to measure ─ double-entry accounting has been used since it was popularized by Bartolomeo de Pacioli in Venice in the 1400s. Measuring the softer side, CE, has proven to be very difficult. The best attempt to date is NPS.

Why are so many business executives now focused on CE? Smartphones changed everything. The Internet is the greatest communications advance since the invention of movable type, also in the 1400s. The Internet was already disruptive when only available on desktop or laptop screens, then Steve Jobs stuffed it into a mobile phone, and now there is no escape. Today customers are only a couple of taps or clicks or swipes away from a competitor’s marketing information, blog postings, reviews – the good, the bad and the ugly. Now we must anticipate what our customers will do. Be proactive rather than reactive, otherwise even longtime customers may defect before you knew they were even a potential flight risk.

As a result, the science of measuring CE has moved front and center. NPS is lagging and not keeping up, so we need a better answer. We need to build a better and actionable measure that reflects the recognizes that a customer’s relationship with brands area ongoing journey, not a destination. NPS captures a snapshot of a single point in time, this is not adequate. You could have dozens of interactions every year with a brand, so that snapshot may become obsolete very quickly. Our CTO, Alain Briançon PhD, highlighted this and other technical shortcomings of NPS in this recent byline:  “The Twilight of the Net Promoter Score.”

Imagine if you buy an SUV, you visit the dealer for your first service visit, and it does not go well, it is a disaster, took two days instead of one, no loaner car, your SUV was filthy when you got it back, etc.  You go on Facebook and Twitter and flame away.  But you are not going to drive your new SUV into a ditch and set it on fire. Your relationship with the global automaker that made the SUV is more complex than just one service visit.  It requires money to buy, investment in a brand, financing capacity, it is complicated.  You certainly might not recommend the brand to a friend anytime soon, but you love the SUV, it is big, it rides great, it takes all the gear and mountain bikes, it is safe, etc.

Even with serious issues, NPS has become a de-facto standard for measuring customer satisfaction because there was simply no other good method to measure CE.  Apple has proven to the world what a loyal and engaged customer base can do for a vendor in good times, and bad.  Customer engagement is very profitable.  Do Apple customers pay more because they are so engaged with the vendor? Good question, and most likely the answer is ‘yes’.

The soft side of business (engagement) is a lot harder to measure than the hard side (success).  Technology must help when you have millions of customers.  Moses came down from Mount Sinai with the Ten Commandments carved on stone tablets, they were not written on an iPad, you work with the technology you have. NPS was the best that we could do, given the limitations of technology applied at enterprise scale. Criticism of NPS falls on deaf ears today, because no alternative has been available until now.

An alternative to NPS

How about if we start with the view that every digitally recorded interaction with a customer should be unified into a single view of the “digital you”? This “digital you” includes all transactions and interactions (marketing, customer support) drawn from all the different IT silos across the enterprise. Another input on the horizon is the telemetry sent from your digital devices, including not just phones but also cars, smart speakers etc., but that’s a topic for another day.  One truth is undeniable: companies already have massive amounts of data about you, and there is always more that could be added to the mix, such as social media, browsing history, even macroeconomic events, etc. The NPS output relies on a single piece of feedback when there are literally billions of data points that could be included in that score. What do we do with all this data?  How about using the power of AI and machine learning to understand the relationship of a customer with a vendor?

Machine learning is a very powerful tool, which is very clever in its simplicity. So, what if we start with a very basic model using all the available data about the digital you? Then select a question of interest and ask the model to predict the answer. In the case of the SUV owner above, we could look at the data for each customer journey and the model would predict what kind of vehicle you are most likely to buy next. With a little data it may be way off the mark, but the model has one major advantage, it does not get tired, it can work 24/7/365.

When the model looks at predicting what you will buy, it does not just look at you, it looks at all the customer journeys in the data set. That could easily be millions of journeys.  What if the model could compare what you are doing with customers with similar journeys? Every time a customer buys a vehicle in this example, the model compares the purchase to its prediction, and if it got it wrong, it then changes the weights it assigns to all the individual interactions in the customer journey.  And it looks at every other customer and makes changes to those journeys which are similar.

In large enterprises, AI models can manage billions of data points, while analyzing trillions of patterns, all with the goal of making better predictions of what the customer will do next. Think of it as the cloud (power) x Moore’s Law (less and less expensive) x Internet (lots of data) = new and better predictions about what you would like to buy, and when.

The latest AI can take as much digitally recorded information as our customers have and use it to inform models which help large enterprises predict what their customers will do. Sinister? Well, if you don’t like getting spam, and vendors can send you offers specially tailored to what you need on a timelier basis, presumably that’s a good thing? Understanding how and when a relationship with a brand is deteriorating can help the company fix the issue before it becomes a problem. AI can measure customer engagement in dollars and cents, providing a view of the value of the brand to the customer right now, in terms of their commitment to continue using and buying those products and/or services. For the first time, a big company can see where it stands with each and every customer.

Why stop there? What if we could use this customer engagement measurement system as a predictive input for models predicting customer success, what you will buy, and when. Your customers may never need to opt out or click “unsubscribe” again. With the right data, we can recommend the best actions a vendor can take across all their customer engagement and customer success KPIs.  AI promises to be a complete game-changer for large-scale enterprises looking for the best way to fix all the problems with NPS and go way beyond current capabilities. 

We agree wholeheartedly with the conclusion of Khadeeja Safdar and Inti Pacheco’s article, that NPS is ready for retirement. AI is waiting in the wings to replace it.

Jean Belanger
Currently CEO of After graduating (LSE), I joined Wood Gundy. I left to start a VC fund, investing in start-ups. 3 went public. After 15 years in finance, I decided to run companies rather than finance them. The first, programming tools vendor, Metrowerks CodeWarrior, built most of the software used on the Mac in the 1990s. When Metrowerks was acquired by Motorola, I was named VP Biz Dev for their semiconductor business, where I invested over $450M in M&A in 14 months. After Motorola, I started data science supply chain software and IOT pioneer, Reddwerks.


Please use comments to add value to the discussion. Maximum one link to an educational blog post or article. We will NOT PUBLISH brief comments like "good post," comments that mainly promote links, or comments with links to companies, products, or services.

Please enter your comment!
Please enter your name here