Is NPS still worth it on a content site?

Fred Reichheld and Bain & Company introduced NPS in the early 2000s as a measure of customer loyalty for a whole company. The idea was simple: people who would recommend you are likely to come back and to bring others, so the share of them should predict growth.

That design shows in the question. It asks about "us", the company, and it asks about a future action, recommending. Both make sense for a bank, an airline or a software subscription that a customer uses every week. They make less sense for a guide someone found ten minutes ago.

The main criticisms fall into three groups.

Note what the Gartner prediction was about: customer service and support teams, not NPS everywhere. It was a forecast, not a count of companies that stopped. It still points at the real weakness: one broad score doesn't tell a team what to change.

The sharpest criticism came from Reichheld himself. In his 2021 book Winning on Purpose, written with Darci Darnell and Maureen Burns, and in their Harvard Business Review article "Net Promoter 3.0", the authors admit that NPS has been badly misused.

They write that "self-reported scores and misinterpretations of the NPS framework have sown confusion". They describe companies linking scores to bonuses for frontline staff, staff pleading with customers for a top score, and one offer of free oil changes for a 10. They also point to "the response biases that plague the results of non-anonymized surveys", where customers know their answer can be traced back to them.

Their fix is a second number, earned growth, based on audited revenue rather than surveys. If you run NPS on a website, the same problems apply to you: a score someone is paid on, or one the customer knows can be traced back to them, tells you less than it seems to.

On a content site or a marketing page, NPS has some extra problems on top of the ones above.

Ask about the page the reader is on, in a form they can answer in one tap. These questions usually tell a content team more than NPS:

The CSAT calculator and the CES calculator work out those scores from your counts, and the NPS calculator does the same for NPS.

NPS is not useless. It still earns its place when:

If your NPS is already part of a board report, you don't have to drop it. Move it to the place where it fits, such as the account page for members, and use page questions everywhere else.

Asklet asks short questions on the page they are about, from one snippet. Its question types include Yes / No, a 1 to 5 star rating, multiple choice and NPS (0 to 10), so you can run CSAT, an effort question and NPS from the same question bank, on different paths.

For NPS, results show the score with promoters, passives and detractors for each page, next to the reasons readers gave. Answers are anonymous, and a page shows only counts until it has 10 responses, so a handful of answers never turns into a dramatic score.