NPS - Net promoter score
Net promoter score explained: what NPS measures, and what it throws away
Net promoter score is a loyalty measure built on a single question: how likely are you to recommend us to a friend or colleague, answered on a scale of 0 to 10. The score is the percentage of promoters minus the percentage of detractors, and it runs from minus 100 to plus 100.
Almost always written as NPS. It is a whole number, not a percentage, which is why writing it as "an NPS of 40 percent" marks the report as second-hand.
How is net promoter score calculated?
Responses are sorted into three buckets, then two of them are subtracted:
- Promoters: scores of 9 and 10.
- Passives: scores of 7 and 8. They sit in the denominator and contribute nothing to the score.
- Detractors: scores of 0 through 6. All seven of them, weighted identically.
NPS = percentage of promoters minus percentage of detractors
Net promoter score: a worked example
A service department gets 200 survey responses back in a month. 118 score 9 or 10, 44 score 7 or 8, and 38 score somewhere between 0 and 6.
Promoters are 59 percent, detractors 19 percent, so the NPS is +40. Note what just happened to the 44 passives: 22 percent of the sample answered, and their answers changed nothing. A customer who scored you 6 and a customer who scored you 0 also count the same, and a store with 118 tens scores exactly the same as a store with 118 nines.
Where NPS came from
Fred Reichheld introduced it in "The One Number You Need to Grow" in the December 2003 Harvard Business Review, developed with Bain & Company and Satmetrix. The claim was strong: one question predicted top-line growth better than the long satisfaction surveys it was meant to replace.
That claim is the reason it spread, and it is also the part that has not held up well.
The honest criticism of NPS
- It discards most of the answer. Collapsing an 11-point scale into three buckets throws away the information you paid to collect, and roughly doubles the margin of error around the resulting figure compared with simply averaging the scores.
- Small samples swing it wildly. Below about 200 responses the error band commonly runs to double digits, so a department-level score built on forty surveys can move ten points on three responses. Detecting a genuine improvement over time needs a sample far larger than most stores collect.
- The growth claim did not replicate. Keiningham, Cooil, Andreassen and Aksoy, writing in the Journal of Marketing in July 2007, used longitudinal data from 21 firms and more than 15,500 interviews in the Norwegian Customer Satisfaction Barometer and could not reproduce Reichheld's finding that NPS predicts revenue growth better than other loyalty measures.
- It is widely gamed. Bain, the metric's own home, names linking NPS to frontline pay as its most common and most damaging misuse. Once a bonus depends on the number, staff start coaching customers toward 9 and 10, the score rises, and the feedback loop the survey existed to create stops working. Reichheld returned to exactly this problem in Net Promoter 3.0 in the Harvard Business Review in 2021.
- It says nothing about why. The number is a direction, not a diagnosis. The free-text answer underneath it is usually worth more than the score, and it is the part that gets left unread.
None of this makes NPS useless. It makes it a rough directional signal that needs a decent sample, a stable method and no money attached to it. Read alongside CSAT and customer effort score, it earns its place. Read alone, on 45 responses, with a bonus riding on it, it is theatre.
Why the manufacturer's survey usually matters more
In a franchised dealership, NPS is your number. CSI, run by the manufacturer, is theirs, and only one of the two is tied to franchise standards, bonus money and sometimes vehicle allocation.
That difference decides where attention goes, and it should. The manufacturer sets the questions, the sample and the scoring, and none of it is negotiable. NPS is worth running in parallel because you control the timing and can survey moments the OEM survey never reaches, like the day after a large repair was approved rather than the week after delivery.
Both are worth watching because both are moving. CDK's Friction Points Study 2026 reported dealership NPS falling from +48 to +29 in a single year, which is a large move by the standards of a metric that usually drifts. Whatever else the number is, it is not currently flat.
Using it without fooling yourself
Three rules make NPS survivable. Keep the method frozen, including question wording, timing and channel, because changing any of them breaks the comparison you are trying to make. Report the sample size next to the score, every time. And never attach compensation to it.
Then treat the score as a trigger rather than a result. A drop is a reason to read the comments and pull the repair orders behind them, which is where the actual answer lives. The number itself has never fixed anything.
