Net Promoter Score NPS (relational vs transactional; tNPS)
A loyalty gauge: the share of promoters minus the share of detractors on the "how likely are you to recommend us" question.
The fork, why two teams get different numbers
First, RELATIONAL NPS (periodic "how do you feel about us overall") vs
TRANSACTIONAL NPS (right after a specific interaction), same 0-10 scale,
different question, routinely compared as if equal. Second, the CALCULATION:
standard NPS = %promoters (9-10) − %detractors (0-6), reported as an integer
from −100 to +100; but some report a 0-100 or a raw percentage, and some
quietly move the cutoffs (counting 7-8 as promoters) to lift the score. Third,
the SAMPLING universe: all customers vs only engaged responders, with heavy
non-response bias. A vendor's "NPS of 62" and yours computed the textbook way
may not be the same statistic at all.
The trap
NPS discards the middle (7-8) and is a difference of proportions, so it is noisy on small samples, a swing of a few detractors moves it several points, inviting over-reading of random wobble as a trend.
- The full trap, worked on a real export
- Every formula variant, spelled out
- The reconciliation anchor, what to tie it to and when to refuse
Reference: Reichheld / Bain & Company / Satmetrix (NPS origin · reference) · survey-methodology convention