I also have some questions/concerns over the NPS. Online surveys, which is effectively the instrument of NPS, typically yield statistically incorrect data due to, often, some flavor of self-selection bias. If you think of an online survey as an experiment, they rarely allow enough control over the sample to mean much. However, that's not to say it's impossible to properly conduct an NPS, just that it's probably very easy to get wrong which may paint a false picture.