This is great question. The short answer is that the more NPS responses you have to work with, the more accurate your NPS score.
How is your NPS Score and Response Rate Correlated?
This is a byproduct of user research that actually has little to do with NPS. With any research or feedback tool the basis of the rule is that the more data you have to work with the more accurate your summary and analysis will be. These tools and surveys get thrown off by outlier data. NPS is no exception to this rule. An outlier in NPS reporting can be someone who did not fully understand the product enough to pass judgement. Alternatively they could have just finished a really bad experience and their score reflects their current experience rather then their overall opinion. These outliers can be very useful to investigate and understand — however they can be dangerous if you let your precious time be consumed by them. If you spend all your time fixing the outliers you will find that you only helped a small portion of your user base while you neglected the concerns of your majority.
How do you fix the Outlier Problem?
NPS Data will be more accurate if you are able to collect more scores. When you collect more scores you are able to put less importance on the outliers and get a better overall picture of your data. Generally speaking if you collect less then 100 responses, then a single person can swing the results by a whole percentage point or more. For that reason we like to stay in the 250 range as a bare minimum then collect as much as we can above that. Many analytics companies will demand a minimum of 1,000 responses before correlating data, however for some companies that range is unrealistic. Another way to fix the NPS outlier problem is to not treat it like a problem. Instead look for patterns and correlations between the data and scores they supposed outliers have recorded. Maybe there is a commonality between them. For example if you explore the data you might find that the users you thought were outliers are actually users who have the same outdated browser and a portion of your service is broken for them using that outdated technology. You can then work quickly with those customers to upgrade their tech and improve their experience.