You can absolutely love a product, and even recommend it but hate individual features about it.
The final number doesn't 'mean' anything. What should I do with an NPS of 40? Use the CES (Customer Effort Score) instead:
Asking hypothetical questions (Would you recommend...) are inherently risky for predictability. The CES asks how easy something was immediately after it happened, examining an event that already happened.
Asking on a feature level gives you very targeted feedback about a very specific tool or feature. A team can get 2 useful insights from this: An average score and splitting it up by vote. (See graph)
You can then see what unites people that voted a 1 compared to those that voted a 5. Are they using the tool differently? Are they a completely different segment? Is it a tool that people are indifferent about? (Big Middle-finger pattern) or something that people hate or love (Strong 1 and 5 distribution)
How to implement
The CES is best asked right after a success happened with a tool. A download, something was saved, do NOT interrupt the flow of the user before they completed the action.
Since the CES is triggered on a feature level you need to be careful not to trigger it too much. I recommend asking it not more than once per 30 days/user but this depends on your product and might even be way too much for high-usage products.
A comment field is optional, and the CES as well, do not force users to answer it and make it easy to dismiss (x) and capture "ESC" as a hotkey and click outside of it.
Measuring Word of Mouth
It is worth mentioning that the CES is not good in measuring overall how strong your word of Mouth is. This is where it makes sense to also ditch the NPS and look at it from the perspective of a framework like the word-of-mouth coefficient:
This article from Yousuf Bhaijee is excellent on the topic and worth a read
But Leah, I really love the NPS
Ok ok. Try this:
Doesn't they measure 2 different things?
In the end, it depends on what you want to tackle in my opinion / or what you are looking for.