There are some features in software I rarely use. But those times I do use them they are utterly essential. If I find such feature has been removed I am incensed.
Metrics can tell that story though so you’re arguing a straw man.
Example: If you see that 99% of users have never used a function ever - you have a pretty good idea that it needs to be reworked or removed. You may also see a function that is used by 80% of users once a month, that you may opt to keep.
It's not so much that ubiquitous telemetry can't identify this, it's whether it's better for this than a focus group. You can have background telemetry with the focus group so you're not just giving customers what they say they want instead of what they need.
I'm not sure. While I understand that developer time needs to be cut down or restrained sometimes - though perhaps not at Amazon in this case, which concerns their core business -, your example could merely turn out to be a way of losing 1% of the users. Usage statistics alone cannot tell you whether your users hate or like a feature. Some features are always going to be used more than others.
What if that feature costs 30% of dev time? Without being able to measure you wouldn’t be able to make a good judgement. Imagine how science would work without experiments?
Wouldn't focus groups work better AND respect your users?
Devs think it is either telemetry or develop blind but in reality software was developed (and possibly was better) before telemetry using focus groups.
Metrics can tell that story though so you’re arguing a straw man.
Example: If you see that 99% of users have never used a function ever - you have a pretty good idea that it needs to be reworked or removed. You may also see a function that is used by 80% of users once a month, that you may opt to keep.