It won't be as black and white as that.

For instance - if it's a weather website and there is a button on there for Report a Tsunami then that's almost never going to get clicked, but that doesn't mean it's not needed.

It's all going to depend on a variety of factors. Business requirements / politics dictating that features need to be present, legal requirements, legacy features for old customers, providing awareness that a feature exists even if it's not used (for example you could have a link to the Twitter account on the page. That may not be clicked on but it shows the user that there is a twitter account for that company). It's never just as black and white as 'very few people clicked this feature over the last 30 days so lets scrap it'.

Sure, use the analytics as some data to trigger some investigation in to why things aren't getting used, but that's all the data should be for - prompting investigation. You shouldn't act on pure analytic data alone.