Applying a filter can help you improve the sensitivity of your metric by refining the sample used in the analysis. Filtering is most often used to provide deeper analysis of how customers progress through a particular flow in your product.
An ecommerce company may seek to drive an increase in purchases, and while it is important to track that purchase rate globally, it is often valuable to see where in the process customers are dropping off. By filtering, you can see the behavior of customers who reach particular points in the funnel, such as abandonment by those who visited a product page, or those who added something to their cart.
Filtering is also used to create metrics that target only users who engage in a particular behavior, for instance observing the support ticket rate of those users who experience an exception or of users who completed the on-boarding process, as shown below: