I wrote about a tough decision we made in 2010 to shut down Blogger's oldest feature: FTP publishing. (Back in 2010 I wrote about the announcement here.) The gist of the post is captured in the quote below, but I encourage you to read the whole thing.
It’s easy to say yes when a customer (or prospect) asks for a new feature: after all, if it’s just a day or two of engineering time, why not? But you quickly lose sight of the product you’re building: your product no longer has a coherent vision, and each new feature brings with it uncertain support costs that will last as long as the feature remains. Much harder — but much more important — is the discipline to question whether the feature is a required piece of what you’re building. New or old, easy or hard — if the feature does not support the overall product goals, it has to go. Customers and team-members alike respond to that discipline — particularly if it results in better support, more predictable development, and a clearer understanding of what it is you’re trying to build.