Say no to low-value features
In most code bases, you can get 80% of the business value with 20% of the features. Imagine if the time spent on the other 80% of features had been spent on high value features - you'd have four times the business value! This is what Google realized in 2011: they needed more wood behind fewer arrows. I have great respect for Google's willingness to kill products that aren't going to have major impact. So many organizations have struggling features that get blindly maintained and never get killed. This is a tragic mistake and a massive tax on the organization.
Engineers are the gatekeepers
"The difference between successful people and really successful people is that really successful people say no to almost everything" - Warren Buffett
Measure feature value aggressively
Nearly dead code is much worse than dead code
Nearly dead code must be maintained and is much harder to remove than dead code. Low-value features are nearly dead code - just waiting to be deprecated. It's not always obvious which features are low-value unless you measure their value from the start. Avoiding adding low-value features is much easier than removing them; removing them is much easier than maintaining them. When you do add a feature, release the simplest possible version in a small roll out. Measure its value against a business objective, and be ready to retract it quickly if it doesn't deliver value.