Quality and impact aren’t the same. For example you can spend extra time making an administrative settings UI look better. Or you could reduce the wait time on an action performed from the administrative screen. Both could be objectively "better" — higher quality — from the builder's POV. But that doesn’t mean the changes make a difference to customers.
Think of each aspect of a design in two dimensions: how much better it is in terms of quality and how much it matters to customers. You can imagine each force working on the design as a vector. The sum is the net win/loss given the trade-offs in the design.
Example: We’re considering adding a new setting to notify people when a to-do is completed. The new feature does a better job than baseline on setting expectations. But it complicates the rules a bit. The complication doesn’t produce a big negative impact and it’s not so bad that we can’t live with it. So we shipped it.