One of the many things music and software have in common is their complex determination of truth.
Both have a low and high bar for truth. The low bar for software is mechanical: does it even run? Does it produce the desired result, and in a reasonable period of time.
The low bar for music corresponds. We can all tell when a performer stumbles, loses the beat, cracks just short of the high note.
It takes a more skilled observer to perceive the high bar.
I saw an interchange on The Voice: while the judges were bantering, Kelly Clarkson sang a line from one of Alicia Keys' songs. Keys immediately asserts, with no malice but total certainty, "That's not how it goes." I couldn't tell exactly what Clarkson missed, but I don't doubt for a second that Keys did.
I wrote recently about how "Louie Louie" gets played wrong all the time. It's a simple I-IV-V chord progression, but the V chord has to be minor. It's a subtle change, but once someone points it out to you, you'll always hear it played wrong.
These are not high bar criteria though. They're subtler, elevated, but they're still about correctness, a binary pass/fail.
They can be measured.
A software program should handle weird input without crashing, use appropriate amounts of system resources. These are measurable criteria.
Good software needs good qualities in addition to its measurable quantities: it should use abstraction in the most useful way, clearly communicate its intent in source code.
Qualitative characteristics can't be measured. Our preferences for different qualities make up our taste.