Why level-matched A/B matters in mastering

A mastered version is usually louder. If I compare it directly with the original, I am testing loudness and processing at the same time—and loudness gets an unfair vote.

There are two honest questions to ask. “What will the finished master feel like at release level?” and “Did the processing itself improve the track?” The first needs a normal before/after. The second needs level matching.

The trap is not stupidity; it is perception

A small gain difference can make a version feel clearer, wider and more detailed before any of those properties actually changed. That is why good mastering decisions are easy to overestimate when the limiter also adds several loudness units.

Our first real raw/master pair made the problem concrete. The supplied master was only 0.112 LU louder than the raw file, while PolishMyTrack’s adaptive result was 1.63 LU louder. A normal A/B correctly demonstrates the release impact. But it cannot tell me whether the low-mid reduction, air lift, stereo move and de-essing were better. For that, the mastered side must be attenuated by the measured loudness difference.

What I listen for after matching

• Transient shape: did drums keep their attack, or did density replace impact?
• Vocal comfort: is sibilance genuinely calmer, or is the whole top end simply darker?
• Low-end definition: can I follow the bass more easily without losing its weight?
• Stereo focus: is the center more stable without shrinking the arrangement?
• Section movement: does the chorus still feel larger than the verse?

I switch frequently and compare like with like: chorus against chorus, the same transient against itself, at the same playback position. Long gaps make auditory memory less reliable.

Why PolishMyTrack names the attenuated side

A level-match button can become another black box. The listener should know which version is being turned down and by how much. PolishMyTrack calculates the difference from measured loudness and labels the attenuated side. It also keeps release-level playback available because the emotional “finished record” check still matters.

The two modes are complementary. Release level answers whether the chosen loudness works. Level matched answers whether the tone, dynamics and image earn their place after loudness stops helping.

A useful result can be “do less”

If the master becomes less convincing after level matching, that is not a failed test. It is the test doing its job. Back off the limiter, reduce the tonal move or keep the original. An automatic system should be able to leave a control at zero without treating restraint as an error.