How A/B level matching works, and why louder sounds better

Play two identical mixes, one 1 dB louder than the other, and most listeners prefer the louder one. It sounds fuller, brighter, closer. Nothing about the mix changed; only the level did. This is one of the most reliable effects in listening, and it quietly decides a lot of mix approvals.

If you are choosing between Mix 19 and Mix 20, you want to hear the mix decisions, not the volume. That is what level matching is for.

Why louder wins

Human hearing is not flat. At lower levels we are less sensitive to low and high frequencies; as level rises, the bass and the top end become more audible (the equal-loudness contours, first measured by Fletcher and Munson in the 1930s and standardised today as ISO 226). So the louder version literally sounds like it has more bass and more air. Our ears read that as “better”.

The difference needed is small. Around 0.5 to 1 dB is enough to bias a comparison, and it is below what many people consciously notice as a level change.

What to match: peak, RMS or LUFS

There are several ways to measure level:

Peak level is the highest sample value. Two mixes with the same peak can sound very different in loudness, especially once a limiter is involved. Not useful for matching.
RMS is an average of the signal’s power. Better, but it treats all frequencies equally, while hearing does not.
Integrated loudness (LUFS), defined by ITU-R BS.1770 and used in the EBU R128 recommendation, measures loudness over the whole programme with a frequency weighting that approximates hearing, and ignores silent passages through gating. It is what streaming services use to normalise playback.

For comparing two versions of the same song, integrated loudness is the right measure. It is a single number per file, and the difference between two numbers tells you how much to adjust.

How the matching is done

Measure the integrated loudness of each version. Say Mix 19 measures −9.4 LUFS and Mix 20 measures −7.6 LUFS.
Take the quieter of the two as the target: −9.4 LUFS.
Turn the louder version down by the difference: Mix 20 plays at −1.8 dB.

Always turning the louder one down, rather than turning the quieter one up, matters. Raising a mastered file by 1.8 dB can push peaks over full scale and clip. Lowering can’t.

The gain is applied only at playback. The files themselves are untouched.

Switching at the same moment

Level is one half of a fair comparison. The other is time. If switching versions restarts playback, you compare your memory of the chorus with the verse that is now playing. Instead:

Keep both versions playing in sync, and switch between them at the same timestamp.
Crossfade very briefly on the switch (tens of milliseconds) so the change doesn’t click.
Loop the section in question, so you hear it several times in each version.

What matching can’t fix

Different arrangements. If Mix 20 has a new bridge, loudness over the whole song includes it. Matching makes the overall level fair, but the section you care about may still differ. Loop that section and trust your ears.
Different dynamics. A heavily limited master and a dynamic mix at the same LUFS still feel different. That difference is real and part of what you are judging.
Different lengths or offsets. Versions must start at the same point to compare at the same timestamp. A bounce with an extra bar of count-in will be out of sync.

The strongest test is one where you don’t know which version is playing. Have someone else switch, or use a tool that randomises A and B, and note your preference before the reveal. It is humbling, and it is the fastest way to stop arguing about a change nobody can actually hear.

In short

Match on integrated loudness. Turn the louder one down, never the quieter one up. Switch at the same timestamp. Then decide.