Why Spotify turned your track down: LUFS, true peak and the one-tap fix

You bounced a master that hits hard. On Spotify it sounds smaller than the track before it. Nothing is broken: the platform measured your track, decided it was louder or quieter than its target, and applied a gain change before anyone heard it.

That is loudness normalisation, and once you know the two numbers behind it, the problem becomes arithmetic instead of mystery.

→ LevelLab measures your bounce and normalises it in one tap, free on Android

The two numbers

Integrated LUFS is the loudness of the whole track, measured to ITU-R BS.1770 with gating so that silence does not drag the average down. It is what platforms compare against their target.

True peak (dBTP) is the highest point the waveform actually reaches between samples, measured with oversampling. It is not the same as your DAW’s sample peak meter, and it is the number that decides whether your track distorts after conversion to a lossy codec.

A third, loudness range (LRA), describes how much the loudness varies across the track. Not a pass or fail, but a useful sanity check: a mastered pop track usually lands between 3 and 8 LU.

The targets

Platform	Target	Ceiling
Spotify	-14 LUFS	-1 dBTP (-2 if you exceed the target)
Apple Music	-16 LUFS	-1 dBTP
YouTube	-14 LUFS	-1 dBTP
Amazon Music	-14 LUFS	-2 dBTP
Deezer	-15 LUFS	-1 dBTP
TIDAL	-14 LUFS	-1 dBTP
Broadcast (EBU R128)	-23 LUFS	-1 dBTP
Podcast (typical)	-16 LUFS stereo, -19 mono	-1 dBTP

Read the mechanism, not just the numbers:

• Louder than the target: the platform turns you down. Everything you crushed to gain loudness is still crushed, but now it is also quieter. You paid in dynamics and got nothing.
• Quieter than the target: the platform turns you up, which lifts your noise floor and any hiss along with the music.

This is why “louder wins” stopped being true. What wins is a mix that sounds good at the target.

How to check your own bounce

Measure the exported file, not the sound in the room. A phone microphone in front of a speaker can tell you something about balance, but the number that platforms use comes from the file itself.

1. Bounce as you normally would.
2. Measure integrated LUFS, true peak and LRA of that file.
3. Compare against the targets above; a decent tool shows pass or fail per platform in one view.
4. Look at loudness over time. A track that is on target overall but has one hot chorus will be turned down by its average, and the chorus will not benefit.

Landing on target without clipping

The fix is not “master louder”. It is:

1. Apply the gain difference between your integrated LUFS and the target.
2. Check what that does to the true peak. If your master sits at -18.7 LUFS and the target is -14, that is +4.7 dB of gain, and a peak that was -1.0 dBTP is now +3.7 dBTP: over the ceiling.
3. Hold the peaks with a true-peak limiter with a few milliseconds of lookahead, so the loud transients are contained rather than clipped. Limiting costs a little loudness; clipping costs the mix.
4. Export at the right depth: 24-bit if the platform accepts it, 16-bit with dither if not.

This is exactly what a one-tap normalise does: gain to target, true-peak limiter under the ceiling, new file out, your original untouched.

The A/B trap, and how to avoid it

Comparing your master to a reference at their own volumes is a rigged test: whichever is louder sounds better for the first ten seconds, every time. The way mastering engineers compare is at matched loudness: set both to the same LUFS, then switch. Suddenly you hear the mix instead of the level, and the reference stops sounding magic.

Measured with LevelLab: integrated, short-term and momentary LUFS, true peak, LRA, a scrubbable loudness timeline, pass or fail against eight platforms, and one-tap normalisation with a true-peak limiter. Free on Android, files measured on-device and sent nowhere. Album mode and matched-loudness A/B are a one-time Pro unlock.