Key findings
- Spotify turns 81% of these releases down at its default setting, by a median 4.2 dB and up to 8.5 dB.
- Apple Music's Sound Check turns 92% down, by a median 5.8 dB. Only 8% sit at or below its -16 LUFS target.
- 62% peak above 0 dBTP and 69% above -1 dBTP, the ceiling the streaming services recommend.
- The median release measures -10.6 LUFS, 3.4 dB louder than Spotify's -14 LUFS reference.
- Loud costs punch: releases quieter than -14 LUFS keep a median 13.6 dB crest factor; releases louder than -9 LUFS keep 10.2 dB.
- Quiet masters don't get rescued: 18% sit below -14 LUFS, and YouTube and Tidal play them a median 1.6 dB quieter because those services only turn tracks down.
Streaming services were supposed to end the loudness war. Every major platform now measures each song and turns loud ones down to a common level, so a crushed master gains nothing in volume and still loses its dynamics. We wanted to see how commercial releases actually land against those targets, so we measured 151 of them.
Every file went through the same measurement the services themselves use: ITU-R BS.1770-4 integrated loudness, plus 4× oversampled true-peak detection to catch the peaks that land between samples. Then we compared each result with each platform's published normalization target.
1Nearly every release gets turned down somewhere
At Spotify's default setting, 81% of the releases play quieter than they were mastered, by a median 4.2 dB. Apple Music is stricter: Sound Check normalizes to -16 LUFS, so 92% are turned down, by a median 5.8 dB. Deezer, at -15 LUFS, turns 87% down by a median 5.0 dB and as much as 9.5 dB.
| Platform | Target (LUFS) | Turned down | Median cut | Largest cut |
|---|---|---|---|---|
| Apple Music (Sound Check) | -16 | 92% | 5.8 dB | 10.5 dB |
| Deezer | -15 | 87% | 5.0 dB | 9.5 dB |
| Spotify (default) | -14 | 81% | 4.2 dB | 8.5 dB |
| YouTube | -14 | 81% | 4.2 dB | 8.5 dB |
| Tidal | -14 | 81% | 4.2 dB | 8.5 dB |
The practical point: making a master louder than the target doesn't make it louder for listeners. The platform takes the extra level back, and what's left is whatever dynamics survived the limiter.
2Most releases already peak above 0 dBTP
62% of the releases have true peaks above 0 dBTP, and 69% exceed -1 dBTP. A sample-peak meter can read below 0 dBFS while the reconstructed waveform goes over it. When a streaming service encodes those files to a lossy format, peaks like these can turn into audible distortion. That's why Spotify and Apple recommend a -1 dBTP ceiling.
3Louder masters keep less punch
Releases mastered at -14 LUFS or quieter keep a median crest factor of 13.6 dB. Releases louder than -9 LUFS keep 10.2 dB, about 3.4 dB less. Once a platform turns the loud master down to its target, that difference is what listeners hear: the quieter master hits harder at the same playback level.
4Loudness by genre
| Genre | Releases | Median LUFS | Median true peak (dBTP) |
|---|---|---|---|
| Metal | 10 | -14.7 | -2.0 |
| R&B | 8 | -13.0 | -1.4 |
| Indie | 8 | -12.7 | -2.5 |
| Rock | 8 | -12.4 | -2.5 |
| Classical | 8 | -12.2 | +0.0 |
| Country | 10 | -11.6 | -0.8 |
| Jazz | 9 | -11.2 | +1.3 |
| EDM | 10 | -11.0 | -1.0 |
| Hip-hop | 10 | -10.7 | +1.2 |
| Latin | 10 | -10.7 | +0.8 |
| Acoustic | 8 | -10.2 | +0.8 |
| Electronic | 10 | -10.2 | +0.9 |
| Pop | 10 | -10.1 | +0.5 |
| Gospel | 10 | -9.9 | +1.6 |
| Folk | 9 | -9.7 | +1.5 |
| Reggae | 5 | -9.3 | +1.6 |
| Bedroom pop | 8 | -8.4 | +1.6 |
Metal was the quietest genre in this set and bedroom pop the loudest. With 5 to 10 releases per genre, treat these as directional rather than definitive.
What this means for your next master
- Check true peak, not just sample peak. Aim for -1 dBTP or lower. The free LUFS check measures both with the same BS.1770-4 math.
- Pick a target on purpose. Around -14 LUFS for most streaming, -16 LUFS for Apple Music. For vinyl, about -16 LUFS with -1.5 dBTP and mono bass below 100 Hz.
- Don't chase loudness the platform will remove. Past the target, extra loudness mostly costs punch.
Use this data
Journalists, educators and creators are welcome to quote these findings and republish the charts with credit and a link. The summary numbers are available as a CSV download, and each chart links to a downloadable SVG. Suggested citation:
Questions, interview requests or a custom cut of the data: jeff@luvlang.studio.
Measure your own master the same way.
Upload a mix, hear the mastered version free, and see its loudness, true peak and score before you pay.
Open the studio