True Peak on Spotify: Half of 111,538 Releases Sit at 0 dBTP

We measured the true peak of 111,538 released tracks as the decoded Spotify stream delivers them. 48.5% sit at or over 0 dBTP, 81.2% are above the -1 dBTP that every mastering guide recommends, and only 18.8% leave that headroom. The median release peaks at -0.01 dBTP. House, dance, afrobeats and hip-hop run the hottest. Jazz, score and ambient leave the most room. Pop sits almost exactly on the line.

I have been mastering records for 20 years, 19 Billboard Top 20 albums, and "what true peak should I set" is the question I get most often after "how loud." The usual answer is -1 dBTP. So I wanted to know what the records people actually release do, not what the forums say they should. Here are the numbers.

True Peak by Genre

19 lanes, 99,403 tracks. The remaining 12,135 tracks sit in genres too small or too mixed to lane and are in the overall figures only.

Read the first column as the genres where the limiter is doing the most work. Dance, house and hip-hop have six in ten releases touching or passing 0. The club genres are the clearest case: house has 92% of releases above -1 dBTP, the highest share of any lane, and its middle half runs from -0.16 to +0.28. There is no headroom convention in that world at all.

The bottom of the table is the acoustic end. Jazz and score leave a median 1 to 2 dB, and more than half of each lane sits at or below -1 dBTP. Singer-songwriter is interesting: a median of -0.48 and only 16% at the ceiling, with the same loudness pressure as pop and a lot less appetite for it.

Where the Whole Field Sits

Four in ten releases land in a half-decibel band just above zero. That is not an accident and it is not clipping in the master file. It is a master set to a 0 dBFS ceiling, or close to it, and then the codec's reconstruction adding its overshoot on the way out. The 8.2% over +0.5 are the ones where that overshoot has real teeth.

What Headroom Costs

Run down the right-hand column. Each step toward the ceiling is worth about 1 dB of integrated loudness: -13.3 at the -1 to -0.3 band, -12.3 just under zero, -11.1 just over. So the honest cost of a -1 dBTP ceiling, against the hot end of the field, is about 1 to 2 dB on the meter.

Spotify then normalizes most of that away. If you have not read it yet, the loudness numbers show how little the extra decibel is worth once the platform turns everyone down to the same level. The true peak is the part the platform does not fix: an overshoot that already exists in the file goes through the encoder and comes out worse.

Why -1 dBTP Is Still My Ceiling on Most Records

Not because the field does it. The field clearly does not. Because of where the record is going to be played. A master that touches 0 on the way into Instagram, TikTok or YouTube gets re-encoded at a lower bitrate than Spotify uses, and the overshoot grows with every pass. On a phone speaker that comes back as grain on the transients and a splat on every kick. I wrote up the whole chain in why your music sounds bad on social media, and the first fix in that list is the ceiling.

So the rule I use: records that will live on phones and in video get -1 dBTP. Club records cut for a system, where the field sits at -0.3 and above, I will take closer to the line, because the playback chain is a known quantity and the extra half decibel is audible on a big rig. Either way the balance comes first and the level comes last. A master that reads right at -1 dBTP is a mix that was already right, not a limiter setting.

If you want to know where your own master sits against the lane it is competing in, the Sonic Analyzer reads the true peak and the loudness of your file and puts it on this table. And if the number comes back hot and you would rather not learn limiting the hard way, mastering is the part of the job where this gets decided.

How This Was Measured

The source. Every figure above comes from the Sonic Analyzer rig at Freshly Baked Studios, which plays each released track from Spotify and measures the decoded stream. That is deliberate: it means the true peak includes the intersample overshoot the codec adds, which is what a listener's player actually sees, rather than the cleaner number a mastering engineer reads off the WAV.

The set. 111,538 tracks scanned between March 20 and October 18, 2026, each with a true peak reading and a primary genre. The weeks in which the rig was later found to be reading the stream twice were excluded in full, and any track in the rig's quarantine list was excluded as well. Genre lanes are built from each track's primary genre label; 99,403 tracks fall into the 19 lanes in the table and the rest are counted in the overall figures only.

Loudness. Integrated loudness is the rig's reading of the same capture, used here only to compare the true-peak bands with each other. The genre loudness study above used a larger and earlier frame and projects to whole-track values, so its medians are not the same numbers as the band medians here.

What it isn't. A release-year view. The scan date is when the rig read the track, not when the record came out, so there is no decade trend in this set. Nothing is estimated.

Where the numbers come from: every figure above was computed by direct SQL against the Freshly Baked Studios analysis database on October 18, 2026.