Does Master Loudness Affect Whether a Song Breaks Out? We Measured 51,868 Releases

At the extremes, yes. Of 51,868 releases we tracked, the ones mastered between -13 and -9 LUFS were followed by the artist at least doubling their Spotify monthly listeners within 90 days 7.9% of the time. Everything outside that window: 6.7%. Quieter than -14 LUFS, 6.6%. Louder than -8 LUFS, 4.5%. With the artist's starting size and genre held constant, a master in the window carried 26% higher odds of a breakout, and it held in every size tier under 250,000 listeners and on 20 out of 20 random half-splits of the data. Same releases and the same yardstick as the breakout rate study.

I want to be careful with this one, because I've spent twenty years telling people I don't chase loudness, and I don't. The gear is good enough now that I can get a record to whatever number it needs as long as there's enough dynamic in it to begin with. So this isn't a post about turning it up. It's a post about what happened to the records that landed outside the range everybody competent masters to, in both directions, and the answer is that both ends lost.

The Data: Breakout Rate by Master Loudness

Every release below was measured by our analyzer for integrated loudness on the released file. "Broke out" means the artist at least doubled their monthly listeners within 90 days.

Integrated loudness	Releases	Share	Broke out	Breakout rate
Quieter than -16 LUFS	4,826	9.3%	300	6.2%
-16 to -13	10,484	20.2%	756	7.2%
-13 to -11	17,996	34.7%	1,442	8.0%
-11 to -9	15,314	29.5%	1,180	7.7%
-9 to -7	2,851	5.5%	172	6.0%
Louder than -7	397	0.8%	13	3.3%
All	51,868		3,863	7.4%

A hump, not a line. Two thirds of releases sit in the -13 to -9 window, and that's where the breakouts cluster. The window against everything else, by the artist's size on release day:

Artist's monthly listeners at release	In the window (-13 to -9)	Outside it	Releases in the row
500 to 2,000	30.5%	27.9%	2,434
2,000 to 10,000	20.1%	16.3%	7,156
10,000 to 50,000	9.8%	7.3%	11,986
50,000 to 250,000	4.0%	3.3%	14,817
250,000 and up	1.2%	1.1%	15,475

And the quiet end on its own, because it's the bigger group. Releases quieter than -14 LUFS, one in five of the sample:

Artist's monthly listeners at release	Quieter than -14	-14 and louder
500 to 2,000	25.8%	30.5%
2,000 to 10,000	16.1%	19.4%
10,000 to 50,000	6.7%	9.5%
50,000 to 250,000	3.4%	3.8%
250,000 and up	1.4%	1.2%

Size-adjusted, the quiet master carried 20% lower odds of a breakout. The loud end: releases past -9 LUFS broke out 5.7% of the time, past -8 LUFS 4.5%. That top group is 1,098 releases, 49 breakouts, so the count is thin, but the direction held on all 20 half-splits, and inside electronic music, where most of those records live, it was 3.1% against 6.7%.

Each lane has its own centre: electronic sits at -11.2 LUFS, rock -11.3, folk and country -11.5, pop and hip-hop -11.6, R&B -11.8, metal -10.7, jazz -14.8, classical -20.6. Measured against the lane's own median, records within 1 LU of it broke out 8.6% of the time, records 3 LU or more louder than their lane 6.1%, records 3 LU or more quieter 7.5%. The full lane table is in the average LUFS by genre study.

What I Take From It

The window is the window because it works. -13 to -9 is where a competent master of a dynamic record lands without anybody chasing anything. That two thirds of releases sit there isn't an accident, and that the breakouts cluster there isn't either. If your master is in it, this post has nothing to tell you. Go write the next one.

The quiet end costs more than I expected. One release in five came in under -14 LUFS, and for an artist between 2,000 and 50,000 listeners that's the difference between 9.5% and 6.7%. I know where a lot of those masters come from: the number -14 got passed around as "Spotify's target" for years, and people mastered to it. I wrote about why that reading is wrong in why your master sounds quieter than commercial tracks and why your song is quiet on Spotify. The short version: normalization doesn't lift a quiet master on every surface, and when it does, it lifts the noise and the softness with it. A record that sits 3 LU under everything on the playlist reads as smaller, and smaller doesn't get the repeat.

The slammed end loses too, and that's the part I'd underline. Past -8 LUFS the rate falls off a cliff, and it isn't because loud is bad. It's because getting past -8 on most records means the dynamic was shaved off, and what's left is the same thing for three minutes. I can get a record loud and keep it breathing, because I've spent twenty years finding every last bit of headroom in a mix before the limiter ever sees it. That's the craft. The number is easy.

It's your record. If the thing you're making wants to sit at -16 because it's a piano record and the silence is the point, that's the record. The jazz lane sits at -14.8 and the classical lane at -20.6 and nobody in those rooms is wrong. What I'd want you to know is that for a pop, rock, electronic or hip-hop record, the ones that doubled were overwhelmingly in the window, and the quiet exception is rarer than the internet says.

What I'd Do With This

Check the number on the released file, not the session. If it's under -14 and the lane you're in sits at -11, the master is leaving something on the table that the mix already paid for. If it's past -8, ask what the limiter is holding and whether the record still moves. And if it's in the window, stop reading about LUFS and go make the record better, because at that point the number isn't what's deciding anything.

If you'd rather have someone else hold the limiter, the rate calculator prices the mix and master to where you are as an artist. The window is where my masters land; keeping them dynamic inside it is what you're paying for.

How This Was Measured, and What It Doesn't Say

The data. 51,868 releases from August 2025 through April 2026, the same per-release table as the breakout rate study: one row per release with the artist's monthly listeners on release day, their peak within 90 days, a breakout flag when the peak was at least double, and the analyzer's integrated loudness (LUFS) on the released file. 3,863 releases (7.4%) carry the flag.

The tests. The bins are counts. The claims (the window is better, under -14 is worse, over -8 is worse) were each tested on twenty random half-splits of the data: find the direction on one half, check it on the other, report only if the direction agrees on at least 16 of 20 with a median p below 0.05. The window and the quiet end agreed on 20 of 20 both raw and in the size-adjusted model. The over -8 claim agreed on 20 of 20 raw; in the adjusted model the direction held on 19 of 20 but the count is too thin for the p-value to clear the bar, which is why I've called it a direction and not a number. Odds ratios come from a logistic model with starting size (log scale), genre lane and label flag as controls.

Genre is known for 45% of releases. Lanes come from track-level Chartmetric tags mapped through our taxonomy, not from the analyzer's own guess. The lane-relative figures use only the tagged 23,292; the overall bins use everyone.

Loudness range moves with it. The releases in the most dynamic third (highest loudness range, highest dynamic range) broke out less often, 0.87 to 0.92 times the rate of the rest. Those meters move with integrated loudness, so this is one finding, not two. I don't read it as "dynamics are bad"; I read it as the quiet, unprocessed master showing up on a second meter.

Growth is measured at the artist, not the track. If the artist doubled in the 90 days after a release, that release gets the credit, whatever actually drove it.

The source can be wrong. Data is data, and sometimes the data can be wrong from the source. Chartmetric's pulling system doesn't always work, which is why I run my own systems next to it. A missed snapshot moves a baseline, and a moved baseline moves a growth number.

Correlation. Nothing here says that mastering to -11 LUFS makes a record break out. It says the records in the window broke out more often in this window of releases, inside every size tier that has enough breakouts to count, and on every held-out half we tried. I'm standing behind that sentence and no further.

Where the numbers come from: every figure above was counted from the study's per-release table (51,868 rows) on 2026-10-08. The loudness bins, the window and quiet-end tables by tier, the lane medians, the lane-relative bins, the twenty-split holdouts and the size-adjusted odds ratios are in the study folder as study_01.json.