For most of the CD era, the loudest record won. Not the best mixed one, the loudest. Labels pushed engineers to master albums hotter than whatever was already on the radio, because a track that jumped out of a car speaker felt more exciting than the one before it, and excitement moved copies. Engineers obliged by squeezing the gap between the quiet parts and the loud parts until almost nothing was left. The industry called it the loudness war, and it ran for roughly three decades before the technology that everyone now uses made it pointless.
Understanding what got lost takes one idea, and it is not complicated. Dynamic range is the distance between the softest and loudest moments in a recording. A drum hit has a sharp spike at the front, called a transient, and that spike is most of what makes a snare sound like it was struck rather than pressed. Compression and limiting pull those peaks down so the whole track can be raised without distorting. Do it gently and you get a record that holds together on cheap speakers. Do it hard for the whole song and every hit lands at the same level, which the ear reads as flat, tiring, and strangely small.
The clearest casualty is famous. The 2008 Metallica album Death Magnetic was mastered so hot that the waveform looked like a solid block, and fans found that the versions released inside the game Guitar Hero, mastered separately with more room, sounded noticeably better. Thousands signed a petition asking for a remaster. It became the case study everyone points to, because the same performances proved that the mastering decision, not the band, was the problem. Similar complaints have followed dozens of reissues where an older, quieter master beat the new one.
Then streaming changed the math without asking anyone. Spotify, Apple Music, YouTube, and Tidal all apply loudness normalization, which measures how loud a track sounds over its whole length and adjusts playback so every song hits a similar level. The common target sits near negative fourteen on a scale called LUFS, though platforms differ and some have shifted their numbers over time. The effect is the same everywhere. A track mastered far hotter than the target does not play louder for the listener. It gets turned down to match.
That single behavior flipped the incentive completely. Two songs now arrive at your ears at roughly the same perceived volume no matter how they were mastered. The one that was crushed shows up quieter than its creator intended and carries all the damage anyway, since normalization only lowers the level, it does not restore the peaks that were flattened out. The one mastered with breathing room shows up at the same loudness with its punch intact. Crushing a mix in 2026 buys nothing and costs the thing that made the drums hit. The war was won by a setting in a playback engine.
Plenty of people in music have not adjusted. Artists still ask for masters that compete with a reference track they love, unaware that the reference is being turned down on the same platform. Some engineers cite the listener who turns normalization off, or the club system, or the streaming service that handles quiet tracks differently. Those are real edge cases and they are shrinking. Normalization is on by default nearly everywhere, and most listeners never open the settings menu, which means the default is the market.
There is something useful here for anyone who records anything, including people making podcasts and video. Loud is not the same as clear, and the instinct to fix a weak mix by raising the whole thing has almost always been the wrong instinct. Space is what makes a recording feel large. If a chorus is supposed to feel bigger than the verse, it has to actually be bigger, which requires the verse to sit lower. That was true before the loudness war and it stayed true through it. The difference now is that the machinery finally stopped rewarding the shortcut.
