What is Normalization?

There are two fundamentally different types of normalization, and confusing them leads to bad decisions. Peak normalization finds the highest sample value in a file and scales the entire signal so that peak hits a target level — typically 0 dBFS or -0.1 dBFS. Every sample is multiplied by the same gain factor, preserving the relative dynamics exactly as they were. It is a simple, linear level change that makes the file as loud as possible without clipping. Loudness normalization is a more sophisticated process. Instead of targeting the peak, it measures the integrated loudness (typically in LUFS) and adjusts the overall level so that the perceived loudness matches a target value — for example, -14 LUFS for Spotify. This means a quietly mastered track gets turned up and a loudly mastered track gets turned down, so everything on the platform plays back at roughly the same perceived volume. The crucial difference is that loudness normalization penalizes over-compressed music — if your track is mastered to -8 LUFS, Spotify turns it down by 6 dB, and all that dynamic range you sacrificed for loudness is just gone for nothing. Streaming platform normalization happens automatically on the playback side — you do not apply it yourself. The platforms analyze your track's integrated LUFS after you upload it and apply a gain adjustment during playback. This adjustment is purely a volume change; the platforms do not compress, limit, or otherwise process your audio. Understanding this distinction is essential: normalization does not change the dynamics of your track, it only changes the playback volume.