Why there's a limit at all
Digital audio has a maximum representable amplitude — 0dB full scale — and any attempt to push the signal louder than that doesn't make it louder, it clips, flattening the peaks of the waveform into a shape that reproduces as harsh, crackling distortion rather than a genuinely louder sound. This is a hard mathematical ceiling, not a setting you can push past with a stronger boost; going over it just adds distortion on top of the loudness increase.
This is why a naive volume boost applied uniformly can make quiet passages nicely louder while simultaneously clipping and distorting the already-loud passages — the file's loudest existing peaks are what determine how much headroom is actually available before hitting that ceiling.
How much headroom a file actually has
A quiet recording — a voice memo recorded far from the microphone, or an old file recorded at low levels — often has significant headroom, meaning it can be boosted a good amount before clipping becomes a risk. A recording that's already close to full volume has very little headroom, and any additional boost risks clipping immediately. Checking a file's existing peak level before boosting tells you how much room you actually have.
Normalization vs. straight volume boost
A straight percentage or dB boost applies the same multiplier everywhere, which is simple but risks clipping if any part of the file is already loud. Normalization instead finds the file's loudest peak and calculates the exact boost that brings that peak up to (but not over) the maximum without any manual guessing — making it the safer default when the goal is simply “make this as loud as possible without distortion” rather than a specific creative loudness target.
Perceived loudness isn't just peak level
Two files can have the same peak level but sound noticeably different in loudness because human perception of loudness is more closely tied to average (RMS) level and dynamic range than to peak level alone — this is the whole basis of the “loudness war” in music mastering, where heavily compressed tracks sound louder overall despite having the same peak ceiling as more dynamic recordings. For most everyday boosting needs, though, getting the peak level right without clipping is what matters, and finer loudness shaping is a more specialized editing task.