Why your mix sounds quieter than everyone else’s.

Streaming and mix hosting platforms measure the loudness of what you upload and turn it down, or occasionally up, so that everything plays back at a similar level. SoundCloud publishes minus 14 LUFS integrated as its reference and normalises playback towards it.

The consequence is worth stating plainly: squashing your master to be louder does not make it louder for the listener. It makes it quieter after normalisation, and it arrives with less dynamic range than the mix next to it. The competitive advantage that limiting used to buy has been taken away by arithmetic.

LUFSis loudness as ears perceive it, standardised in EBU R128 and ITU-R BS.1770. Audio is K-weighted, which is a filter that approximates the ear’s sensitivity, then averaged over 400 millisecond blocks. Quiet blocks below an absolute and a relative gate are excluded, so a long fade in does not drag the number down. The result is the integrated loudness of the whole set.

True peakin dBTP is the highest level the waveform actually reaches between samples. Your mixer’s sample peak meter cannot see it, because the peak often falls between two samples and reappears after conversion. Meters measure it by oversampling, typically four times, and reconstructing the curve.

Clipping is the recording already being damaged, not merely loud. A run of consecutive samples pinned near full scale is a flat top on the waveform, and the distortion is baked in.

Around minus 14 LUFS integrated is a sensible target for an uploaded mix, because it is where the major platforms normalise. MixGrade scores against exactly that: 3 points off per LU away from minus 14, capped at 40, plus a flat 12 point deduction if true peak exceeds minus 0.3 dBTP.

Keep a decibel or so of true peak headroom below zero. Lossy encoding adds peaks that were not in your file, so a mix that just touches zero before upload will overshoot after transcoding. Leaving that headroom costs you nothing you can hear.

A set is not a single track. Some dynamic range across an hour is the point, so do not chase a flat loudness curve. The energy sub-score is measuring shape, and shape needs contrast.

Almost always gain staging. Channel gains pushed until the meters sit in the red, a master output above unity, both basslines running at once during a blend, or a recording input on the controller that was already hot before the first record played.

The fix is unglamorous. Set channel gains so peaks land around the top of the green with the loudest track in your set, leave the master at unity, and check the recording level before you press record rather than after. Then use the bass swap so two low ends never stack.

MixGrade counts a clipped moment as four or more consecutive samples above 0.985 of full scale, merging events less than 50 milliseconds apart so one distorted passage is reported once rather than fifty times. Each moment costs 18 points from the cleanliness sub-score, and each one arrives with a timestamp, because the fix is at the source, not in the file.

Clipping removes information. A limiter applied later reduces the peaks around the damage without restoring what was flattened, and normalising a clipped file just moves the distortion to a different level.

This is the one measurement on the report that is genuinely non-negotiable. A tempo drift is a skill you build. A clipped recording is a set you play again.