Mix problem or master problem?

Something is wrong with the file. It sounds small next to the reference, or harsh on earbuds, or the low end wobbles on a decent system. The instinct at this point is to do something: push the limiter, add a shelf, try a different chain. That instinct is usually what makes it worse.

The question to ask first is simpler and almost nobody asks it. Is this a mix problem or a master problem? Because the two have completely different fixes, and applying the wrong one costs you a day and leaves the original fault in place.

What a master can actually change

Mastering works on the finished stereo file. Everything it can do, it does to the whole thing at once. That is the constraint, and understanding it answers most of the question on its own.

A master can change overall tonal balance — more air, less mud, a tilt one way or the other. It can change perceived loudness and how the record holds together under a limiter. It can widen or narrow the stereo image, control peaks, and make a set of tracks sit at consistent levels. All of that is real and all of it matters.

What a master cannot do is reach inside the balance. It cannot turn the vocal up without turning the snare up with it. It cannot fix a bass part that is fighting the kick, because both live in the same region of the same file. It cannot remove the sibilance from one word. It cannot undo a reverb decision.

If the fix requires touching one element without touching the others, it is a mix fix. There is no other test.

That single sentence resolves most cases. Before you do anything else, describe the problem out loud and notice whether your description names one element or the whole record.

Four symptoms and what they point to

Some faults are ambiguous. Most are not, and they fall into recognisable shapes.

• It sounds quiet, but pushing the limiter makes it worse. Master-side symptom, mix-side cause. Something in the mix is eating headroom — usually an unruly low end or one uncontrolled transient — and the limiter is reacting to that rather than to the record.
• The vocal disappears in the chorus. Mix problem, every time. No amount of broadband EQ on the stereo file brings a vocal forward without dragging the guitars and cymbals with it.
• It is harsh on earbuds but fine on monitors. Usually master-side, and often fixable with a gentle cut in the two to four kilohertz region. Sometimes it is one bright element in the mix, in which case you will hear the harshness follow that element.
• The low end is loud on one system and absent on another. Almost always a mix problem, and frequently a phase problem between kick and bass rather than a level one.

The pattern in all four is worth noting. When the problem is describable as a property of the whole record, the master can address it. When it is describable as a property of one part of the record, only the mix can.

Testing it without guessing

There is a quick check that settles the ambiguous cases. Take the file into any editor and apply a single broad move — a two decibel shelf, a touch more limiting, a slight narrowing of the sides. Then listen for whether the problem moves proportionally with everything else, or whether it stays put.

If the fault scales with the move, it belongs to the whole file and the master can reach it. If you can make everything else better or worse while the fault sits there unchanged, it is embedded in the balance and no stereo processing will lift it out.

The other half of the test is measurement, and it is faster. Integrated loudness, loudness range, true peak, and phase correlation between them describe most of what a master can and cannot do about a file. A track sitting at minus eight LUFS with a loudness range of two has already been flattened — whatever else is wrong, adding more limiting is not the answer. A track with correlation dipping below zero in the low end has a phase issue that lives in the mix.

The third answer nobody offers

There is a case that is neither, and it is more common than either of the other two: the file is fine and the reference is not comparable.

You are A/B-ing against a commercially released record that has been mastered loud. It plays back louder. Louder reliably sounds better to human hearing in a direct comparison, regardless of quality, and the effect is strong enough that trained engineers fall for it. Unless you have level-matched the two files, you are not comparing masters — you are comparing volumes.

Match them by integrated loudness, not by ear and not by peak. Then listen again. A meaningful proportion of the time the problem evaporates, and the correct action was to do nothing.

Level-match before you diagnose. Half the problems people bring to a master are a volume difference wearing a costume.

What to do with the answer

If it is a mix problem, go back to the session. This is unwelcome when the mix is signed off and the deadline is Friday, but the alternative is paying someone to fail to fix it on the stereo file. A good mastering engineer will tell you the same thing and will usually tell you exactly which element to look at.

If it is a master problem, it is tractable. Broad tonal shaping, controlled limiting, a sensible true peak ceiling, and consistency across the set. That is a well-defined job with a known method.

And if it is neither, ship it. The hardest skill in this part of the process is recognising a finished record and stopping, and the only reliable way to get there is to ask the diagnostic question before you reach for a plugin rather than after.

Two minutes of diagnosis saves the afternoon. It is the cheapest thing in the entire process and the step most often skipped.