Why “is it good” is the wrong question
Play a demo to a friend and you get a verdict: it is good, it is not quite there, it needs something. Play it to someone whose job is evaluating music and you get a location: the chorus arrives too late, the second verse repeats the first, the vocal is buried from 1:20.
The difference is not sensitivity. It is that professionals do not evaluate a track as one thing. “Good” collapses six separate judgements into one word, and collapsing them destroys the only information you needed, which is what to do next.
What follows is those six passes, in the order they are actually made. The order matters: each one assumes the previous is settled, and a fault found in pass two makes passes four and five irrelevant until it is fixed.
You can run all of this on your own material. The obstacle is not skill, it is that after the fortieth listen you no longer hear the track, you remember it. Every pass below therefore includes a way of checking that does not depend on your ears being fresh.
Pass 1: the first thirty seconds
This gets its own pass because it is where the decision is actually made, by A&Rs and listeners alike. Nobody working through a hundred submissions listens to all of them in full, and streaming platforms register a play at around thirty seconds anyway.
- What happens in the first second? If the answer is a slow fade or an atmospheric wash, the clock is running and nothing has been offered yet.
- When does the track first lift and hold? Not a single hit, a section that sustains. Under 30 seconds is healthy; past 45 you are relying on patience.
- Is there a reason to stay? A voice, a sound, a phrase. Something the listener has not heard exactly that way before.
This is measurable rather than a matter of taste: the point where energy first rises and sustains can be located precisely in the audio. Five tests for whether your chorus is actually memorable covers what to do when the answer is bad, and the cheapest fix is almost never rewriting. It is cutting the intro.
Pass 2: the song underneath
Strip away the production and ask whether anything is there. The classic version of this test is playing it on one instrument with one voice: if it collapses, the arrangement was carrying it.
- Structure and length. Does the chorus arrive when it should, does the bridge earn its place, is the track between 2:30 and 3:30 for a streaming context? The wider structural framework.
- The lyric. What is it about in one sentence, is the language concrete or abstract, does anything change between the first verse and the last? Six questions a critic asks.
- Repetition. Count the actual occurrences of the hook phrase. Eight to twelve across a track is the working range, and writers are consistently surprised by how far below it they sit.
- Prosody. Say the hook line as speech and notice the stressed syllable. If the music emphasises a different one, the line will not lodge.
A fault here is expensive, because it cannot be fixed downstream. This is the pass where the honest answer is sometimes that the song is not finished, and no amount of production will make it so.
Pass 3: the performance
A&Rs listen to performance separately from writing, because a strong song with a flat performance and a slight song delivered with conviction are completely different problems.
There are five dimensions, and treating them as one is what makes vocal feedback useless. Pitch, timing, diction, delivery and breath, each with its own fix:
- Pitch. Measurable in cents: a professional take averages 5 to 15, beyond 25 errors become audible. How to check intonation without perfect pitch, and which deviations are actually phrasing.
- Timing. Sitting behind or ahead of the beat is a choice; drifting between them across sections is not.
- Diction. If the words are unclear in solo it is the take; if they are clear in solo and unclear in context it is the mix.
- Delivery. The only dimension with no technical remedy, which is why it deserves the most attention while recording is still possible.
- Breath. Running out at the end of long lines produces flat notes and lost consonants, and gets misdiagnosed as a pitch problem.
Pass 4: the production
Now the recording rather than the song. The question is not whether it sounds expensive but whether anything is getting in the way of what was written.
- Is it clear? Congestion in the low mids makes everything indistinct. Where mud lives and how much to cut.
- Does the lead survive? A vocal that disappears in the chorus is almost always masking rather than level. Why turning it up never solves it.
- Is the problem the mix or the master? Faults between two elements belong to the mix; faults affecting everything equally belong to mastering. A symptom by symptom guide.
- Does it hold up against the genre? Frequency balance, dynamics, stereo imaging and transients.
Concrete action: compare against two reference tracks at matched loudness. Almost every A and B comparison people make is invalid, because the louder version sounds better regardless of its actual quality.
Pass 5: the technical floor
This pass is different from the others: it is pass or fail rather than a matter of degree, and it is the one an A&R makes without consciously noticing. Audible distortion in the first ten seconds ends the listen, and the song never gets evaluated as a song.
The six values are objective and repeatable, computed from the file rather than judged. The full pre-release pass covers them with the order to fix them in:
- No clipped samples, and true peak at or below -1 dBTP. The two kinds of loudness damage, one permanent and one that only appears after encoding.
- Integrated loudness around -14 LUFS. What each platform normalises to, and why louder gains you nothing.
- Loudness range above roughly 3 LU, phase correlation positive, spectral balance in line with the genre.
The reason this pass exists separately is that it is cheap to fix and fatal to ignore. Everything above involves judgement and effort; this involves measurement and an afternoon.
Pass 6: where it sits
The last pass is the one artists skip and A&Rs make first, silently, before the music even starts: what is this, and who is it for?
- Comparable artists. Three names, correctly sized, on a stated dimension. How to build a set that works in a pitch rather than one that flatters you.
- Audience and listening context. Who this is unmistakably for, and where they would play it. How to define an audience that excludes people.
There is a consequence most artists never consider: the positioning determines the standard you are judged against. A rough breathy vocal is character in indie folk and a technical problem in contemporary pop. Wide dynamic range is a virtue on an acoustic record and a liability in a club context. Choosing where you sit is also choosing which criteria apply, so it belongs in the evaluation rather than after it.
Reaching a verdict
An evaluation that ends in a number is entertainment. One that ends in a decision is work. There are only three outcomes:
- Release. Passes one through five are clear and the positioning is decided. Nothing is gained by another month.
- Revise. One specific pass failed, and you know which. Fix that, then re-run from there rather than from the top.
- Shelve. Pass two failed, meaning the song itself is not finished. This is the hardest verdict and the one that saves the most time, because every stage after it multiplies whatever you started with.
One caution on scores, including ours. Technical measurements are objective and repeatable: the same file always produces the same numbers. Creative and commercial assessments are indicative estimates that can vary between runs, and they are not forecasts. What is genuinely knowable about a song before release draws the line explicitly, and it is worth reading before treating any number as a verdict.
Used properly, a score is a prompt to look somewhere specific. A high one is permission to stop worrying, not a prediction of success. A low one is a reason to check, not a judgement. The value was never in the number: it was in having made six separate passes instead of one vague one.