How to Analyze a Mix: A Practical Guide for Producers

To analyze a mix, you do three things in a loop: listen critically with a fresh set of ears, measure what you heard with a couple of meters, and test your fixes level-matched against the original. Skip any one of them and you end up guessing. This guide walks through the whole loop, from the five things worth checking to the traps that make good mixes worse.

Aubiomix

aubiomix.com

Get Clearer Feedback on Your Mix

Upload your song to Aubiomix for detailed feedback and actionable steps on its mixing and mastering.

Visit Aubiomix

TL;DR

Judge a mix on five fundamentals: level balance, frequency content, dynamics, stereo image and space.
Meters tell you what is happening. Your ears decide whether it matters, so use both, and use at least two analysers.
Check translation on several playback systems, plus mono, before you call anything finished.
Make one change at a time, level-match every A/B, and stop when the problem is solved, not when the meter looks pretty.
Automated analysis is brilliant at spotting patterns and weak at the odd glitch or the creative call, so verify what it says by listening.

Start with what you are actually analysing
Reset your ears before you listen
Step 1: Analyse the frequency balance
Step 2: Check the dynamics and loudness
Step 3: Check the stereo image and mono compatibility
Step 4: Use reference tracks properly
Step 5: Test translation on real playback systems
Step 6: Turn findings into changes without overcorrecting
What automated analysis can and cannot do
Our perspective on mix analysis
Put it together: a 30-minute mix analysis routine
FAQ
Sources

Start with what you are actually analysing

A mix is a set of decisions about five things. Avid's mixing guide frames them as balance of levels, frequency content, dynamics, stereo imaging and space, and I find that list a handy checklist when a session starts to feel overwhelming. Everything you hear going wrong in a mix belongs to one of those buckets.

Balance: can you hear every element at the volume you intended, in every section of the song?
Frequency content: is the tonal balance pleasing, or is there a hole, a pile-up or a harsh edge?
Dynamics: how much does the volume move, and does the song still breathe?
Stereo image: panning and width give you left-to-right placement, and the centre stays solid.
Space: reverb and delay create depth. Too little sounds flat, too much sounds like a wash.

Write those five words on a sticky note. Every time you analyse a mix, you are just asking each question in turn.

Reset your ears before you listen

The biggest source of bad mix analysis is not missing tools. It is a tired, biased listener. Hearing is adaptive: time in a noisy environment can make it harder to perceive a full stereo image, and heavy bass-boost listening can dull your sense of the low end. If you have been inside the same session for four hours, your ears have quietly recalibrated to its flaws.

Soundfly's advice is to sleep on a mix and listen at different times or in new environments, because emotional reactions to your own work swing wildly. We have a deeper look at the numbers in what 6,000+ reports reveal about ear fatigue while mixing, and a practical routine in our guide to reviewing your mix objectively.

A simple reset routine:

Stop mixing and leave the room for at least ten minutes. Overnight is better.
Come back and play the song from the top at a modest volume, before touching anything.
Take notes on feelings first ("chorus feels small"), then translate them into technical suspects ("vocal too far back? low mids crowded?").

Step 1: Analyse the frequency balance

Frequency balance is where most amateur and semi-pro mixes go wrong, usually in one of three places: mud, harshness or a weak low end.

When you listen, ask whether the whole spectrum feels even, and whether instruments are fighting for the same space. The classic trouble zones are the low mids, where buildup makes a mix boxy and cloudy, and the upper mids, where harshness causes listening fatigue. Masking is the usual culprit: two sounds sharing a range so neither is clear.

A good workflow:

Loop the busiest section, usually the final chorus.
Solo nothing. Masking only shows up in context.
Use a spectrum analyser to confirm where the energy sits, then use your ears to decide whether it is a problem.
Cut before you boost. Fixing a clashing element by carving the competing one is often cleaner than lifting the quiet one.

We lay out a full routine in our 10-step frequency balance analysis for mud and harshness, and the broader idea of level and tonal relationships is in the producer's guide to mix balance.

Step 2: Check the dynamics and loudness

Dynamics tell you how hard a track has been squeezed. Mike Senior notes in Sound On Sound that peak and RMS readings often show how heavily dynamics have been squashed, and he recommends a reliable LUFS meter plus comparison with reference material.

Here is how to read it:

Big gap between peak and average level: lots of life, but possibly too quiet next to commercial releases.
Small gap: loud and dense, with a real risk that the drums and vocals feel flattened.
Loudness against the genre: different styles live at different levels. Our measured loudness and mix benchmarks show what acoustic material looks like, and there are separate pages for techno, deep house and post-punk.

Don't chase loudness in the mix stage. A mix with healthy headroom gives mastering room to work, and a mix that is already clamped leaves nothing to do but make it worse.

Step 3: Check the stereo image and mono compatibility

Stereo imaging is panning and width; the centre should hold your foundation. Kick, snare, bass and lead vocal normally sit there, and width comes from the supporting parts around them.

The test that catches the most problems is quick: fold the mix to mono. If the guitars thin out, the pad vanishes or the vocal gets hollow, you have phase trouble. Sound On Sound's checklist puts mono compatibility and phase correlation at the heart of this check, because many listeners still hear music on a single speaker or a phone.

If you want a measured approach, our guide to stereo imaging analysis, with three meters and a mono check shows how to read the tools, and what stereo width really is clears up a lot of myths about wide being better.

Step 4: Use reference tracks properly

A reference is a ruler. Pick two or three finished songs in your genre that you love, match their playback level to your mix, and flip between them. Don't just ask "which is better", ask specific questions: is my low end bigger or smaller, is my vocal further back, is my top end brighter?

Sound On Sound makes two useful cautions. Spectral analysers or EQ-matching tools can compare your tonality to a reference, but different keys cause measured differences in the low end, so a bass-heavy key in your track will show a bump that has nothing to do with your mix. And Sam Inglis warns against pushing tonal matching into extreme EQ moves that better recording or better reference choices would solve. For a measurement-first take on choosing a handful of reference tracks in mastering, that approach pairs well with what you are doing here.

Which meters to trust

No single meter is gospel. Different analysers can show different results for the same signal because of different algorithms and integration times, so the sensible habit is to use at least two and look for agreement. When two disagree, it is not an error, it is a reminder that these are views of the signal, not verdicts on the music.

A sensible minimum meter set:

Meter	What it tells you	What it cannot tell you
Spectrum analyser	Where energy sits across frequency	Whether that energy sounds good
LUFS and peak meter	Average loudness and peak level	Whether the dynamics feel musical
Correlation or phase meter	Mono risk and phase issues	Whether the width feels exciting
Reference A/B	How you compare with finished records	Whether your song needs to be like them

Step 5: Test translation on real playback systems

A mix that only works on your monitors is not finished. Understanding Audio suggests testing on a car stereo, iPhone speakers, a laptop, headphones, earbuds, a TV, and even from outside the room. Each one exposes something different: the car shows bass balance, the phone shows whether the vocal and the mids carry the song, and the hallway shows whether the balance survives when detail is stripped away.

Seven playback systems for checking mix translation

Headphones have their place too. They are excellent for catching clicks, bad edits and pops, even if they are not recommended for main balance decisions. If headphones are your main tool, read our piece on mixing on headphones so it translates.

While you are at it, scan for technical faults: digital clipping on loud notes, distortion, hum, DC offset, and polarity problems. Open-source tools such as Phantom check recordings for those issues before you start mixing, which saves you from building a mix on a flawed foundation.

Step 6: Turn findings into changes without overcorrecting

This is where analysis earns its keep, and where most people undo their own work. A list of twelve observations is not a to-do list for one afternoon.

A safe process:

Rank the findings. Fix the biggest problem first. Often fixing one issue, like low-mid mud, solves three smaller complaints.
Be specific. "Cut 3 dB around the boxy region of the guitars" beats "make it clearer".
Level-match every A/B. A louder version always sounds better. The testing advice is to match levels within about 0.5 dB using a trim plugin, or 0.2 dB if your meter is steady enough.
Change one thing at a time, and check at low and normal volume.
Stop when the issue is solved. A meter that looks like a reference curve is not a reason to keep going.

Here is a worked example. Your analysis says the chorus sounds small and the vocal is hard to hear. Before reaching for the vocal fader, check whether the guitars and synth pad crowd the same range as the voice. Carving 2 dB from the pad often does more than lifting the vocal by 2 dB, and it leaves headroom intact.

What automated analysis can and cannot do

Automated analysis is excellent at the objective layer: levels, spectral balance, stereo correlation, clipping. Research on automatic mixing is honest about the limits. Steinmetz and colleagues point out that mixing is a subjective, creative process that cannot be fully expressed through objective metrics, which is why perceptual listening tests are the standard for evaluating such systems. Research is also showing progress: a recent academic paper describes a critic trained to retrieve vocal-mix issues and rank their severity.

The gaps are real, though. A mastering engineer notes that automated tools do not listen for a click at a reverb tail, an export dropout, or a phase issue hidden by room modes. That is why we always say: treat any report, ours included, as a second opinion that you verify by listening.

Our perspective on mix analysis

We built AubioMix because most producers don't lack tools, they lack a trustworthy second pair of ears. Free analysers give you numbers. Numbers are useful, but they rarely tell you what to do on Monday morning.

Our approach combines audio analysis with the experience of our founder, a multi-award-winning mix engineer with 20+ years in the craft. Every report gives both subjective guidance, the "this is how it feels and why", and objective measurement, so you can see the evidence behind each suggestion. Our evaluation framework is regularly updated to reflect current genre standards, because a punk record and a techno record should not be judged against the same ruler.

What we believe, plainly:

A mix is never "correct", it is either serving the song or it isn't.
Feedback is only worth something if you can act on it today.
Progress beats perfection. Seeing your balance, dynamics and stereo scores improve across ten songs teaches more than any one perfect mix.

— AubioMix

Put it together: a 30-minute mix analysis routine

Rest (10 minutes or more). Leave the session and reset your ears.
First listen (5 minutes). Play through at moderate volume, notes only.
Meters (5 minutes). Spectrum, loudness, phase. Use two tools where you can.
Mono and reference (5 minutes). Fold to mono, then A/B with level-matched references.
Translation (5 minutes). Phone, earbuds, car or laptop.
Plan. Pick the top two problems and fix those, then repeat the loop.

If you want the same process with expert-level feedback and no guesswork, you can upload your song to AubioMix. A single Mix Review costs £4.99, a Reference Pass is £2.99, and if you mix regularly the Professional plan is £19.99 per month and Professional + is £34.99 per month. You get visual and written reports across 15+ areas, from compression and EQ to phase, loudness and device translation, with progress tracking and downloadable PDFs. You can see what a report looks like on a real record in our analysis of a Fred again.. track.

FAQ

How do I check if my mix is good?

Start with a fresh listen, then check balance, tonal balance, dynamics, stereo image and space. Test in mono and on at least two other playback systems, and compare with a level-matched reference. If it holds up everywhere, it is in good shape.

Can AI analyse my mix reliably?

It is reliable for measurable things such as loudness, clipping, spectral balance and stereo correlation. It is less reliable for creative intent and odd glitches, which is why you should verify its suggestions by listening and with level-matched A/B tests.

How many meters do I need?

A spectrum analyser, a loudness meter and a phase or correlation meter cover the essentials. Use at least two analysers where possible, since different algorithms can show different readings for the same signal.

Should I fix everything a mix analysis finds?

No. Rank the findings and tackle the biggest one first, because it often resolves smaller ones. Make small changes, level-match your comparisons, and stop as soon as the problem is solved.