tl;dr: IEMs are rated on Timbre and Tonality, Bass and Dynamic, Resolution, Soundstage and Imaging. The four ratings come from A/B testing against gate IEMs at each performance band. A different number tells you where an IEM sits in the established hierarchy, not whether I personally like it. For how I think about IEMs and reviews in general, see philosophy.
Last update: 1 July 2026
I read and watch reviews to know whether a shiny new IEM is worth picking up alongside what I already own. Most reviews are bad at telling me this, because the “ruler” they use moves with the reviewer’s taste and experience. An “average IEM” to a daily-driver of flagship IEMs is not the same product as an “exceptional IEM” to a reviewer who lives in the entry-level tier.
There is also an over-focus on tonality that crowds out the things I find more interesting in this hobby: stereo imaging, dynamic response, separation, detail resolution, the way an IEM holds up when the music gets busy.
In-Ear Gems runs a review method called Subjective Benchmark-based Rating (SBR), designed to make the ruler stable and to weight the things I actually grade on. The method depends on three things: gates (a fixed set of reference IEMs at known performance boundaries), dimensions (the four rating axes), and a measurement rig (my own, used to keep my impressions honest).
What SBR is
Benchmark-based means an IEM’s score reflects how it compares to a known, well-performing IEM. The system can tell you that an IEM resolves better than the Moondrop Aria but worse than the Moondrop Blessing 2, instead of the useless “good detail for the price.”
Subjective means the comparison is done by ear and brain rather than a measurement rig. This is necessary because a lot of what makes an IEM interesting (soundstage depth, the way a mix stays composed under load, the texture of decays) does not live in a single microphone capture.
This does not mean subjective equals preference. The reference is other IEMs, not what I like. I treat personal preference as mostly irrelevant to your purchase. The stance is spelled out on the philosophy page.
Frequency response is part of this method, not separate from it. I run my own measurement rig and use the graph to keep my impressions honest. When the measurement and my impression disagree, I get curious, because one of the two is wrong, or I am missing something.
The four dimensions
Each IEM is rated on four dimensions. They correspond to how I think about the listening experience, the things I find interesting in this hobby.
Scores run 1 to 5 in 0.5 increments. Half-steps exist because ties happen and they are useful to record.
Timbre and Tonality
How the IEM colours the music: the relative loudness and timbre of different elements in a mix. Tone shapes what an IEM emphasises. I keep the rating conservative here, because tonality is largely a matter of preference, and a well-judged warmer or brighter presentation is fine if it is delivered without sacrificing the precision of the other dimensions.
- 5/5 (outstanding): Improved upon 4/5. No EQ is necessary.
- 4/5 (great): Tuning that does not sacrifice lower-mid for artificial separation. All frequency ranges are present. Lower frequencies may be too much on some tracks.
- 3/5 (good): Generally balanced and enjoyable. No timbre issues, though there may be some tonal imbalance, such as cutting the lower-mid to create artificial separation.
- 2/5 (not enjoyable): Unbalanced sound such as too much bass and too little treble (dark). There may be timbre issues. EQ is necessary.
- 1/5 (train wreck): Tuning causes pain or discomfort.
Test tracks for this dimension are on the test tracks page.
Bass and Dynamic
The quality of bass and drum hits, and the way the IEM reproduces transients and dynamic swings. It is about how the bass and percussion behave under the music, not just how loud they are. A 4/5 IEM with less bass extension than a 5/5 IEM will still rate lower on this dimension, even though its bass control might be just as good. The dynamic part of this dimension is about the IEM’s ability to convey loudness swings: a quiet passage to a crescendo, a held breath to an impact.
- 5/5 (outstanding): Same control as 4/5 with more bass quantity and subbass extension. You can feel the bass hit in your throat.
- 4/5 (great): Fast and well-controlled bass that keeps up with complicated basslines. Quantity may be lacking, either in midbass or subbass extension.
- 3/5 (good): Decent bassline with a physical sensation. May be slightly mushy or struggle with complicated basslines.
- 2/5 (not enjoyable): Boomy or mushy bass that does not have a strong, clean impact.
- 1/5 (train wreck): Broken headphones or IEMs with zero bass.
Test tracks for this dimension are on the test tracks page.
Resolution
The IEM’s ability to reveal fine detail and keep elements of a mix from mushing into one another. Three things feed into this:
- The sharpness of note attacks. Does the IEM have the speed to start a transient on time?
- The separation of overlapping instruments and vocals, especially when the mix gets busy.
- The texture in the decay side of notes: reverb tails, breath, things you only hear on more resolving IEMs.
- 5/5 (outstanding): More resolving than 4/5. Pin-point separation and layering.
- 4/5 (great): Excellent detail and separation, with strong texture in voice and instruments.
- 3/5 (good): Adequate detail and separation, with some smearing remaining in busy sections.
- 2/5 (not enjoyable): Some larger details are presented, but micro-detail and nuances are lacking, and elements overlap.
- 1/5 (train wreck): Mushy mess with little detail.
Test tracks for this dimension are on the test tracks page.
Soundstage and Imaging
Stereo imaging is the IEM’s ability to project sound elements in space, left-to-right and near-to-far. Soundstage is the resulting illusion of the field around the listener. Exceptional IEMs can paint a layered, three-dimensional picture, where instruments form distinct objects in depth rather than blending on a flat plane.
- 5/5 (outstanding): Wide and deep soundstage with a complex, layered image. Some reviewers call this “holographic.”
- 4/5 (great): Exceptional left-to-right panning or above-average depth and layering, but not both at the same time.
- 3/5 (good): Wide soundstage, but depth and layering lacking.
- 2/5 (not enjoyable): Small soundstage with little to no depth.
- 1/5 (train wreck): Mono sound. No soundstage.
Test tracks for this dimension are on the test tracks page.
Gate IEMs
The four dimensions alone give you four scores, but they do not place the IEM relative to the wider market. To do that, I use gate IEMs: reference units at the entry boundary of each performance band.
The current gates:
- AFUL Cantor. Entry gate of the 5/5 band. Replaced the 64 Audio U12T, which held this gate before Cantor.
- CFA Andromeda 2020. Entry gate of the 4/5 band.
- Moondrop Blessing 2. Entry gate of the high 3 band.
- Moondrop Aria. Entry gate of the low 3 band.
When a new IEM comes in for review, I A/B it against the gate IEM closest to where I think it might land. A 4/5 here does not mean “as good as the Andromeda 2020.” It means “in the same performance band as the gate, on the dimensions I am rating.” Gates have largely held their positions over the past few years, with Cantor being the one explicit replacement I have made so far.
I chose the gates through a round of A/B “championship” testing a few years back, pitting reference IEMs against each other to find natural band boundaries. I do not change gates lightly, because they are what let ratings written years apart stay comparable.
How ratings appear in a review
Each review’s conclusion block carries four pieces:
- Per-dimension scores. One score for each of the four dimensions, in 0.5 increments.
- Absolute Sonic Quality Rating. The average of the four dimension scores, rounded to one decimal place. This is the headline numeric rating on the review.
- Bias Score. A 1 to 5 number capturing how the IEM actually lands for me, independent of its technical merit. It is allowed to disagree with the technical rating, because bias can.
- What I like / What could be improved. Two short bullet lists that summarise the trade-offs in plain English.
Bias Score scale:
- 5/5. I love this IEM.
- 4/5. I like this IEM.
- 3/5. I am lukewarm about this IEM.
- 2/5. I do not like this IEM.
- 1/5. I dislike this IEM.
A worked example from a past review:
- Timbre and Tonality: 4/5 (great)
- Bass and Dynamic: 4/5 (great)
- Resolution: 5/5 (outstanding)
- Soundstage and Imaging: 5/5 (outstanding)
- Absolute Sonic Quality Rating: 4.5/5
- Bias Score: 3/5 (I am lukewarm about this IEM)
In this example, the engineering is at the 4/5 band but my personal reaction is muted, which is exactly the kind of split the Bias Score is meant to surface.
Review setup
The rig changes as the test budget allows. The current setup:
- DAP: HiBy R6 Pro Max, used for portable listening tests.
- USB DAC: L&P W4, my daily driver for desk listening.
- Desktop DAC: HiBy FD5, for the A/B chain in long review sessions.
- Source content: Lossless FLAC files, mostly at red book standard.
- Measurement rig: IEC-711-compliant coupler microphone. Every graph in my graph database is my own measurement.
What matters for ratings is that the comparison unit and the test IEM are heard through the same chain, with switching fast enough that aural memory does not fade.
Review procedure
Every review IEM goes through daily driving for at least one day first, to get the brain tuned to the sound and surface any comfort issue. The actual review then runs as multiple A/B tests, working up from the lowest benchmark (Apple EarPods) through the gates.
Sometimes you will see impressions on the blog instead of full reviews. Impressions are carried out similarly but on a shorter time budget, generally for loaned or tour units that I do not get to keep.
FAQs
What is the Absolute Sonic Quality Rating? The average of the four dimension scores, rounded to one decimal place. It is the headline number on each review, but it is just a summary. The dimension scores carry the more useful information.
What is the Bias Score? My own subjective reaction on a 1 to 5 scale. It is allowed to disagree with the technical numbers, because actual enjoyment is not a technical measurement. A high Bias Score with a low Absolute is “I want to like this but the engineering is not there.” A low Bias Score with a high Absolute is “the engineering is there but I personally do not enjoy it.” Both happen.
Why have the gate IEMs changed? I established the bands through a round of A/B championship testing. The bands have largely held, but I updated the 5/5 gate from the 64 Audio U12T to the AFUL Cantor when Cantor established itself as the new entry to that band in my testing. I do not change gates often because doing so would make ratings across reviews harder to compare.
Is detail the same as separation? On this site, no. “Resolution” combines both into one dimension because they share the same underlying mechanisms: the speed of transients, the cleanliness of decays, the precision of timbre. An IEM can resolve well but separate poorly (the older multi-BA designs), or separate well but be smooth on detail (some warm-tuned flagships). Both feed into Resolution.
Is detail just treble? No. Treble energy pushes the perceived presence of detail, but an IEM cannot make detail that is not in the recording, or that its driver cannot physically resolve. The KZ ZSN Pro has a lot of treble energy yet resolves very little.
Is separation a function of frequency response? Largely yes, but not only. Dynamic range, distortion under load, and imaging cues from the recording itself also contribute.
How did you establish the benchmarks? Through a round of A/B testing a few years back where I pitted reference IEMs against each other to find natural band boundaries. The bands have held since, with one explicit gate swap to date.
What about measurements, are you ignoring them? No. I run my own measurement rig using an IEC-711-compliant coupler microphone, and the graph database is entirely my own work. Frequency response encodes most of what I grade on. When the measurement and my impression disagree, I get curious about which is wrong.
Why do you have headphones in the database? The four dimensions apply to headphones too, so I include some headphone ratings for perspective. Headphone reviews are not a focus of this site.
How does a 5/5 IEM sound? It would have energetic tuning that still keeps the midrange clear. All frequencies would be well presented. It would place elements of a mix in a small dome around the listener’s head with pin-point precision. The bass would hit hard and fast with plenty of rumble, yet would not lose composure when the music gets crowded.