Guide · the method behind a fair A/B
Level matching: why the louder version always wins
When you compare two pieces of audio, the slightly louder one almost always sounds "better" — fuller, clearer, more exciting — even when it's otherwise identical. Level matching removes that bias by setting every version to the same loudness before you listen. Without it, you're judging volume, not sound.
The loudness trap
Our hearing equates "louder" with "better." A touch more level adds perceived bass and presence and a sense of impact, so in any uncontrolled A/B the louder file wins — regardless of which mic, plugin, mix, or converter actually sounds best. This is exactly why mastering engineers, codec researchers, and gear reviewers level-match before trusting their ears, and why "it sounded better after I turned it up" is a warning sign, not a verdict.
What "level matching" actually means
Level matching means setting every version you compare to the same loudness — not the same peak — so volume can't sway your judgment. Two files can hit the same peak level and still feel very different in loudness, so peak-normalizing isn't enough. Instead, measure each version's loudness and bring them into line:
- RMS — average energy over time; a simple, solid loudness proxy.
- LUFS (EBU R128) — perceptually weighted loudness, closer to how loud something actually feels to a listener.
For an A/B, either beats not matching at all — what matters is that you match loudness and stay consistent. The safest approach is to match everything down to the quietest version: because you're only ever attenuating, the signal can never clip.
How to level-match your comparison
- Measure loudness (RMS or LUFS) for each version — not the peak.
- Match to the quietest. Turn the louder ones down to meet it; never turn up, so nothing clips.
- Re-match after any change. Process a version and its loudness shifts — measure and match again before you trust the comparison.
- Compare blind, so expectation bias can't pile on top of loudness bias.
Common mistakes
- Matching peaks instead of loudness — they're not the same thing.
- Comparing at different monitor volumes — that reintroduces the bias you just removed.
- Forgetting to re-match after EQ, compression, or mastering changed the level.
- Letting the louder option "win" and calling it the better one.
How PickFav handles it
PickFav does this for you: it matches every variant by RMS to the quietest one — attenuation only, so the output can't clip — and has you compare blind, with labels hidden and reshuffled each pass. You judge the sound; the volume and the names are out of the way.
FAQ
RMS or LUFS — which should I match?
Either is far better than not matching. RMS is a simple average-energy measure; LUFS (EBU R128) is perceptually weighted and closer to how loud something feels. For an A/B, the key is matching loudness, not peaks, and being consistent. PickFav matches by RMS.
How big a level difference actually biases a comparison?
Small ones. Differences well under a decibel are enough to make the louder version sound "better," which is why even careful A/Bs are unreliable without level-matching.
Does PickFav level-match automatically?
Yes. It matches every variant by RMS to the quietest one (attenuation only, so it can never clip), and you compare blind. You can toggle it off, but on is the honest default.