Set up a fair listening test

Audivane makes the audio pair fair: same source, alignment, gain matching, verified delivery. The other half of fairness is the physical setup, which no app can control. These checks keep your result about hearing rather than about circumstance.

Fix the output path

Pick one output device and keep it for the whole test. Route changes are deliberately blocked once the plan locks, but plan around them:

• Disable OS audio “enhancements”, equalizers and loudness normalization. These process everything you hear and sit inside the result.
• Wired output removes one variable. Bluetooth adds an OS and codec path that your test cannot see: Audivane’s checks cover digital routing and playback behavior, not what your adapter and driver do downstream. If you are testing over Bluetooth, treat the whole chain — encode and all — as the thing under test, and repeat after pairing changes or driver updates.
• Keep volume fixed between sessions. The app gain-matches the pair, so neither side is louder; your overall level should still be comfortable and the same across runs.

Choose material that suits the question

The app cannot tell you whether a track can reveal the difference you’re looking for.

• Use several familiar tracks, not one. A single file — especially one chosen because it allegedly “always exposes” a codec — turns the test into a drill for one artifact and inflates confidence.
• Dense, full-range material with vocals and transients gives more to compare than silence-heavy recordings.
• Remember your source’s ceiling: if the file is already an MP3, both chains degrade the same lossy source. Comparing formats means starting from a clean source for both sides of the question, not from the format you suspect.

Control yourself

• Decide the trial count before you start, when you have no opinion about the outcome. Stopping when it “feels conclusive” is the easiest way to fool a fixed-trial test — the results guide covers why.
• Test when you are rested. Fatigue, colds and a noisy room are real limits on hearing; a failed run in bad conditions is a scheduling fact, not a finding.
• Listen alone. Other listeners’ expectations spread, and an ABX result is about one pair of ears.
• Expect nothing. After the plan locks, the labels tell you nothing about which side is which chain, and re-reading this page mid-run is fair game — the mapping stays hidden until you answer.

What fairness does not mean

A fair setup and a passing preparation do not certify audibility or codec conformance. They mean the comparison was constructed honestly, on your machine, under the conditions you chose. Hardware limits, hearing limits and luck remain part of every number, which is exactly why the statistics are reported the way they are.