How We Rate Music: Five Dimensions, One Public Rubric
By Dan Cohen · Published 2026-04-04
Short version: every song on this site is scored on the same five dimensions — dynamic range, sudden changes, texture, predictability, and vocal style — against a public rubric you can read in full on our methodology page.
The ratings are model-assisted. They are derived from each song’s documented musical properties — its arrangement, instrumentation, production style, vocal delivery, and published tempo data — combined with human review. We do not run audio-signal analysis on recordings, and the methodology page says so plainly. Where a dimension can’t be assessed honestly, we leave it blank rather than guess.
Two design choices matter more than anything else in the system:
The rating is conservative on purpose. The top-line Safe / Moderate / Intense level is AND-gated on the Safe side and OR-gated on the Intense side: a song must clear every bar to be called Safe, while tripping any single bar makes it Intense. A song that could blindside you should never be labeled Safe; a gentle song mislabeled Intense costs you a candidate, not your calm.
The ratings are falsifiable. Every rating is published, per song, next to what it claims. If a song we call steady lurches on you, tell us — ratings get revised, and the methodology page keeps a standing list of what we will not fabricate: musical key, chord progressions, per-second structure, lyrics. If we can’t stand behind a number, we don’t ship it.
What the five dimensions mean, how the top-line level is derived, worked examples on real songs, and the honest limits of the whole approach — all of it lives on the methodology page. This page exists so the answer to “how do you rate music?” is one sentence: five dimensions, one public rubric, model-assisted with human review, conservative by design.