Single-artist versus multi-artist recommendations

A single seed asks for one neighborhood. Multiple seeds ask where several neighborhoods overlap. We ran both methods with the same taste profile to show what changes.

Author and reviewer: TrackMoose operator
Published: July 14, 2026
Last reviewed: July 14, 2026

The test

The profile contains Radiohead, Bon Iver, and Phoebe Bridgers. The first run uses Radiohead alone. The second uses all three. Both were calculated from TrackMoose similarity lists on July 14, 2026, using the same scoring code documented in the methodology.

This is not a claim that one mode is universally better. It is a controlled comparison of how adding two real preferences changes the question.

Run A: one seed

Seeds: Radiohead

Rank	Artist	Score	Tag	Seed support
01	Thom Yorke	100%	electronic	1 of 1
02	Atoms for Peace	54.1%	electronic	1 of 1
03	Jeff Buckley	35.3%	singer-songwriter	1 of 1
04	Muse	33.3%	alternative rock	1 of 1
05	The Strokes	28.2%	indie rock	1 of 1

Captured TrackMoose output from July 14, 2026. Scores can change when cached similarity data refreshes.

Run B: the full three-artist profile

Seeds: Radiohead, Bon Iver, Phoebe Bridgers

Rank	Artist	Score	Tag	Seed support
01	Thom Yorke	50%	electronic	1 of 3
02	Volcano Choir	50%	indie	1 of 3
03	boygenius	46.9%	indie rock	2 of 3
04	Justin Vernon	41.5%	folk	1 of 3
05	Julien Baker	32.3%	indie	2 of 3
06	Lucy Dacus	32%	indie	1 of 3
07	Better Oblivion Community Center	31.6%	No tag	2 of 3
08	Atoms for Peace	27%	electronic	1 of 3

Captured TrackMoose output from July 14, 2026. Scores can change when cached similarity data refreshes.

What changed, and why

The single-seed run stays close

Thom Yorke and Atoms for Peace dominate because they have direct, strong relationships to Radiohead. The rest of the list moves through familiar alternative-rock neighbors. Every result is supported by the same one seed, so raw similarity largely determines order.

The multi-seed run becomes a blend

Volcano Choir, boygenius, Justin Vernon, Julien Baker, Lucy Dacus, and Adrianne Lenker enter the top ten. Candidates supported by two seeds gain coverage, while strong one-seed matches remain possible at a lower normalized score.

When to use each

Use one seed for focused exploration

Choose one artist when the question is specific: more artists near this sound, era, scene, or project. It is fast and easy to interpret, but it inherits that seed's blind spots.

Use multiple seeds for personal taste

Choose several artists when no single name explains the goal. Mix seeds you genuinely want represented. Shared candidates rise, but distinct branches can still survive if their individual signal is strong.

Limits and failure cases

A single obscure seed can return little or nothing when its similarity data is sparse.
Adding weakly related seeds can split the result into separate branches instead of creating useful overlap.
Side projects and close collaborators can dominate a focused run. Exclude known artists when they add no discovery value.
Captured rankings can change after the 90-day cache refresh. Treat the result as a dated run, not a permanent chart.

Put the guide into practice

The recommendation method is documented separately and kept aligned with the live scoring code.

Read the methodology

Compare your own seeds