Skip to main content

Worked example

Luma Audio: one catalog, three strategies.

A fictional audio discovery platform with a 54-item catalog, five audience segments and a synthetic behavioural log. Every figure on this page was computed by the same engine that runs inside the product, at seed 424242.

Everything here is fictional

Luma Audio does not exist. Every title, creator, audience segment and behavioural event was created for TasteLab so the product can be evaluated without uploading anything.

Simulation

Discovery Balanced · Discovery Explorer

Discovery Explorer

Taste constellation for the Discovery Balanced strategy, audience Discovery Explorer. Relevance proxy 71%, novelty 73%, diversity 91%, catalog coverage 72%, popularity concentration 0.49. Exposure reached 12 categories. Top recommendations: 1. Estuary Bells by Solveig Rand; 2. Small Hours Percussion by Batid Collective; 3. Interference Pattern by Ilya Bregman.

Recommendation stream

  1. 1Estuary BellsSolveig Rand · Field RecordingAffinity63
  2. 2Small Hours PercussionBatid Collective · PercussionAffinity70
  3. 3Interference PatternIlya Bregman · ModularAffinity75
  4. 4Uphill Both WaysThe Halcyon Unit · Jazz FusionAffinity78
  5. 5Cold OpenPriya Halloran · Spoken WordAffinity50
  6. 6Two RiversAdaeze Marchetti · OrchestralAffinity58
  7. 7Second Person PluralChoir of Small Hours · ChoralAffinity53
  8. 8Housing CooperativeNils Aberdeen · Lo-fiAffinity54
  9. 9Repeat CustomerKestrel Ono · Synth PopAffinity58
  10. 10Held Note for a Bright RoomVera Ostlund · AmbientAffinity49

Everything we measured

Good matches

71%+21

Variety

91%+12

Beyond the hits

73%+61

Catalog reached

72%+50

Crowding

0.49-36

Repeats

0%-60

Happy surprises

63%+58

Brand-new users

53%+21

Creators reached

100%+50

Categories reached

100%+50

These are real results from a made-up catalog that ships with TasteLab, so you can try everything without uploading anything. They show how a strategy behaves, not what your actual users will do.

The five audience segments

Five made-up types of listener, covering the cases recommendations usually get wrong: the explorer, the creature of habit, the trend follower, the person who digs deep, and someone who just signed up and you know nothing about.

  • Discovery Explorer

    Actively seeks novelty, moves between genres in a single session and responds well to emerging creators. Punishes repetition faster than any other segment here.

    Likes the unfamiliar
    88
    Follows the crowd
    22
    Open to new things
    85
  • Comfort Listener

    Returns to familiar moods and expects consistency between sessions. Tolerates repetition well and reacts poorly to sharp genre changes.

    Likes the unfamiliar
    20
    Follows the crowd
    55
    Open to new things
    18
  • Trend Follower

    Strongly influenced by what is popular and recently released. The segment where a popularity baseline looks deceptively good.

    Likes the unfamiliar
    35
    Follows the crowd
    90
    Open to new things
    32
  • Deep Catalog Fan

    Prefers less obvious items and long exploratory sessions. The clearest signal of whether a strategy can surface anything below the popular head of the catalog.

    Likes the unfamiliar
    78
    Follows the crowd
    15
    Open to new things
    62
  • New User

    Cold start

    Almost no behavioural history and a single stated preference. Included so every strategy is measured against the case it usually handles worst.

    Likes the unfamiliar
    50
    Follows the crowd
    60
    Open to new things
    50

The three strategies, side by side

MetricPopularity FirstHybridDiscovery Balanced
Good matches49%65%71%Best
Variety79%92%Best91%
Beyond the hits12%47%73%Best
Catalog reached22%74%Best72%
Crowding0.850.510.49Best
Repeats60%0%Best0%Best
Happy surprises4%35%63%Best
Brand-new users32%47%53%Best
Creators reached50%100%Best100%Best
Categories reached50%100%Best100%Best

These are real results from a made-up catalog that ships with TasteLab, so you can try everything without uploading anything. They show how a strategy behaves, not what your actual users will do.

Start here

Prototype the experience
before building the infrastructure.

Explore how different recommendation strategies shape discovery, relevance and catalog exposure using a reproducible simulation.

These results come from a simulation, not from real people. They show how a strategy behaves on this catalog, which is useful for comparing options — but they are not a prediction of what your actual users will do. Try anything promising with real users before you rely on it.