Model notes
Methodology before output.
Catalog Underwriter starts with observable evidence, identifies what is missing, and exposes every material valuation assumption. The financial engine is deterministic.
- Methodology
- public-underwrite-1.2.0
- Engine
- dcf-1.0.0
§1
What the model collects
MusicBrainz provides catalog evidence: canonical artist identity, release groups, dates, types, and genre metadata. Last.fm supplies cumulative scrobbles and listener counts as demand evidence. When configured, YouTube supplies cumulative views for conservatively matched official music videos, official audio, and artist-topic uploads.
§2
What remains unavailable
Royalty statements, private ownership contracts, platform-reported annual consumption, recoupment positions, negotiated distribution terms, and buyer bids are generally not public. The application labels those fields as unknown, modeled, or user assumptions; it never treats missing information as zero, and every annualization of a cumulative observation discloses its calibration.
§3
Catalog normalization
The economic model distinguishes compositions, recordings, and commercial releases. The current live provider returns release groups and removes identical identifiers. The stored fixture applies a deeper development normalization; recording-level live normalization is intentionally marked incomplete.
§4
Observed consumption
Last.fm observations support relative artist scale, track concentration, breadth, long-tail strength, and persistence. YouTube view counts add another cumulative demand signal. Both are labeled cumulative observations; on their own neither is an annual stream count, royalty statement, or cash-flow estimate.
§5
Modeled consumption
Annual audio streams are modeled from Last.fm listeners summed across an artist’s top 25 tracks. A calibration on 29 artists, measured against each artist’s whole-catalog daily Spotify streams, puts the 10th / 50th / 90th percentiles at 65 / 150 / 445 annual Spotify streams per top-track listener, a range that holds 23 of the 29 artists; Spotify’s reported 65% share of global audio streams grosses that up to all platforms. Individual artists can miss by up to 5×, because Last.fm under-represents mainstream pop, hip-hop, and Latin audiences and over-represents indie and rock. Annual YouTube views take each matched official upload’s lifetime-average views per year (over at least one year) and scale them by the trailing-year ratio measured on 93 official videos: 0.35 / 0.75 / 2.05 (10th / 50th / 90th percentiles). Both ranges are labeled modeled annualizations, and the YouTube range is a floor because unmatched uploads are not counted.
§6
Economic assumptions and cash flow
RightsEconomics bridges annual consumption to annual rights revenue, modeling audio streaming, YouTube, publishing, and other revenue separately. Defaults come from public payout reporting: $2.60 / $3.20 / $3.80 recorded revenue per 1,000 audio streams, $0.45 / $0.80 / $1.50 per 1,000 YouTube views, publishing at 24–31% of recorded revenue, and other revenue at 15–43%. Every value is editable; blanking one turns the estimate off. The resulting modeled cash flow pre-fills the valuation and is labeled as modeled; switching to Known cash flow replaces it with a verified or hypothetical figure.
§7
Valuation
The deterministic DCF projects annual cash flow for ten explicit years and discounts each year at the selected risk rate. Defaults follow how the catalog market prices income: the base case uses Citrin Cooperman’s 8.5% discount rate with +5% growth for five years and +2% after, about 17× annual cash flow, in line with 2024 multiples for older catalogs. The bear case uses Shot Tower’s 9.63% re-valuation rate (about 10.6×) and the bull case the 2021–22 peak (about 23.6×). A separate market approach multiplies the rights-weighted annual cash flow by an editable multiple whose starting range is derived from curated transactions with disclosed income multiples.
§8
Reconciliation
The interface shows DCF value, market-multiple value, and a reconciled indicative range. The range spans the selected DCF result and the transaction-derived market band; it is not a mechanical average. Rights scope and evidence quality still require underwriting judgment.
§9
Comparable transactions
The repository-managed dataset links to transaction reports and labels rights scope, confidence, comparability, and disclosed income-multiple basis. Benchmarks are validation data, never targets for fitting Last.fm to sale price. Missing prices and multiples remain undisclosed.
§10
What the model cannot know
The evidence chain is kept explicit: observed consumption → modeled consumption → economic assumption → estimated cash flow → valuation. This tool cannot determine actual artist income, legal ownership, contractual restrictions, future cultural relevance, or a definitive market-clearing bid. It supports judgment; it does not replace diligence, legal advice, or verified financial statements.
Catalog Underwriter is an educational analytical tool. Estimates are not verified royalty statements, ownership interests, market offers, investment advice, or legal advice.