Research
Studies, with the method attached.
Every study on this page ships with its sample size, its method, its limitations and its raw data. If a study is still running, it says so and it is not linked as finished. Nothing here is behind a wall.
The most lexically complex rappers, measured
Token-level and rhyme-density analysis across the studio catalogues of the most-streamed rappers of the decade, built on MusicBrainz metadata. Lyrics are computed on and never republished, so only derived metrics are released.
Investigations
Published, and sourced to the documents
An investigation is not on a cadence. It runs when the primary documents say something the published consensus does not, and every quotation below was read from the named organisation's own filing, terms or documentation rather than from coverage of it.
Six systems, one genre, and a factor of 366
Every Noise at Once recognises 366 rap genres. Spotify's own public API recognised one. Six maintained classification systems disagree by a factor of 366, the famous "230 kinds of hip hop" turns out to be a literal string match that excludes trap and drill, and the shape of the disagreement is a better map of hip-hop's subcultures than any list of scenes.
Quiet hands and 1,967 ad-libs: two cultures that decided repetition was good
One culture held children's hands down until they stopped moving them. The other counted 1,967 ad-libs on a single album and built a website visualising the noises. Autistic writers quoted in their own words, the Jamaican lineage credited where it belongs, and no claim made about anybody's neurology.
The triple entendre has never had a test. Here is one, applied.
Rap criticism counts layers of meaning like a score and has never published a rule a candidate could fail. Five rules, applied to the canonical claims, with the failures printed, including the Jay-Z line that started the whole format. Plus the check nobody performs: what MF DOOM said about his own method, which was that he was counting sounds.
Credited and unpaid: what rap's ghostwriting scandal was actually about
Quentin Miller was printed in the liner notes of four Drake songs at release, months before anyone accused anyone of anything. Rihanna is not a credited writer on "Umbrella". A ledger of who is actually credited on what, with the tier of evidence printed on every row, and the reason one genre treats a credited co-writer as a moral failure while the genre next door treats it as Tuesday.
The conversion table: what a Hot 100 entry is actually worth
Rap's chart presence stopped converting into hits. Pooled top-10 conversion falls from 18.1% to 7.4% across debut cohorts holding almost identical entry totals, and Drake is the only artist above 200 career entries who still converts better than 20%. The full table for 27 artists, with the chart mechanics that moved the denominator.
403 and zero: rap set its chart record the year it vanished from the top 40
Drake has 403 career Hot 100 entries and nobody else has ever reached 300. In October 2025 the Hot 100 top 40 contained no rap songs at all, the first shutout since 1990. Hip-hop's top 25 by chart entries, read from Billboard's own per-artist pages, with a 19-entry hole below the top 20 and every source disagreement printed rather than smoothed away.
Sampling and beats without getting burned: the legal routes nobody explains
The clearance system was built for people with lawyers. What actually happens at a small scale, and the free and cheap legal routes that exist: free-for-profit beats, what a non-profit licence really permits, replay, and what the royalty-free grants actually say.
How to get your master recordings back: the clock is not always 35 years
The statute contains two termination clocks, not one, and for a record deal the second can open the window years earlier than the 35 figure everyone quotes. Read from 17 U.S.C. §203 and §304 directly, including the recordation step that voids the whole thing if it is skipped.
What you actually bought: nine beat and sample licences, read from the terms
BeatStars' unlimited lease expires at ten years. Content ID swapped sides in 2026. Royalty-free is usually in the marketing and not in the grant. Nine licences read from the platforms' own terms, with every unanswered cell marked UNKNOWN rather than inferred.
Spotify says there is no per-stream rate. Every distributor publishes one anyway.
Spotify's own documentation denies that a per-stream rate exists. Distributors publish $0.003 to $0.005. Both are sourced here, along with the 2024 threshold that makes every published figure inapplicable to most of the artists reading it.
There is no per-stream rate
Published figures for the same platform disagree by up to 21x. That is not mostly dishonesty: payouts are pool-divided, so the rate is an output of a division rather than a price. Open CC-BY dataset of published figures, method and limits stated.
Where credits actually live
A song's credits exist in six registers that were never designed to agree. The ones carrying financial force contain no production credits; the ones people actually read carry no financial force. Open CC-BY dataset comparing all six.
Who answers the money questions in music
We audited who ranks for the highest-volume money-in-music questions and classified each publisher by what they sell. On "how do artists make money on Spotify", the top two results are both print-on-demand merchandise companies. Open CC-BY dataset, method and limits stated.
In the pipeline
Scoped, not published
These are designed and queued. They appear here so you can see what is coming and hold us to it, not to pad a research index. None of them has a number attached yet, because none of them has been run.
What a stream actually pays, by platform
Per-stream payouts reconciled against real distributor statements rather than headline rates, with the variance shown instead of averaged away.
The opener's cut
What support acts are actually paid across tour tiers, who absorbs production and travel, and what a support slot costs an artist net.
Who owns the masters
Ownership chains for catalogues that changed hands, traced through filings rather than press releases.
The deal-term decoder
What the recurring clauses in a recording contract actually do to the money, read from contract language rather than from summaries of it. Annual cadence, because the templates move slowly and the changes are the story.
Chart mechanics, decoded
How bundling, variants and playlist placement translate into chart position, and what each is worth in units.
Standing offer
Sitting on data nobody has published?
Statements, contracts, settlement figures, tour settlements. We will treat a source properly, we will state what we can and cannot verify, and we will never publish something we cannot stand behind.
How we handle sources →