---
title: "Method statement: how a catalogue rating is reached"
description: "The plain-English method behind a TrackForge catalogue rating, rule book v2.4: what is surveyed, the five components, how works are weighted, how the letter is read and capped, estimated inputs, money, sources, what is not checked, and what each letter means."
image: "https://trackforge.studio/docs/img/trackforge-social-card.png"
---

> Documentation Index
> Fetch the complete documentation index at: https://trackforge.studio/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# Method statement: how a catalogue rating is reached

:::caution[Draft rule book, not yet calibrated]
This statement describes rule book **v2.4**, which is a **draft**. Its numbers are starting values, calibrated by running the method on real catalogues and adjusting it until the results are right; they can change before the rule book is published. No rating has been issued under it. When it is published, the rule book's hash is timestamped on Bitcoin before the first rating under it, and this page names that hash.
:::

This is the method a TrackForge catalogue rating follows, in words. The rule book is the same method written as rules and numbers; this page and the rule book say the same thing, and where a term below has a plain name in the rule book, this page uses it. Terms in **bold** are defined in the [glossary](/docs/certification/rating/glossary).

## What a rating says

A rating describes the **registration condition** of a music catalogue on a stated date: for each work the owner says is theirs, whether the registrations that make money flow are in place at each society that collects for it, from the evidence TrackForge read.

Every letter carries one fixed line:

> A letter describes the registration condition of the catalogue on its date. It is not a valuation, a credit rating or a legal opinion on title.

A rating proves what the evidence showed under these rules on that date. It does not prove the evidence is true; a reader checks that by looking the largest works up at the societies, and the Rating Report lists where.

## Levels

| Level | Name | Letter | What goes in |
|---|---|---|---|
| Tier 0 | Catalogue check | **No letter** | Public registries and identity data only |
| Tier 1 | Catalogue rating | Letter | Tier 0, plus the owner's attestation, the owner's exports from the societies it collects through, and statements, agreements and registration history where supplied |
| Tier 2 | Asset integrity assessment | Letter | Tier 1, plus a cash bridge from statements to bank, reconciliation of the population, exceptions and independence disclosures. Defined, not yet offered |

A tier 0 check is a check: it reports what public sources show and what they could not show, and never carries a letter. Coverage minimums apply only at tiers 1 and 2.

## What is surveyed

**The population.** The owner's attestation lists the works the rating covers before any lookup runs. The attestation fixes the population; it decides no state. A work recorded under the owner whose income belongs to someone else under an agreement (for example a work sold on with the owner left on the chain) can be removed by a **grading scope**, with a reason, the name of the person attesting, a date and an evidence reference. Removed works are never scored, and the rating discloses each one and the weight it would have carried.

**Rights positions.** Compositions (performing, digital and physical mechanical, sync) and sound recordings (neighbouring rights, master distribution) are surveyed separately, never in one denominator. A position the owner does not hold is not a defect: the work has no survey for it.

**Routes.** Scope is set by collecting **route**, not by where a work is used:

| Route | Collects |
|---|---|
| PRS for Music | Performing, including overseas performance income reaching a member through reciprocal agreements |
| MCPS | Digital and physical mechanicals |
| ICE | The registration hub for PRS and MCPS; takes no income of its own |
| The MLC | US digital mechanicals |
| PPL | Neighbouring rights, including PPL's reciprocals |
| SoundExchange | US neighbouring rights |

Out of scope, for example: direct memberships of other societies, income collected abroad by a sub-publisher, and mechanicals outside The MLC. CIS-Net, the ASCAP bulk data and the US Copyright Office show that a work exists and who is attached to it; they never decide a route's registration.

## The four states

Each work is judged on each component, and on registration at each route, as one of four states:

| State | Means |
|---|---|
| **Sound** | The evidence read shows it in order |
| **Defective** | The evidence read shows a named defect |
| **Not checked** | No evidence could answer, for a recorded reason |
| **Not relevant** | The question does not arise, for example a right the owner does not hold |

A work's overall condition is defective if any component is defective, otherwise not checked if any is not checked, otherwise sound.

## Why a work is not checked

Not checked is never a choice. It is always the outcome of a recorded limit, and it always carries one of six reasons. Each reason sits with one party:

| Reason | Who | What it means |
|---|---|---|
| Not supplied | The client | The client has not supplied the evidence that would answer: the society export, statement or agreement this check reads. For risk, no reading of the work at PRS, MCPS or The MLC |
| Too recent to appear | The society | The work or release is newer than the society's record has caught up with. Nobody has failed |
| Waiting for the society to publish it | The society | The registration is in the society's queue and not yet published |
| Not reached in this run | TrackForge | TrackForge's run did not read the work at this source: it stopped short, or never read this code there |
| Source could not be read | TrackForge | TrackForge could not read the source, or the source did not say whose the work is |
| Reading past its age limit | TrackForge | TrackForge's only readings are older than the source's age limit, so none may decide |

Where a work's registration is not checked at more than one route for different reasons, its registration state carries the reason that comes first in this table.

A not-checked work is left out of the component's score and counted as a shortfall in **coverage** instead, so it neither flatters nor harms the score.

## The five components

| Component | Asks |
|---|---|
| Identity | Can each work and recording be told apart and matched: ISRC, ISWC, writer IPIs, one registration per work |
| Title | Is the owner's claim to each work backed by a PRS agreement that binds the work |
| Registration | Is each work registered, under the owner, at every society where it earns |
| Collection | Is the money the registrations should bring in actually arriving |
| Risk | Disputes, competing claims, duplicate shells and sources that disagree |

**How a component is scored.** Each defect has a class, and each class a factor: a full deduction, a partial one or a minor one, or, for a share registered below what is owned, the size of the gap. A work's deduction on a component is the largest factor among its defects there. The component's score is, over the works assessed (sound or defective), the weight kept after deductions as a share of the weight assessed, out of 100.

**Registration is scored route by route.** A registration defect costs a work only that route's share of its income, not the whole work. A route that was not read for a work is left out of that work's sums: neither credited nor charged.

**Collection needs statements.** Without them it reads "collection not assessed", and the composite is taken over the other four.

**Risk counts as sound only where a risk check was read.** PRS, MCPS and The MLC can show a dispute or competing claim; PPL and SoundExchange cannot. A work is sound for risk when at least one of PRS, MCPS or The MLC read it and raised no risk. A work none of them read is not checked, reason not supplied, whatever other routes read it. So sound on risk means no known conflict was found in what was read, not that no risk exists.

## How works are weighted

A hit should count for more than a demo nobody plays, and a small stake for less than a whole one. Each work's weight is:

- **Demand, the larger part.** Within each income channel, the work's share of the catalogue's demand there, weighted by the catalogue's channel mix.
- **An equal part.** An equal share for every work, so works with no measured demand still count.
- **Scaled by the attributable share.** The blend is multiplied by the owner's share of the work and the weights rescaled to sum to one. A 1% stake carries 1% of the weight the work would carry at 100%.

**The demand measure.** For each channel the measure is statement lines where they are supplied; otherwise, for each work, Spotify lifetime plays per recording, from a dated snapshot, where Spotify shows a count on any of the work's recordings; otherwise YouTube views per recording. Spotify is searched by ISRC, then by title and artist. A title match counts only when it names one recording: the same base title with version text removed, the same artist, and a length that agrees closely. The snapshot keeps how each recording was matched, and a work sums its recordings. A refresh is a new dated snapshot, never a rewrite, and a snapshot that cannot be read stops the run rather than falling back to YouTube.

**Mixing plays and views.** Where a channel has Spotify plays for some works and only YouTube views for others, the views are converted into Spotify-equivalent plays. The ratio is plays per YouTube view, measured on the catalogue's own works that have both counts: the median of each such work's plays divided by its views. With too few works carrying both counts the views are not used where plays are, and those works get the equal share only. A channel where no work has plays keeps its raw views. The ratio, the number of works it was measured on and both snapshot dates are frozen into the rating.

Physical mechanicals use distributor sales, MCPS statement lines, then the release footprint; sync uses statement lines only. Income never sets a weight on its own: demand is the floor and income can only raise it. A work with no demand in any channel carries only its equal share and is reported in the unweighted tail. The measure and its date are frozen into the rating.

**Where the attributable share comes from.** In order: declared by the owner; from the owner's Workspace draft; from the society's chain; otherwise assumed whole. Every work states its source and the rating counts works by source, so a reader can see how many shares were assumed.

### Largest works

The largest works are the works with the highest weight under the rule above: demand first by statements, then Spotify lifetime plays, then YouTube views; mostly demand, with an equal part for every work; scaled by the owner's share. The Rating Report shows the condition of the top 20 and lists where a reader can look each one up.

## Routes and income

The channel mix says how much of the catalogue's income each channel carries; it is the owner's declaration, or statements where supplied. With no declared mix, a recording catalogue is taken as all neighbouring rights and any other as all digital mechanical.

A route's **expected-income share** is the sum of the mix over its channels. Two channels divide between routes:

- **Digital mechanicals**: The MLC takes the catalogue's US share and MCPS the rest. Undeclared, the US share is set deliberately above the US alone, so that leaving it undeclared never shrinks The MLC.
- **Neighbouring rights**: SoundExchange takes the catalogue's SoundExchange share and PPL the rest. Undeclared, it takes the same starting value.

If only one of the pair is in scope it takes the whole channel. A route below the materiality threshold of expected income is **immaterial**: its defects are found, listed and cost their share in the registration score, but it never caps the letter. With few lifetime streams across the catalogue and no US statements, The MLC is immaterial by default. The rating discloses each route's share and whether it is material.

A route that carries a rights type's main income must be in scope whenever the catalogue shows use in its channels: The MLC for compositions with digital releases, PPL and SoundExchange for recordings.

## From scores to a letter

**Composite.** The weighted mean of the component scores, over the components assessed.

**Letter before caps.** The composite read against the scale: each letter starts at a set composite, AAA highest and D lowest. The Rating Report shows where the catalogue's composite fell.

The letter issued is the letter before caps, lowered by any cap that applies. The lowest cap wins, and every cap is disclosed.

**Coverage minimums.** At tiers 1 and 2, coverage below the minimum withholds the letter, with the shortfall named. There is a minimum for:

- Identity: by count of clusters, with an alternative test by weight. A cluster is a work with its versions and stems, counted once.
- Title: by weight.
- Registration: by weight at each main route.

**Coverage cap (ladder).** Above the minimum, title coverage by weight, and registration coverage by weight at each route, cap the letter in steps: the less of the catalogue that could be checked, the lower the highest letter it can carry. The worst step met applies, and the Rating Report names it.

**Floors.** Some defects are too serious for an average. Each floor caps the letter at a set level when a set share of the catalogue carries the defect, whatever the composite says. There are floors for:

- works lacking an ISRC on an exploited recording (by count);
- a route's income unregistered, at two levels, the larger share capping lower;
- works without a PRS agreement binding them (by weight, where PRS is in scope);
- a route's income going to the wrong party;
- a termination notice or dispute among the works carrying the most weight.

The Rating Report names every floor that capped the letter.

"Unregistered" means not registered where exploited, registered under another party only, or a recording absent from PPL or SoundExchange. Floors on a route's income use each work's **income weight** there, the demand part of its weight alone, so unplayed works cannot trip a floor; they are applied per route, scaled by the route's share of income, and never at an immaterial route.

**When no letter is issued.** No letter is issued when a coverage minimum is not met, or when any single work carrying a material share of the weight is not checked because evidence was not supplied, the work was not reached in this run, or a source could not be read.

**When a rating cannot be issued.** A run can reach a letter and still not be issued. An issued rating has no lookup limit: issue is refused while any work at all, whatever its weight, is not checked anywhere because it was not reached in this run. That limit is TrackForge's own cap on lookups, a gap of TrackForge's making, so it is never a reason on an issued rating. The same reason covers a code TrackForge never read at the PRS works database, PPL or SoundExchange, so those gaps refuse issue too. A draft or rehearsal run may still stop short and keeps its letter; TrackForge finishes the reading before issuing.

## Estimated inputs

Some inputs are declared rather than measured: the US share, the SoundExchange share and the channel mix. Each states its value, its basis and a plausible range. The catalogue is re-rated at every combination of the ends of the ranges, and the letter issued is the **lowest** of those and the letter at the declared values. The composite and components shown are those at the declared values; the Rating Report lists the letter at each end. An input measured from settled statements says so, with a range of one value. A rating cannot be issued on a declared input with no range.

## Money

A rating shows only **measured amounts**: a figure a society shows for the work, or a statement line, each with its source, in whole pence shown as pounds. Amounts attach to the registration leaks, usage unmatched at The MLC and unclaimed shares. Where there is no such figure the rating says **no amounts**, never £0. Amounts are disclosure: they never change a score or a letter. Estimates of what works are worth belong in the separate Fix plan, as labelled ranges, outside the rating.

## Sources and dates

Every source is named with the date its reading was true. Where two sources of one kind answer, the newer wins. Where sources of different kinds disagree within their age limits, the higher-ranked source decides and the disagreement is recorded as a risk, so a manufactured disagreement cannot hide a defect. Each kind of source has an age limit, and a reading older than its limit never decides. Identity readings do not age. An acknowledgement proves only that a file entered a society's queue, never that it was registered.

A table TrackForge reads live is a dated read, pinned by a hash of the rows read, which travel with the rating. The MLC's bulk data is pinned by the hash of its manifest. MusicBrainz takes its date from its version. Spotify plays and YouTube views each take the date of their snapshot, and the plays-per-view ratio is fixed with the number of works it was measured on.

## What is not checked

Thirteen defect classes are defined in the rule book but nothing reads them yet. A sound result says nothing about them:

- Identity: no ISWC where the route allocates one.
- Title: the society holding a lower share than documented; the society holding a higher share than documented; a reversion or term expiry coming soon.
- Collection: no income despite usage; income to the wrong payee; an account on hold.
- Risk: a claim window closing; an administration or sub-publishing agreement ending.
- Opportunities: registered but not yet exploited; registered at a society outside the routes; sync not represented; physical exploitation not registered or claimed.

Collection is assessed only when statements are supplied.

## Leaks, risks and opportunities

A **leak** loses money now; a **risk** could lose money later; an **opportunity** is income not yet earned. Leaks and risks move the grade. Opportunities are listed, never scored.

## In flight, the projected letter and the outlook

Work sent to a society and acknowledged but not yet published never changes the letter. The Rating Report shows it beside the letter, not in it: a **projected letter** with acknowledged, uncontested items treated as landed, and an **outlook**. These are committed in the issue plan and are not yet rules in the rule book; see [in flight](/docs/certification/rating/in-flight).

## The scale

A letter describes the registration condition of the catalogue on its date. It is not a valuation, a credit rating or a legal opinion on title.

| Letter | Means |
|---|---|
| AAA | Registration is in order; only isolated minor items remain. |
| AA | Very few defects, none on the main earners. |
| A | Sound overall; defects are minor and not concentrated. |
| BBB | Adequate: most income is sound, with some material defects to fix. |
| BB | Noticeable income is exposed; expect a remediation plan before relying on it. |
| B | Substantial defects across the catalogue. |
| CCC | Widespread defects; much of the income is at risk now. |
| CC | Most of the catalogue is defective. |
| C | Nearly all of the catalogue is defective. |
| D | The registration position cannot be relied on. |

The scale is not yet calibrated: the composite at which each letter starts is a starting value, reviewed after the first tier 1 ratings.

## The rule book

The rule book is versioned. Each change takes a new version; a published version never changes. It is published by its SHA-256 hash, which is timestamped on Bitcoin before the first rating under it, and every rating names the version and hash it was reached under. The rule book itself, the scoring code and the reference model stay private; a party relying on a rating can have it recomputed under a separate agreement.

Source: https://trackforge.studio/docs/certification/rating/method-statement/index.mdx