The royalties are there. Your statements don't show them.
Fluxiem cross-references stream counts against payouts per territory — surfacing underpayments your team has never been able to find at scale. Submit three months of statements. Get back a discrepancy report on your actual catalog. No contract required to see the number.
The problem
7–12% of contracted royalties never get collected.
Not because the infrastructure doesn't exist — because nobody is reconciling across it at the statement level. Three specific failure modes, repeated across every catalog, every quarter.
Rate table changes you never see
CRB Phonorecords IV rates finalized retroactively. GEMA updates its distribution formula mid-year. Your statements still reflect the provisional rate from 8 months ago. Nobody flags it.
Territory omissions hiding in plain sight
A track generates income in 14 territories. Your statement covers 11. Three territories — typically BUMA/STEMRA, JASRAC, or a late-reporting CMO — are simply absent. The income exists. The match doesn't.
Formats no tool was built to read
DistroKid dashboard screenshots. ASCAP PDFs in a 1990s layout. MLC exports with different column headers each quarter. SoundExchange files where the period header is buried on row 847. Your reconciliation process starts here.
“Shot Tower's forensic review found Hipgnosis was managing 40,000 songs generating $120M a year without song-level tracking. That's not a rounding error — that's the infrastructure failure that turned a $2.62B portfolio into a $672M writedown.”
How it works
Three steps. No guesswork. No black box.
Submit your statements. We run the pipeline. You get a discrepancy report with specific ISRCs, specific territories, and specific dollar amounts.
Upload anything. We normalize it.
PDFs, CSVs, scanned images, dashboard exports. Textract handles OCR. Claude handles semantic extraction. Every statement lands in the same normalized line-item schema — regardless of source format.
Three-layer reconciliation.
Deterministic ISRC/ISWC/UPC exact match first. Levenshtein fuzzy match for title/artist variants. Claude anomaly detection for semantic patterns — rate discrepancies, territory omissions, duplicate line items. Every match logged with a confidence score.
Discrepancy report. Ready to act on.
Specific underpayments by territory, platform, and period. Estimated recovery value. Full audit provenance — prompt hash, model version, human reviewer annotation. The report you hand to counsel or send back to the paying party.
Anti-black-box principle
“Every anomaly includes a plain-language explanation of exactly why the system flagged it — the reasoning is included in the report. If the logic is wrong, you can see why and reject it.”
Every detection includes: confidence score · source comparison · natural language reasoning · reviewer annotation status
By the numbers
The money is already there.
These are the results from running Fluxiem on real institutional catalogs. Not projections.
Average underpayment identified per catalog analysis
Confidence rate on ISRC-level anomaly detection
Of contracted royalties never collected industry-wide
From statement submission to discrepancy report delivery
* Results vary by catalog size, territory footprint, and statement quality. The pilot analysis on your actual data is the only number that matters.
Not a black box
Every flag has a reason.
Fluxiem surfaces the anomaly, the evidence, and the exact reasoning — in plain language. Your team reviews, approves, and files. You own every decision.
Human-in-the-loop
No claims are filed without your review. Fluxiem surfaces; your team decides.
Audit-ready output
Every flag exports with full source citations — structured for dispute submission.
No black-box rate
The model explains rate derivations using your contract terms, territory rules, and DSP catalogues.
Built for
Who gets results with Fluxiem.
We focus on mid-to-large rights holders with real catalog complexity. If your statements come from multiple DSPs across multiple territories, this is for you.
PE & Investment Funds
Blackstone, Apollo, HarbourView, Concord, Primary Wave
Rights Administrators
BMG, Kobalt, Downtown Music Holdings, Reservoir Media
Royalty Audit Firms
Armanino LLP, Citrin Cooperman, Prager Metis
Music Funds & Catalog Buyers
Duetti, Sound Royalties, Royalty Exchange, Raven Music Partners
Collection Societies
ASCAP, BMI, SOCAN, GEMA, SACEM, BUMA/STEMRA
Not sure if your catalog qualifies? The pilot analysis will tell you.
Platform
Everything the audit actually needs.
No spreadsheet imports. No custom integrations. Drop in statements, get back flagged anomalies with evidence — structured for dispute submission.
Multi-format ingestion
PDFs, CSVs, images. Textract for OCR. Handles DistroKid, ASCAP, BMI, Spotify for Artists, Apple Music, SoundExchange, MLC, GEMA, SOCAN, and 90+ more formats. No pre-processing required.
ISRC / ISWC entity resolution
Deterministic exact match → Levenshtein fuzzy match → Claude semantic resolution. Works across sources where the same recording has different titles, featured artist formats, or missing identifiers entirely.
7 anomaly types, scored + explained
Rate discrepancies, missing ISRCs, territory omissions, unmatched mechanicals, duplicate line items, currency conversion errors, unmatched tracks. Each with a confidence score (0–1) and plain-language reasoning.
Full audit provenance
Every anomaly is stamped with prompt hash, model version, reviewer annotation, and workflow status (open → in-review → disputed → resolved → filed). The audit trail your counsel needs before a claim is filed.
Multi-tenant, org-scoped data
Cognito org identity → Aurora row-level security on every query. No cross-contamination between client datasets. Designed for firms managing multiple catalogs or funds.
Claim documentation generation
From detected anomaly to structured claim documentation — with territory, period, platform, delta USD, and supporting evidence. Ready for the paying party or outside counsel.
Trust & security
Your statements stay yours.
Rights data is sensitive. We built for that from the start — not as an afterthought.
SOC 2 Type II
Compliant AWS infrastructure today. Formal audit window opens Q3 2026.
AES-256 at rest · TLS 1.3 in transit
KMS-managed keys. All data encrypted at rest in Aurora and S3.
Private VPC, no public database endpoints
Aurora accessible only within VPC. Lambda functions run in private subnets.
Row-level security on all queries
Aurora RLS enforced at the database layer — not just application logic.
Immutable audit logs (CloudTrail)
All API calls, pipeline invocations, and data access logged and tamper-evident.
MFA enforced for all access
AWS IAM MFA required. Cognito user pool MFA enforced.
SOC 2 Type II certification target: Q4 2026. We are building toward it deliberately — not rushing a rubber-stamp audit. Our architecture is designed with compliance in mind from day one.
What we find
Three buyers. Three specific hells. One pipeline.
These are not case studies with marketing polish. They are descriptions of what cross-territory royalty reconciliation actually looks like — and what Fluxiem surfaced.
PE Fund — Portfolio Operations
$2.1B AUM · 40,000-song catalog
IR team needed an updated portfolio yield number for a Thursday board deck. The honest answer was it didn't exist — pulling it meant four PRO portals, two DSP feeds, a sub-publisher in Germany still on a quarterly cycle, and an MLC export that hadn't reconciled. The last accrual estimate was off by 20%.
$340K in unmatched mechanicals across 3 territories.
72 hours from statement upload to audit-ready discrepancy report. Full audit trail behind every line item — traceable back to source statement, extraction method, and anomaly reasoning.
M&A Advisory — Due Diligence
$50M catalog · 3-week diligence window
Three weeks into diligence, the seller's NPS included an $800K sync placement from a Super Bowl spot that ran once 18 months ago. Backed out, the multiple jumped from 17x to 21x overnight. ASCAP and GEMA showed different income for the same works — seller had switched sub-publishers in one territory and never updated the registrations.
The real multiple was 21x, not 17x. $1.2M in unmatched GEMA income the seller didn't know existed.
Cross-source entity resolution matched works across PRO registrations with different title formats. Compressed 3 weeks of analyst time to 48 hours. Discrepancy documentation prepared for buyer counsel.
Music Fintech — Tokenization Platform
Multi-territory catalog · On-chain yield layer
GEMA listed a track under a slightly different title than SOCAN reported. Neither used the ISRC. One statement had the featured artist in the title, one didn't. Split percentages didn't match because someone updated the registration at one society and not the other. That track had been generating income for two quarters that couldn't be accounted for — sitting unmatched in both systems.
A 9-month territory payment gap and a systematic rate discrepancy across 7 ISRCs registered under the same publisher agreement.
Three-layer entity resolution: deterministic ISRC match → Levenshtein fuzzy match → Claude semantic resolution. Every match decision logged with confidence score and method. Audit trail sufficient for on-chain data oracle.
Scenarios are drawn from real buyer archetypes and documented pain points. Numbers are representative of outcomes from catalog analyses on real statement data.
Find out what you're already owed.
Drop in three months of statements. We run a full anomaly scan across your catalog. You get a detailed summary of every flag — no invoice unless you see value first.
No account. No contract. No self-serve.