1. Processing that survives scale
Royalty runs grow with the catalog. A platform should process a period without multi-day batch jobs or crashes that force restarts, and let you validate numbers in a draft state before anything is published. Look for a draft, review, publish, and close workflow — calculation-only runs you can inspect before a single statement goes out.
2. Imports you can preview, not pray over
Sales files arrive from every DSP and distributor in different shapes. The difference between hours and minutes is whether the platform lets you map columns with a live preview of transformations before processing — and whether those mappings are saved as reusable templates instead of rebuilt each period.
3. Contract depth without a developer
Splits, advances, recoupments, escalations, and cross-release terms should be expressible by the royalty team directly. Tree-structured deal models let you define terms once and apply them across artists and releases — and change them later without filing a ticket to IT.
4. Reconciliation without re-imports
Unmatched or unreported sales are a fact of life. If handling them means extracting rows and re-importing them next period, you have a duplicate-data risk built into the workflow. The platform should carry unreported sales forward into the next period automatically.
5. Auditability by default
Who changed the contract, who requested the payment, who approved it — a serious platform answers these from its audit log, with permissions that can be limited by responsibility and payment thresholds. If auditability is an add-on, statements become arguments.
6. Artist-facing transparency
Artists increasingly expect to see their own numbers. Statement portals, payment requests with label-side approval, self-bill invoicing, and payment-file export reduce the invoice-checking and manual admin that consumes label teams around payout time.
7. A migration path you can believe
The best software is worthless if you can't get onto it. Ask any vendor how historical statements and opening balances come across, whether a parallel run is supported, and how a cutover aligns with reporting-period boundaries. Vague answers here predict painful migrations.