SubSweep

Agencies that own the subscription and billing layer stop selling hours and start governing a client's cash flow, which converts project work into multi-year retainers. The strategic risk is lock-in: the platform that automates invoicing, tax, and dunning also becomes the system a client cannot leave without a migration project. Choosing the billing stack is therefore a client-retention decision, not a procurement decision.

Recurring Revenue Custody is the question of who holds the payment relationship, the tax liability, and the renewal trigger inside a client's subscription business. Three custody models exist. The agency can hold custody directly, running invoicing and dunning on infrastructure such as WHMCS or Blesta. A merchant-of-record can hold it, with Paddle absorbing global tax compliance across 300+ markets while the agency loses direct customer contact. Or the client retains custody and the agency operates the tooling, as with MemberPress or Memberful deployments on client-owned WordPress installs. Custody determines what happens when the engagement ends: an agency holding the billing relationship keeps leverage and data, while one operating inside a client's account walks away with nothing. Custody also carries liability. A class action filed in September 2026 accuses Anthropic of overselling Claude subscription capacity through deceptive usage multipliers, a reminder that whoever sells the subscription absorbs the dispute. Map custody before signing, not after.

The Merchant-of-Record boundary is the line where tax liability, chargeback exposure, and payment failure handling stop being the agency's problem and start being the vendor's. On one side sit platforms like Paddle, which acts as the legal seller across 300+ markets and absorbs VAT, sales tax, and fraud disputes. On the other side sit gateway-based stacks like Chargebee or MemberPress, where the agency's client remains the merchant and owns every compliance obligation. The framework matters because agencies routinely quote a billing build without pricing the compliance work it creates. A client selling digital memberships into the EU can face registration thresholds in each member state; moving that client onto an MoR model removes the filings but adds roughly 5% of transaction value in fees. The trade is not cost versus no cost. It is predictable margin compression versus unbounded administrative exposure, and the right answer changes as the client's revenue mix shifts.

Dunning Recovery Window treats every failed renewal as a timed decision rather than an accounting event. Involuntary churn is recoverable only inside a narrow band: card retries, in-app prompts, and backup payment methods work in the first days, then recovery odds fall sharply and the client relationship resets to a sales conversation. Agencies that own this layer protect retainer continuity, because a client whose own subscribers churn from failed cards blames the agency running the billing stack. The window differs by model: a WordPress membership built on MemberPress or Paid Memberships Pro can retry through Stripe and PayPal over several days, while a Merchant-of-Record setup such as Paddle absorbs tax and fraud handling but still hands the agency the recovery sequence. Chargebee and Recurly ship dunning automation, yet the sequence, timing, and messaging remain agency work. Treat recovery rate as a deliverable metric in the retainer, not a platform setting nobody reviews.

Before signing or migrating, test the exit: export the full customer and invoice history, confirm the pricing models you will need in 18 months, and price the switch in hours and dollars.

Before signing a client onto any billing platform, document the data export format, the migration path, and the cost of leaving, then price that exit against the projected operational savings.