Choosing a Merchant of Record: Costs, Control and Exit Terms

The distinction between a MoR and a "seller of record" arrangement confuses most teams, because some providers now sell both. Passport, which offers each, describes the split as follows: the MoR is the legal seller and typically manages payments, tax collection and remittance, and disputes, while a seller-of-record model keeps tax compliance outsourced but leaves payments, checkout and cash flow with the brand (Passport on merchant of record versus seller of record).

That split matters for cash flow. Under a full MoR, gross revenue lands with the provider and you are paid net on a schedule. Under a seller-of-record or processor-plus-tax model, funds settle to your own merchant account on normal processor timelines and the tax vendor invoices you separately. Finance should see both before the contract is signed.

There is also a liability spectrum inside the MoR category. Paddle separates "complete MoR platforms" that take liability unconditionally from "specialist or hybrid" arrangements that assume liability in most scenarios with documented carve-outs where the seller keeps responsibility (Paddle's evaluation framework). Read those carve-outs. They are where your assumptions break.

Numeral is blunt about the cost: handing off obligations simplifies expansion and cuts compliance burden, but it also means giving up parts of the customer experience, which can hurt brand cohesion (Numeral on whether SaaS should use a MoR). Decide in writing which of those you can live with.

Never compare headline percentages. Build the all-in cost per dollar of gross revenue, then compare it with the fully loaded cost of doing the work yourself. Collect four components from every candidate: the percentage of transaction value, the per-transaction fixed fee, chargeback and dispute fees, and currency conversion and payout charges. Card payments typically run 1.5% to 3.5% per sale for credit cards, with online transactions priced higher than in-store because of fraud risk, and chargeback penalties sitting on top as a separate line (SHOPLINE on how merchant fees work).

Then check the hidden lines. SHOPLINE lists PCI compliance fees, gateway fees, cross-border surcharges, statement fees, batch processing fees and early termination fees as distinct merchant-fee categories (SHOPLINE). Ask for a written schedule of every one of these before you model anything.

Here is a worked example. Use your own numbers; the arithmetic is the point.

Annual gross revenue:            $2,000,000
Total transactions:              40,000

Option A: processor + tax add-on
  Processing at 2.9% + $0.30     = $58,000 + $12,000 = $70,000
  Tax software + filing          = $18,000
  Internal finance (0.4 FTE)     = $32,000
  All-in                         = $120,000  (6.0% of revenue)

Option B: MoR
  All-in rate 5.0% of gross      = $100,000
  Per-transaction $0.50          = $20,000
  Chargebacks, 0.5% at $15       = $3,000
  All-in                         = $123,000  (6.15% of revenue)

At this revenue the two are within noise. The MoR wins on risk transfer and time; the processor wins on control and data. Now change one input: if your average ticket were $5 rather than $50, transaction count would be 400,000 and fixed fees would swamp the percentage difference. That is the decision rule. A MoR's per-transaction fee decides the comparison at low ticket sizes; the percentage rate decides it at high ticket sizes.

The real cost of a self-run stack is not software, it is registration surface area. Thresholds differ by state and change often. The Sales Tax Institute's chart shows California at $500,000 with no transaction threshold, Alabama at $250,000, Nevada at $100,000 or 200 transactions, and Connecticut requiring both a dollar and a transaction threshold (the Sales Tax Institute's economic nexus chart). Wolters Kluwer's summary notes that Connecticut and New York require both thresholds to be met, while most states use one or the other (Wolters Kluwer on economic nexus).

Then layer on product taxability. Anrok's rate table shows Arizona and Texas taxing SaaS, California and Florida not taxing it at the time of that table, and Colorado taxing it only in certain localities (Anrok's SaaS sales tax by state). TaxJar's guide shows Connecticut charging the full state rate on personal-use SaaS but only 1% on business use, and Maryland charging 3% on business use (TaxJar on SaaS sales tax). Each rule is another return, another registration and another place to be wrong.

Non-compliance does not just mean paying the tax later. It means interest, penalties and back taxes, and it can escalate into legal issues and reputational damage (PayPro Global on SaaS sales tax nexus). Registering late can produce penalties even where no tax is owed: in Austria, an intentional failure to register can trigger a penalty of up to EUR 5,000 (Stripe's EU VAT guide).

Two structural facts make the DIY option harder than it looks. First, some markets have no small-seller threshold for nonresident digital sellers. In the UK, a nonresident selling digital products to consumers must register from the first B2C sale, at a 20% standard rate, with B2B sales handled under the reverse charge (Anrok's UK VAT guide, and HMRC's guidance on digital services to consumers). Second, simplification regimes are real but partial. The EU One Stop Shop lets you register once, file one return and make one payment for cross-border B2C sales at the customer's country rate, keeping records for up to 10 years in case of audits (the EU's One Stop Shop page). It does not replace domestic VAT returns, and OSS covers cross-border B2C transactions only (Numeral on EU VAT OSS).

So the honest framing is: OSS and its equivalents give you one portal, not zero work. You still own registration, evidence of customer location, filing cadence and audit defense. Price that honestly before concluding the MoR fee is expensive.

Score each surface where the MoR's name replaces yours. The question for every row is whether you can be the only name shown.

• Checkout. Can you host checkout on your own domain, or are you redirected to a hosted page? Ask for the exact flow.
• Descriptor. What string appears on the statement? This drives dispute rates and support tickets.
• Invoicing. Who issues the invoice, with whose tax number? B2B customers often cannot process an invoice that is not from their supplier.
• Pricing and packaging. Can you change plans and create custom contracts without provider sign-off?
• Billing logic. Usage metering, proration, credits, dunning. Paddle's business-profile framework flags native usage metering and pricing flexibility as deciding factors for usage-based businesses (Paddle's evaluation framework). If you bill on tokens or credits, test this first.
• Support boundary. Billing support often routes through the MoR rather than your queue (Stripe).

The decision rule: if any of these surfaces is a core differentiator, do not hand it to an MoR for your primary market. Use the MoR for the long tail of markets where compliance cost exceeds revenue. Some arrangements support transaction-level or geography-level coverage, letting you route specific markets through the provider rather than switching the whole stack (Stripe).

Ask who owns the customer record before you ask about fees, because this is the item you cannot fix after signing. Under a MoR, the end customer's contract is with the MoR, which affects your own marketing, analytics and renewal defense. Put six questions in writing:

1. Do we receive the full customer record (name, email, billing address, tax ID, country evidence) in real time, or only on export?
2. Is there a webhook or API for every subscription event, with a stable payload?
3. Can we export transaction-level data including tax amount, rate, jurisdiction and legal basis on demand, without a support ticket?
4. Who is controller and who is processor under data protection law, and is there a signed data processing agreement?
5. Can we use the data for our own marketing, including after termination?
6. For how long after termination will historical records stay retrievable, and at what cost?

On question 3, retention is a legal requirement, not a preference: EU OSS requires records kept for up to 10 years in case of audits (the EU's One Stop Shop page). If your MoR holds those records and your export is a partial CSV, you are exposed. Test the export during the trial. On question 1, make a test purchase and check whether your own systems receive the customer email and country evidence as a structured event. If the answer is "you can see it in their dashboard," you do not own the data in any useful sense.

Exit is where the cheapest quote becomes the most expensive decision. Read four clauses in this order.

Termination for convenience and notice period. Can you leave without cause, and with how much notice? Ask whether notice runs from the date you email them or the date they acknowledge.

Minimum term and early termination fees. SHOPLINE lists early termination fees as a distinct merchant-fee category, and there is no reason to assume MoR contracts are gentler (SHOPLINE). Get the formula, not the principle.

Migration assistance, priced. What exactly do they hand over: a data export, documented API mapping, a named contact, a parallel-running window? For how many days, and is any of it billable? Ask for the migration schedule as an appendix.

Live subscriptions and stored payment credentials. This is the item that sinks migrations. Your recurring customers' card credentials belong to the MoR's merchant account. Ask what happens to active subscriptions on termination date, whether they can be migrated with consent, and who collects renewals in the gap. Without a clean answer, assume involuntary churn across your subscriber base.

Then cover two tails: refunds and chargebacks on transactions the departing MoR processed (who funds them, from what reserve, and when is any rolling reserve released), and tax liability if an authority assesses transactions processed during the term. A practical method: write the migration plan yourself, as the operations lead who has to execute it, then ask the vendor to correct it in writing. Their corrections are your real contract terms.

Score each candidate on four weighted dimensions, and set weights before you see quotes so the exercise does not become a rationalization.

Run the model twice: at current revenue and at roughly double your current international share. Paddle's guidance is to profile where you expect to be in 12 to 18 months rather than where you are today, because the criteria that matter shift with business model (Paddle's evaluation framework).

Two final gut checks. If the DIY option wins on cost but you cannot staff the filing cadence, you have not found a cheaper option, you have found an unfunded liability. If the MoR wins on every dimension, check that you have pushed hard enough on exit; a strong sales process produces weak scrutiny of the termination clause.

Once the payment and tax stack is settled, the next build is usually the product surface around it: pricing pages, usage dashboards, support tooling and the internal agents that draft invoices or reconcile settlements. If your team is wiring models into those workflows, Quanta Inference gives you text, image, video, audio and embedding models through one OpenAI-compatible API key with prepaid US dollar pricing.