A split payment is three transfers wearing one name
Checkout shows a buyer one number. Behind it sit at least three movements: the acquirer pulls funds from the buyer's card or account, those funds land somewhere under someone's name, and then separate payouts leave for the seller, the platform commission, and often a shipping or tax recipient. Marketplaces call the whole thing a split payment. Banks see three unrelated transfers, and the timing gaps between them are where most operational problems live.
That gap is not a detail. If you settle sellers before the acquirer settles you, the platform is lending money it has not received. If you settle after, sellers call support because their dashboard shows a sale they cannot spend. The design question is not "can we split a payment" — most processors can — but "whose balance sheet carries the float, and for how long."
Who holds the money between checkout and payout
Three models dominate, and each shifts risk to a different party.
Platform as merchant of record. The platform takes the payment into its own account, then pays sellers as a supplier. Simple to launch, and it makes the platform the legal seller — which drags in tax obligations, refund liability, and chargeback exposure. It also turns seller payouts into ordinary vendor payments, so you inherit supplier onboarding and payment-rail limits for every seller.
Separate accounts per seller. Each seller gets a sub-account or virtual account under the provider's umbrella. Funds are legally segregated, the buyer's money never touches the platform's operating account, and payouts become internal transfers. This is the cleanest model for compliance, and the most work to onboard: every seller needs KYC, and dormant sellers create a long tail of accounts that still need monitoring.
Licensed intermediary or marketplace facilitator. The provider holds a licence and acts as the collection agent, splitting at the source. Fastest to market if the provider supports your corridors, least flexible when a seller wants a rail the provider does not cover.
Whichever you pick, write down two numbers before you launch: how many days funds sit between capture and payout, and what that float is worth at your monthly volume. Holding three days on $2 million of monthly gross merchandise value is roughly $200,000 of someone's money at any moment. Decide deliberately whether that is yours, the sellers', or the provider's.
The settlement order that decides who waits
Split payments fail most often on sequencing, not on technology. Four clocks run at once: card authorisation, acquirer settlement (T+1 to T+3 for cards, same day to T+2 for local rails), the marketplace's own payout schedule, and the seller's bank posting time.
A workable default looks like this:
- Capture immediately, credit the seller's ledger immediately, but mark it pending. Sellers see the sale; the balance is clearly not yet withdrawable.
- Release funds only after the acquirer's settlement is confirmed — not after the authorisation, and not after a fixed timer that assumes settlement happened.
- Batch payouts on a schedule, not per order. Per-order payouts multiply fees and reconciliation work by the number of sellers.
- Hold a rolling reserve for new sellers — commonly 5–10% for 30–90 days — instead of holding everyone's money.
The failure mode to avoid is a fixed "payout three days after delivery" rule that ignores how the money actually arrived. If a buyer pays by local bank transfer that settles in 24 hours, you are holding seller cash for two extra days for no reason. If a buyer pays by a card that settles in three, you are paying sellers from your own pocket for a day. Both are policy bugs, not payment bugs.
Chargebacks, refunds and the seller who goes negative
Card chargebacks can arrive up to 120 days after the sale, long after the seller has been paid. Refunds requested by the buyer have the same shape: the money must leave immediately, but the seller's balance may already be zero.
Three controls keep this from becoming a collections problem:
Reserve per seller, not per platform. Track a rolling reserve tied to each seller's dispute history. A seller with a 0.9% chargeback rate needs a bigger buffer than one at 0.1%, and pooling the reserve hides the difference until it is expensive.
Allow negative balances, then enforce them. If a refund pushes a seller below zero, the next sales should offset it automatically before any payout leaves. Blocking the payout entirely is better than chasing an invoice.
Decide who absorbs the loss in the contract. Marketplace terms usually make the seller liable, but if the seller has left the platform, the platform absorbs it. Price that residual risk into commission rather than discovering it quarterly.
What breaks between 50 and 5,000 sellers
At 50 sellers, splits are a spreadsheet with an API attached. At 5,000, four things break at once.
Onboarding throughput. KYC per seller, with document review and exceptions, becomes the bottleneck. Sellers who cannot be verified cannot be paid, and they file tickets that all land on the same two people.
Payout file size and rail limits. A single ACH or SEPA batch has practical limits, and a rail that works for 200 payouts may need splitting, staggering or a different provider for 5,000. Some local rails also cap per-transaction value, which forces a high-value seller onto a wire and a different fee line.
Reconciliation. Every payout must map back to orders, fees, refunds and reserves. Without a stable payout ID per seller per batch, finance ends up matching bank lines to spreadsheets by amount and date — which works until two sellers get paid the same amount on the same day.
Support surface. Sellers ask "where is my money" in the language of their own rail. If your support team cannot see the rail-level status, every question becomes an escalation.
The scaling fix is boring: one ledger of record, an append-only transaction journal, and payout objects that carry their own status and evidence. Everything else — dashboards, support macros, tax reports — reads from that.
Matching payout rails to seller geography
Sellers care about one thing: when the money lands, in what currency. The rail is your problem, not theirs — until you pick a rail they cannot receive.
| Seller location | Practical payout rail | Typical timing | Watch out for |
|---|---|---|---|
| United States | ACH, same-day ACH, domestic wire | 1–3 days; wire same day | Account type checks; wire fees on small payouts |
| Euro area | SEPA credit transfer, SEPA Instant | Same day to next day | IBAN name matching; instant limits per bank |
| United Kingdom | Faster Payments | Minutes | Confirmation of Payee mismatches |
| Southeast Asia | Local transfers, e-wallets | Minutes to 1 day | Wallet caps; required beneficiary fields vary by country |
| Latin America | Local bank transfer, instant schemes | Same day | Tax IDs required; FX spread on conversion |
| Cross-region, high value | SWIFT or local collection account | 1–3 days | Correspondent fees; intermediary deductions |
Two mistakes repeat here. First, paying every seller in USD because it is easy for the platform — sellers then absorb the conversion cost and quietly move volume elsewhere. Second, adding a rail per country without a settlement account behind it, which produces payouts that are "sent" but arrive three days late with fees deducted by an intermediary bank.
The questions worth asking before you commit
Most marketplace payment decisions are made on a demo and a rate sheet. Six questions expose the gaps faster:
- Under whose name do buyer funds sit while they wait to be paid out, and what happens to that money if we stop operating?
- Can payouts be batched per seller with a stable reference that survives a rail change?
- What happens to a seller's negative balance — automatic offset, hold, or invoice?
- Which corridors can you pay in local currency, and which convert at your rate?
- How do you handle a refund that arrives after the seller's account is closed?
- What is the settlement evidence we get per payout, and can finance reconcile from it without asking you?
If the answers are vague on float and negative balances, the pricing is not the real cost. Wondergate's global payments rails cover collection and multi-rail payout in one integration, and the payout API integration guide walks through the ledger and webhook design in more detail. If you are still deciding whether to hold a licence or work through one, the comparison of EMI, bank and PSP models covers the trade-offs, and you can book a corridor review to map your seller geography against available rails.
Frequently asked questions about marketplace split payments
Q1: Does a marketplace need its own payment licence to split payments?
A: Not necessarily. Many marketplaces operate as a merchant of record or work through a licensed provider that acts as collection agent. The licence question becomes unavoidable when you hold funds on behalf of sellers for any length of time, or when you want to pay out in currencies where your provider has no coverage.
Q2: How long should a marketplace hold funds before paying sellers?
A: Long enough to confirm the acquirer settled and to cover the dispute window you cannot control, short enough that sellers do not feel penalised. Most marketplaces release after settlement confirmation plus a rolling reserve for new sellers, rather than a single fixed delay for everyone.
Q3: Who pays the chargeback when a seller has already been paid out?
A: Contractually usually the seller, operationally almost always the platform first. The workable pattern is to offset the loss against the seller's next sales, hold the reserve, and only escalate to collections when the seller has stopped transacting.
Q4: Can sellers be paid in their local currency?
A: Yes, but the cost depends on whether the payout rail has a local settlement account behind it. Paying in local currency through a correspondent bank adds intermediary fees and delays; paying through a local rail settles faster and usually cheaper, at the cost of more integration work per corridor.
