Food Delivery & Quick Commerce: split carts, instant tips, and clean money moves between three parties

Food Delivery & Quick Commerce

Food delivery is a triangle: the customer pays, the merchant earns, and the courier gets compensated—often in three different currencies and on three different clocks. If your payments layer doesn’t mirror that triangle exactly, you’ll bleed margin through refunds you can’t justify, courier payouts that bounce, and merchant statements that spark weekly arguments. Here’s how to wire the cash so operations move fast and finance sleeps.

The order is three transactions (at least)

Treat every basket as a bundle, not a blob:

  • Customer side. Authorize the full estimate (items + delivery + platform fee + tips if pre-tipped). Capture in parts: the cooked subtotal at dispatch, adjustments at handoff (out-of-stock swaps, weight items like produce), and the tip when it finalizes. Keep the tip on a separate sub-ledger; don’t net it into revenue.
  • Merchant side. Commit their gross (menu price × quantity) minus platform commission, promo funding (who pays?), and taxes you’re legally responsible to collect as a marketplace. If items were out, reflect that on the same day—merchants hate “mystery netting” two cycles later.
  • Courier side. Fare = base + distance/time + surge/zone + incentives + tip (where allowed). Pay on a cadence the labor market expects (weekly baseline, same-day or instant as an opt-in), and make every incentive an explicit line with its own rule.

When these three views reconcile by ID and timestamp, disputes go down and you stop writing long Slack threads to explain five-euro differences.

Checkout and rails that match the street

Cards convert, but local bank rails save real money and make refunds painless at dinner rush.

  • Methods mix. Keep cards and major wallets for ubiquity; add account-to-account rails where they’re mainstream—SEPA Instant (euro area), Faster Payments (UK), Pix (Brazil), ACH-equivalents domestically. Route by approval rate and cost per successful order, not by headline fees.
  • Rate discipline. Tourists and expats pay cross-currency. Lock the exchange rate at authorization and reuse it for partial captures and refunds inside a short window so “you short-paid me” tickets don’t swamp support.
  • Descriptors that stick. City + brand + “delivery” beats clever names. Friendly fraud falls when the statement matches what the customer remembers.
  • 3-D Secure / SCA. Use exemptions for low-risk baskets; step up instantly on risk signals (device mismatch, new address + high ticket, card BIN from a different region).

Tips: don’t improvise the most emotional part of the flow

Tipping drives driver supply at peak times. Get the money mechanics right:

  • Separate object. Tips are their own ledger lines with timestamps (pledged at checkout, finalized at handoff, or added post-delivery), their own tax treatment, and their own refund rules (usually “no refund” unless order never delivered).
  • Payout timing. Post-delivery tips should flow into the next courier payout, not sit as a month-end adjustment. Instant-tip cash-outs are a powerful retention perk—price the speed with a small, transparent convenience fee.
  • Transparency. Couriers see tip amount and value date; customers see where the money went. Opaque tip handling invites PR trouble and chargebacks framed as “I never tipped.”

Refunds and “make-goods” without trench warfare

You’ll issue refunds. The only question is whether they’re explainable.

  • Tight taxonomy. “Never arrived,” “late beyond SLA,” “missing/incorrect,” “quality,” “technical cancel.” Each reason maps to who funds (platform vs merchant), how much (full vs partial), and which rails.
  • Evidence pack. GPS breadcrumb, handoff photo or code, courier motion in the last 5 minutes, store ticket, and chat transcript. For weight-priced SKUs, store the weigh-in delta. Support shouldn’t guess; they should choose from facts.
  • Refund rails. Default to the original method. Where bank rails are instant, use them for goodwill—customers remember minutes, not policies.

Win rate on chargebacks jumps when you submit these artifacts as a bundle instead of free-typing a story.

Merchant settlements the kitchen can read

A restaurant will ignore your finance theory; they want a statement that matches their night.

  • Cut-off clarity. Freeze at a predictable hour (local time), settle on a calendar (e.g., T+2), and show line items: orders, cancellations, adjustments, promos (who funded), fees, VAT/GST, and net.
  • Virtual accounts. Give each merchant a dedicated virtual IBAN or local virtual account; inbound adjustments and ad credits land with identifiers your auto-recon can match without reading PDFs.
  • Reserves with sunlight. Hold a corridor-sensitive rolling reserve to cover post-delivery refunds and disputes; publish balance, release dates, and reasons in the merchant portal. Blanket reserves breed support tickets; targeted reserves change behavior.

Courier payouts that don’t bounce

Your supply side quits if payday wobbles.

  • KYC/KYB depth. Individual vs fleet operator matters. Capture IDs or business docs, tax numbers, and beneficiary bank verification (name-match/Confirmation of Payee where schemes support it).
  • Rails. Use domestic instant rails by default; wires only for high values. Failed payouts get expensive quickly—name mismatch and closed accounts are the usual culprits; fix with upfront verification and cool-offs on payee changes.
  • Incentives as code. Peak-hour boosts, quest bonuses, new-zone incentives—each has math your ledger can recompute. If you can’t replay an incentive, you’ll overpay quietly.

Promotions, vouchers, and who actually pays for them

Every discount needs a sponsor.

  • Funding tags. Label each promo as platform-funded, merchant-funded, or shared. That tag follows the order to both the merchant statement and your P&L.
  • Stacking rules. Prevent double-dipping (voucher + bundle + loyalty) unless you intend it.
  • Tax posture. In many regimes, discounts reduce the taxable base only if seller-funded; marketplace-funded promos often don’t. Encode the rule per jurisdiction so your VAT/GST returns don’t drift.

Tax and invoicing: marketplace responsibilities aren’t optional

In several countries the platform is the deemed supplier for VAT/GST on B2C food. Elsewhere the restaurant stays the supplier. Either way:

  • Emit compliant e-receipts to the customer (tax inclusive, line-level rates), and settlement statements to the merchant showing tax lines and responsible party.
  • Separate your platform fee (with its own tax treatment) from the menu value.
  • For groceries/quick commerce, mixed tax baskets are normal (reduced rates for staples, standard for others). Don’t flatten; calculate line by line.

Fraud patterns that actually bite—and the counters that work

  • Triangulation and card testing. Block BIN/geo mismatches, velocity on new devices, and bulk small tickets at odd hours.
  • First-order abuse. Heavier KYC on “first high ticket + new address + no prior history.”
  • Collusion. Merchant–courier splits on “fake delivered” leave GPS and motion tells; weight the model with those signals.
  • Refund farming. Require RMA-style photos for repeated “missing item” claims; throttle goodwill per account.

Run experiments; let ops see the business impact of each toggle instead of arguing hypotheticals.

FX and treasury: volatility hides in the tail

Cross-currency shows up in travel hubs, expat neighborhoods, and corporate accounts.

  • Natural hedge. Spend what you collect—pay local couriers, local ad buys, and lease dark stores in the same currency.
  • Coverage. Roll short-dated forwards on predictable net outflows; don’t push 100% through one roll date.
  • Metadata. Stamp the exchange source and timestamp on every capture, refund, and payout. When rates move over a weekend, finance can explain the pennies.

Reconciliation without archaeology

You don’t need more people; you need better shims.

  • Identifiers. One order ID everywhere—checkout, dispatch, courier app, merchant POS, bank file.
  • Bank data. Ingest ISO 20022 statements; they preserve remittance details your parser can trust.
  • Auto-match. Aim for 98% by count, 95% by value. Bucket the rest: FX drift, acquirer fee mismatch, partial pay, duplicate prevention, missing remittance. Owners + SLAs, not “misc.”

What to watch if you care about margin

  • Approval rate by corridor and cost per successful order (rail + acquirer + FX + ops minutes).
  • Time-to-refund (median) and refund reason mix.
  • Merchant statement disputes per 1,000 orders and average time to close.
  • Courier on-time payout and failure rate (name mismatch, closed account).
  • Tip attach rate and instant-tip take-up (with unit economics).
  • FX bps on GMV and on courier/merchant payouts; covered vs uncovered ratio.
  • Auto-recon rate and exception aging.

When to bring in a payments intermediary

If your geography map looks like a patchwork—euro area, UK, Brazil, MENA—and you’d rather not stitch a dozen bank connections, a specialist like Collect&Pay can provide multi-currency accounts, virtual IBANs per merchant cohort, and local instant rails for both refunds and courier payouts, with line-item fee/FX breakdowns your finance team can defend. Choose on corridor breadth, failure handling, and reporting clarity—not on headline price alone.

Leave a Comment