Choosing a Payment Gateway for Delivery Orders: 3 Critical Factors

Most advice about choosing a payment gateway assumes the customer is standing at a counter, or clicking “buy” on a box that’s already packed. Neither is true for you. Your customer pays on Tuesday night for a Thursday morning drop, the florist swaps two stems that were meant to be in the arrangement, the wholesale account wants an invoice instead of a card, and a driver ends up holding a phone at somebody’s back door.

That gap is the whole problem. A payment gateway for delivery orders gets judged on what happens in the hours between checkout and handoff, not on what happens at checkout. The gateway that wins a generic comparison chart can be the wrong one for you on exactly that point.

This guide covers the three factors that actually decide it: when the money moves, what the money costs, and what the gateway connects to. The mechanics underneath it are covered in the guide to payment APIs for order fulfillment, and if a chunk of your customers order the same thing every week, recurring billing for delivery routes is a separate decision that runs alongside this one.

The Bottom Line

• A gateway is the piece that captures card details and asks the bank for approval. It isn’t your processor, and it isn’t your POS. Comparing the wrong layer is the most common mistake here.
• The single most important question for delivery is whether the gateway separates authorization from capture. If it charges at checkout, every substitution and every weight change becomes a refund.
• Authorization holds expire. The card network maximum sits around 144 hours, roughly six days, with Visa typically allowing capture within seven (Solidgate, retrieved September 2026). A delivery slot booked for next week can outlive the hold.
• Headline rates hide the real number. Stripe’s US online benchmark is 2.9% + 30 cents, PayPal Checkout runs 3.49% + 49 cents, and Square is 2.6% + 10 cents in person (Dodo Payments, retrieved September 2026). Your channel mix decides which of those you actually pay.
• On wholesale invoices, ACH is the entire argument: roughly $0.50 to $1.50 on a $10,000 invoice against about $250 on a card (Resolve, retrieved September 2026).

What a Payment Gateway Does, and What It Doesn’t

The word gets used for three different things, and picking the wrong layer wastes weeks.

The gateway captures the payment details and asks for authorization. The processor moves funds between the banks. The merchant account is where the money lands. Plenty of companies sell all three as one product, which is why the distinction feels academic right up until you try to switch one of them.

What matters for a delivery business is that the gateway is the layer holding the decision you care about: whether a charge is approved now, held now and taken later, adjusted, or released. Everything below hangs off that.

Two clarifications before the factors. A gateway is not the same thing as a payment processor, and the difference shows up on your statement rather than in your checkout. And a hosted gateway, where the customer jumps to the provider’s page to type a card, isn’t automatically worse than an integrated one. It’s a trade of checkout polish against how much PCI scope you want to carry.

Factor 1: Whether the Gateway Splits Authorization From Capture

This is the one that separates a gateway built for retail from one you can run a delivery operation on.

Authorization asks the bank to approve and reserve an amount. Capture is the instruction that actually takes it. A retail gateway usually does both in the same breath, because in a shop the goods leave with the customer. In a delivery business, hours or days pass between those two moments, and things change inside that gap.

Three things change constantly:

• The order changes. A substitution, a short pick, a case that turned out to be 14 pounds and not 15. If you charged at checkout, correcting this means a refund, which costs you the processing fee on the original amount at most providers and shows up on the customer’s statement as two transactions.
• A tip gets added at the door. If the gateway can’t increment an existing authorization, the tip becomes a second, separate charge against the same card, and your deposit stops reconciling cleanly against your order list.
• The delivery fails or gets rescheduled. Nobody should be charged for a box still sitting in your van.

So the question to put to any provider is blunt: can I authorize at checkout and capture a different, final amount at fulfillment, and how long can I wait? The answer is time-limited. The network maximum is about 144 hours, and Solidgate recommends settling inside 120 hours to stay clear of trouble, with Visa typically expecting capture within seven days (Solidgate, retrieved September 2026). Delay past that and you face higher interchange, a declined capture, or an expired authorization that leaves you chasing a customer who thought they’d paid.

If you sell delivery slots more than a week out, auth-and-capture alone won’t carry you. You need either a stored, reusable payment method charged on the delivery day, or a deposit now and a balance later. Both are gateway capabilities to confirm before you sign, not after.

Factor 2: The Fee Structure You’ll Actually Pay

Published rates are written for whichever channel makes the provider look best. A delivery business usually runs three channels at once, and the rate is different in each.

Those card figures come from a 2026 comparison of Stripe, PayPal, Square and others (Dodo Payments, retrieved September 2026), where PayPal Checkout sits higher than both at 3.49% + 49 cents.

The ACH gap is the one worth doing arithmetic on. On a $10,000 invoice, ACH runs roughly $0.50 to $1.50 against about $250 for a card at 2.5% (Resolve, retrieved September 2026). If a meaningful share of your revenue arrives as monthly wholesale invoices, a gateway with weak ACH support is expensive in a way no percentage-point difference on card rates will offset.

Three costs that never appear in the comparison table:

• Refunds. Many providers keep the original processing fee when you refund. If your substitution rate is high and you charge at checkout, you pay twice on a slice of orders.
• Chargebacks. A flat fee per dispute, typically $15 to $25, plus the disputed amount. Delivery generates a specific kind of dispute, the “it never arrived” claim, which proof of delivery defends well and a bare card receipt defends badly.
• Monthly and per-transaction gateway fees charged on top of the processor’s rate. This is how a 2.9% quote becomes 3.2% in practice.

Factor 3: What the Gateway Connects To

A gateway with a perfect rate that doesn’t talk to anything else will cost you more in labor than it saves in fees.

The connections worth checking, roughly in order of how much time they save:

• Your online store or ordering page, natively rather than through a plugin somebody else maintains.
• Your accounting software, at the payout level. One bank deposit covering forty orders minus fees is what has to reconcile, not forty separate transactions.
• Your in-person reader, on the same merchant account, so a card taken at the door lands against the order and not against a blank amount.
• Your order and delivery system, so a captured payment marks the order paid and a failed capture flags before the box goes on a van.

Where a native connection doesn’t exist, the fallback is the API, and how good that API is becomes a real evaluation criterion rather than a developer footnote. What to look for is covered in the payment APIs guide.

One warning about hosted checkout pages. They’re the fastest way to reduce your PCI burden, and for a small operation that’s often the right trade. But some hosted pages can’t be told about a delivery fee calculated after the address is entered, and some can’t collect a delivery window or a gate code. Check that your checkout can still ask the questions a delivery order needs before you take the compliance shortcut.

How the Choice Changes by Type of Delivery Business

The same three factors land differently depending on what’s in the box.

A bakery or florist taking retail orders for local drops should weight authorization flexibility highest. Substitutions are routine, order values are low enough that the fixed 30 cents matters, and tips at the door are common.

A caterer taking large orders weeks out should weight stored payment methods and deposit handling highest. Nothing about a seven-day authorization window helps when the event is six weeks away. You need a deposit captured now and a balance charged against a saved card later.

A wholesaler delivering on standing routes should weight ACH and invoicing highest, and should read the recurring billing guide next, because for that model the gateway decision and the repeat-collection decision are really one decision.

Questions to Ask a Payment Gateway Provider Before You Sign

Send these to anyone you’re seriously considering. The answers sort the field fast.

• Can I authorize one amount and capture a lower or higher one, and what’s the maximum delay before the authorization expires?
• Can I increment an authorization to add a tip or a weight adjustment at the door?
• Do you keep the processing fee when I refund?
• What does ACH cost, and how long does it take to settle?
• Can I store a customer’s card and charge it on a scheduled delivery day without them re-entering it?
• Does your in-person reader use the same merchant account and the same payout as my online payments?
• What’s the dispute fee, and what evidence format do you accept as proof of delivery?

Any provider that can’t answer the first two clearly is built for a shop, not for a delivery operation.

Making the Decision

Score your shortlist on the three factors in order. Timing first, because a gateway that can’t hold and adjust a charge generates manual work forever. Cost second, weighted by your real channel mix rather than the headline online rate. Connections third, because integration gaps hurt but are usually solvable.

If two candidates tie, pick the one whose in-person hardware and online payments share an account. That single property removes more end-of-month reconciliation than any rate difference you’re likely to negotiate.