Third-party delivery apps solved a real problem. They put your restaurant in front of people who'd never have found you otherwise, and paying a commission for that discovery is a legitimate trade. The frustrating part isn't paying for genuine new customers. It's paying a commission on customers who wanted to order from you directly, tried to call, couldn't get through, and opened an app instead. That's a self-inflicted tax, and it's the specific leak voice AI is actually positioned to close.
Let's separate the commission you should accept from the one you shouldn't, and be honest about what answering the phone can and can't do about it.
The commission math, framed honestly
You'll see a lot of confident, specific numbers about what delivery marketplaces cost. The truth is that it varies by platform, by plan tier, by market, and by which add-on fees apply. As a rough frame, commissions on marketplace orders commonly sit somewhere in the 15% to 30% range per order, and the all-in cost can run higher once processing and service fees stack on top. Treat that band as an approximation rather than a quote.
The number that actually matters is your own, and it's on your statements. Pull what you paid in commissions and fees last month, divide by the orders that flowed through those channels, and you have your real per-order cost. Hold that figure next to the same order taken directly over the phone, where the marginal cost is close to zero, and the gap is the thing worth managing. On thin restaurant margins, that gap is often the difference between a profitable order and a barely-breakeven one.
It's also worth doing this per platform rather than in aggregate, because the mix usually isn't even. Groups that run three marketplaces frequently find one of them costs meaningfully more per dollar of revenue than the others, and that finding changes what you do about promotion and menu placement long before it changes anything about your phone.
The reroute problem: a missed call becomes an app order
Here's the mechanism that quietly moves customers off your direct channel. A caller wants to order from you. They pick up the phone during your Friday rush, the exact moment your team has the least slack to answer it. Two rings, four rings, voicemail. So they do the thing that reliably works: they open a delivery app and order from you there.
You didn't lose the sale. You converted a commission-free phone customer into a commission-charging app customer, on that order and often on every future one, because the app is now their default way to reach you. We walk through this in detail in the real cost of a missed call, and the reroute is the most expensive part. A missed call isn't one lost order. It's a relationship moved onto a channel that taxes it from then on.
Sizing your own reroute rate
You can estimate this rather than guess at it, and you should before spending anything.
Pull two weeks of call logs and two weeks of app order timestamps and lay them side by side. You're looking for a pattern: an unanswered call, then an app order from the same neighborhood within the next ten or fifteen minutes. You won't be able to match individual customers without detail you don't have, but the correlation during your peak hour is usually clear enough to be useful. Count the pairs, multiply by your average app commission per order, and you have a defensible monthly figure for what the reroute costs you.
Do that arithmetic before you read another vendor page. If the number lands below a monthly plan, the phone isn't your leak and you've saved yourself a subscription. The single-location economics breakdown works through the same comparison in more detail, including the costs people forget to put on the other side of the ledger.
The channel you actually own
A direct phone order is the only ordering channel you fully control. You keep the full ticket minus your ordinary payment processing. You keep the customer's phone number and the relationship. You set the experience end to end. Nothing about that order is intermediated by a platform that can change its rates, bury you in search results, or own the customer data.
That's what makes the reroute sting. Every caller who defaults to an app because they couldn't reach you is a customer you had on your best channel and handed to your most expensive one, for no reason other than a phone that didn't get picked up. Protecting the direct channel isn't anti-app. It's refusing to give away margin you didn't have to.
One thing to settle while you're here: whether your phone prices match your app prices. Most operators mark up on the marketplaces to absorb commission, which is defensible, but price parity across channels is a decision to make deliberately rather than discover when a caller reads your app listing back to you mid-order.
Where voice AI fits: answer before the app does
This is the narrow, honest job a voice agent does here. It answers every call, including the ones that come in during your peak rush and after you've closed, the two windows where reroutes concentrate most. The caller who would have hit voicemail and opened an app instead gets their order taken, priced against your live menu, paid for over the phone, and confirmed by text. They never had a reason to switch channels, so they didn't.
That's the entire play: catch the direct order before the missed call pushes it to a marketplace. It's not a discount war with the apps or a clever pricing scheme. It's making sure "call the restaurant" reliably works, so it stays the path of least resistance for the customers who'd have taken it anyway.
What it doesn't solve
Be clear-eyed about the limits. A voice agent does nothing for genuine discovery. The customer who finds you by browsing an app for the first time was never going to call, and you're not saving a commission you'd never have avoided. The apps still earn their cut on that. It also can't recover customers who prefer app ordering for its own sake, for the tracking, the one-tap reorder, and the delivery logistics you may not run yourself. And it won't help if your phone number is hard to find or your direct ordering experience is worse than the app's.
Delivery commissions aren't all the same. Paying for real discovery is a fair trade. Paying commission on a customer who tried to call and couldn't get through is money you gave away. So check your statements for what a marketplace order actually costs you, count the reroute pairs in two weeks of logs, and let that number decide. If it's small, fix something else. If you want to talk through what you find, get in touch, but start with your own data rather than anyone's headline percentage.