Delivery stopped being a side project for food businesses a while ago. It is now the channel most of your orders travel through, and the one that quietly decides whether a busy week made money.
That shift is what makes food service industry delivery its own operating problem. You are not just cooking, baking, or packing anymore. You are running a small transport operation with temperature rules, labor costs, and a cost per stop that nobody printed on a menu.
This guide covers the whole picture for a local operator: which parts of the industry deliver, what each delivery model really costs per order, the food safety rules that apply once a box leaves your building, and the routing decisions that move your cost per delivery up or down.
The Bottom Line
Total U.S. restaurant and foodservice sales are projected at $1.55 trillion for 2026, and nearly 75% of restaurant traffic now happens off-premises (National Restaurant Association, 2026).
Third-party apps advertise 15% to 30% commission, but operators who add packaging, processing, promotions, and refunds usually land at an effective 30% to 40% per order.
Your own drivers cost roughly $1,920 to $2,400 a month each at $12 to $15 an hour, plus $300 to $1,000 a month for a leased vehicle, so the model only pays off when you have enough stops to keep them busy.
Once food leaves the kitchen, the 41°F to 135°F danger zone and the four-hour cumulative clock still apply, and staging time waiting for a driver counts against that clock.
Route density, not driver speed, is what sets your cost per delivery. One food distributor cut cost per case from $2.40 to $1.92 by re-planning routes (Paragon Routing).
Save 80% of delivery management time
We handle everything:
- Dedicated operations manager
- Real-time tracking dashboard
- Automated customer notifications
- Urgent issue resolution
Which parts of the food service industry deliver?
Almost all of them, and that is why generic delivery advice fits so badly. The industry splits into commercial operations that sell food for profit and non-commercial ones that feed a captive group: school cafeterias, hospitals, corporate dining, senior living. The commercial side is the larger share of revenue, and it is where most local delivery volume sits.
Inside the commercial side, the delivery job changes completely depending on what you sell:
Restaurants and cafes send single hot orders to homes, usually within a few miles, with a quality clock measured in minutes.
Caterers send large, fragile, time-locked drops to offices and events, where arriving early is nearly as bad as arriving late.
Bakeries and coffee roasters run standing wholesale routes, dropping crates at the same cafes and grocers before those places open.
Wholesalers, farms, and food product brands move cases and pallets to loading docks on fixed days, against purchase orders and signed invoices.
Meal prep and subscription kitchens send batched residential deliveries on a weekly cycle, cold, in insulated packaging.
Food services now account for 51.5% of all employment in the U.S. food system, up from 47.6% in 2010 (USDA Economic Research Service, retrieved 2026-09-26). More of the food economy sits in preparing and serving than ever, which means more of it has a delivery leg attached.
How big the off-premises channel has become
Off-premises is no longer the overflow channel. The National Restaurant Association projects total restaurant and foodservice sales of $1.55 trillion in 2026, with industry employment reaching 15.8 million people (National Restaurant Association, 2026). Nearly 75% of all restaurant traffic now happens off-premises, and 58% of limited-service operators plus 41% of full-service operators say off-premises is a bigger share of sales than it was in 2019.
Read that as a planning number rather than a headline. If three out of four transactions leave your building, then delivery is not a marketing channel you can bolt on. It is a production line that needs its own packaging, staging space, labor plan, and cost tracking.
The three ways food businesses get orders delivered
There are only three real models, and most operators end up mixing two of them. The differences that matter are who owns the customer relationship, who carries the fixed cost, and how the price scales when volume moves.
| Third-party apps | Your own drivers | Courier or delivery partner | |
|---|---|---|---|
| Cost shape | Percentage of each order | Fixed monthly, plus fuel | Per delivery or per route |
| Cost when volume drops | Falls with volume | Stays, whether you deliver or not | Falls with volume |
| Who owns the customer data | The platform | You | You |
| Branding and handoff control | Limited | Full | Partial to full |
| Good fit for | Discovery, new demand, unpredictable volume | Dense routes, daily repeat stops | Wholesale routes, catering, spiky weeks |
| Main risk | Margin compression | Idle drivers and vehicles | Vendor dependence |
The honest summary from operators who have run the comparison: third-party platforms are easier to scale up and down, while in-house delivery gives more control and can cost less when delivery density is high (Lightspeed, retrieved 2026-09-26). Density is the deciding variable, and it comes up again further down.
What third-party delivery apps really cost per order
More than the advertised rate. Published commission tiers in 2026 run 15%, 25%, or 30% on DoorDash, 20%, 25%, or 30% on Uber Eats, and a 15% to 25% marketing commission plus roughly 10% delivery commission on Grubhub, with separate order-processing fees on top (Rezku, 2026).
Once packaging, payment processing, promotional discounts, and customer refunds are added, the effective rate operators actually pay commonly lands between 30% and 40% of the order (OPA!, 2026). Set that against restaurant profit margins that typically sit in the 5% to 15% range and the arithmetic gets uncomfortable fast: on a $50 order, a 30% take is $15 before you have paid for food, labor, or the box it went in.
That does not make the apps a mistake. They buy you discovery from people who have never heard of you, and they absorb volume swings you could not staff for. What they are bad at is carrying your repeat business. A customer who orders from you every Friday is the most expensive possible order to keep paying commission on, and they are the first group worth moving to a channel you control.
What it costs to run your own delivery drivers
In-house delivery converts a variable cost into a fixed one, which is either a gift or a trap depending on your volume.
The typical components, based on figures compiled for small operators (Expert Market, retrieved 2026-09-26):
Vehicle: roughly $300 to $1,000 a month to lease and maintain, or about $8,000 to $30,000 to buy outright.
Fuel: roughly $100 to $300 per vehicle per month, depending on volume and distance.
Driver wages: roughly $12 to $15 an hour, which is about $1,920 to $2,400 per driver per month at 40 hours.
Insurance: commercial auto coverage, which is a separate policy from your general liability.
Do the division before you hire. A driver costing $2,200 a month who completes 300 deliveries is a $7.33 delivery. The same driver completing 90 deliveries is a $24.44 delivery, and no packaging saving rescues that number. This is why in-house delivery works beautifully for a bakery with 40 wholesale stops on a Tuesday morning route and badly for a restaurant with nine scattered dinner orders a night.
The upside is real when the volume is there. Your own driver is trained on your product, wears your name, and handles the customer conversation when something is wrong. That handoff is part of the product for catering and wholesale in a way it never is for a takeout order.
Why route density sets your cost per delivery
The final leg is where the money goes. Last-mile delivery now accounts for about 53% of total shipping costs, up from roughly 41% in 2018 (Net Zero Insights, 2025). Labor alone is around half of last-mile expense, with fuel adding another 10% to 25% (GoBolt, retrieved 2026-09-26).
Notice what is not on that list: your food cost. The lever that moves last-mile cost is how many stops a driver can complete in an hour, and that is a planning problem, not an effort problem. Stops clustered in one neighborhood are cheap. The same number of stops spread across the metro in random order can cost twice as much to serve with the same driver and the same van.
Re-planning routes is the cheapest available fix. A fresh produce processor that moved to route optimization software cut its cost per case from $2.40 to $1.92, a 20% reduction, while increasing weekly routes per truck from 75 to 110 and running 10 fewer trucks (Paragon Routing, retrieved 2026-09-26). Reported gains for food and grocery delivery generally include delivery time cuts of up to 35% and fuel reductions of 15% to 20% (Locus, 2026).
Three practical moves come out of this:
Group orders by geography, not by order time. A delivery window of two hours instead of thirty minutes lets you build a real route.
Set minimums by zone, not one flat minimum. Distant stops need a bigger basket to break even, and your pricing should say so.
Give repeat wholesale stops a fixed day. Predictable stops are the ones that make a route dense.
This is also the practical case for using a courier network instead of buying vans. Metrobi is a local delivery platform built for food, floral, catering, and wholesale businesses, with multi-stop route optimization, real-time tracking, and proof-of-delivery photos, and it covers single-stop jobs as readily as fifteen-stop routes across major U.S. metros. For a business with a dense Tuesday and a quiet Thursday, paying per delivery instead of per month is often the difference between a profitable route and an idle van.
Food safety rules that still apply once the order leaves your kitchen
Your health inspector’s authority does not stop at the door. Under the FDA Food Code, cold time/temperature control for safety (TCS) foods should be received at 41°F or below and hot TCS foods at 135°F or above, with frozen product arriving frozen solid and showing no signs of thawing and refreezing (GoFoodService, retrieved 2026-09-26).
The range between 41°F and 135°F is the danger zone, where common pathogens can double roughly every 20 minutes. TCS food cannot sit in that range for a cumulative total of more than four hours before it has to be discarded, and that clock does not reset when the food changes hands (ConnectedFresh, retrieved 2026-09-26). Prep time, staging time while a driver is en route to you, and transit time all count against the same four hours.
What that means for a delivery operation:
Treat staging as transit time. A tray sitting on a pass waiting for a pickup is already on the clock.
Pack hot and cold separately. Insulated bags do not fight each other’s contents, and a shared bag drags both items toward the middle of the danger zone.
Log temperatures at handoff, not just at prep. A reading taken when the order goes out is the record that protects you in a complaint.
Use packaging that survives a stop and a curb. Tamper-evident seals also settle the “was this opened” argument before it starts.
Schedule inbound deliveries into slow periods so someone has time to check temperatures on receiving.
Licensing, insurance, and staffing for food delivery
Requirements vary by city and state, but the recurring list for an operator adding delivery is a business license, a food service permit, food handler certification for the people touching the food, and commercial auto insurance on any vehicle used for deliveries (Restaurantware, retrieved 2026-09-26). Personal auto policies routinely exclude commercial use, which is the gap that surprises people after an accident rather than before one.
On the staffing side, the delivery role is a customer-facing job that happens to involve driving. Operators who run it well train for food handling and for the doorstep conversation, run background checks, and confirm insurance on any vehicle used for the job. If you use a third-party network instead, the equivalent diligence is checking what proof of delivery you get, how issues are escalated, and whether you can keep working with drivers who already know your route.
How wholesale and catering delivery differs from restaurant takeout
This is the split most delivery advice ignores, and it matters if you sell to other businesses.
Takeout delivery is one hot item, one address, thirty minutes, and a consumer who watches a map. Wholesale and catering delivery is a scheduled drop against a purchase order, often before the receiving business opens, with a dock or a back door, a case count to verify, and an invoice or signature that has to come back. Speed is rarely the binding constraint. What matters is arriving inside the window, with the count right and the paperwork in hand.
Practical differences worth building around:
Delivery windows beat delivery speed. A caterer arriving 40 minutes early has created a problem, not delighted anyone.
Proof of delivery is the payment trail. Photos, signed invoices, or supporting documents close out the order and shorten disputes about what arrived.
Standing routes deserve standing days. Fixed weekly drops are the densest, cheapest deliveries you will ever run.
Someone has to be there. Access details, dock hours, and a named contact belong on the delivery record, not in a driver’s memory.
Setting up delivery: a short sequence that works
Count what you already deliver, by channel, for one month. Orders, revenue, and fees paid, separated into app orders and direct orders.
Work out your true cost per delivery in each channel. For apps, use the effective rate, not the headline commission. For in-house, divide total monthly driver and vehicle cost by actual completed deliveries.
Map your last 200 deliveries. Clusters tell you where in-house or a fixed route makes sense; scatter tells you where a per-delivery model is cheaper.
Price delivery by zone and set minimums that match your cost. A single flat fee across a whole metro subsidizes your least profitable orders.
Fix the temperature and packaging chain before you scale volume. Staging discipline, separate hot and cold packing, and a temperature log.
Move your repeat customers to a channel you own. Commission on a weekly regular is the most expensive line in the whole operation.
Re-plan routes monthly. Customers move, days shift, and a route built in March is rarely still the best route in September.
Frequently asked questions
How much do delivery apps charge restaurants?
Published commissions run about 15% to 30% depending on the platform and plan tier, with separate order-processing fees (Rezku, 2026). Operators who include packaging, promotions, processing, and refunds typically calculate an effective cost of 30% to 40% per order.
Is it cheaper to deliver with your own drivers?
It depends almost entirely on stop density. A driver costing around $2,200 a month is cheap across 300 deliveries and expensive across 90. Dense, repeating routes favor your own drivers; scattered, unpredictable orders usually favor paying per delivery.
What temperature does food need to be at during delivery?
Cold TCS food should be at or below 41°F and hot TCS food at or above 135°F, with frozen items still frozen solid on arrival. Time in the 41°F to 135°F range is capped at four cumulative hours, including staging and transit (ConnectedFresh, retrieved 2026-09-26).
Do I need special insurance to deliver food?
Commercial auto coverage is normally required for vehicles used in deliveries, because personal policies often exclude commercial use (Restaurantware, retrieved 2026-09-26). Requirements for permits and food handler certification vary by jurisdiction, so confirm locally.
How far should a local food business deliver?
Far enough to keep routes dense and short enough to keep food in spec. Rather than a single radius, set zones with their own minimum order and fee, and let the numbers from your own delivery history tell you where the profitable edge sits.
Where to start
The food service industry has already moved most of its volume off-premises, so the question for a local operator is no longer whether to deliver. It is which model you use for which orders, and whether you know your cost per delivery well enough to tell the difference.
Start with the month of data you already have. Separate the channels, calculate the real cost per delivery in each, and look at where your stops sit on a map. Most operators find the same two things: they are paying commission on customers who would have ordered anyway, and their routes are built around order times instead of geography. Both are fixable this quarter, and both show up directly in margin.
