Own Delivery Fleet vs Third-Party — When Does It Actually Pay Off for a Restaurant?

Operators ask the wrong question about own delivery. It is not "should I add a driver?" — it is "which subset of my current third-party orders would be cheaper to do myself?" The two questions have very different answers, because not every delivery is equally expensive on a marketplace.

A short delivery to a frequent customer with a €40 ticket costs the restaurant ~€11 in commission. A long delivery to a one-time customer with an €18 ticket costs ~€5. The two have the same driver cost on your own fleet, but very different savings. Targeting the first type — high-margin, short-distance, repeat customers — is where own delivery economics work; trying to take on the second type is where they break.

The break-even math (worked example)

A part-time delivery driver in EU markets costs roughly €8–€12/hour fully loaded (wage, taxes, fuel allowance, basic insurance). Call it €10/hour for a clean number. A reasonable driver does 4–6 deliveries per hour at urban density, so the per-delivery cost is €1.70–€2.50.

Commission saved per delivery, on a €35 average order at 28% commission: €9.80. So each in-house delivery saves roughly €7–€8 vs the same delivery on a marketplace. A driver doing 5 deliveries/hour for 4 hours per shift saves the restaurant €140–€160 per shift — comfortably more than the €40 shift cost.

The catch: that math only works if the driver is actually doing 5 deliveries/hour. At 2 deliveries/hour (a slow Tuesday), the savings collapse and the shift loses money. Volume — not commission rate — is the real lever.

Three pitfalls that wreck the economics

Idle hours. The biggest cost in delivery is paying a driver to sit. Most restaurants make the mistake of staffing the driver for the whole shift, when they only need them during the 6–9pm peak. Solution: hire on-call/flexible shifts that match the order-volume curve, not a fixed 4-hour block.

No dispatch system. A driver running deliveries from sticky notes loses 30–40% of their potential throughput. The first investment after the driver themselves should be a dispatch screen that auto-batches nearby orders, shows the optimal route, and reassigns when a driver finishes early. Without this, the per-hour delivery count never gets above 3.

Trying to compete on radius. Don't. Marketplaces win the 6+ km radius because they aggregate driver supply across many restaurants. The own-delivery sweet spot is 0–4 km — the urban core where your loyal customers live anyway. Cap your own-delivery zone there and route longer orders through the marketplace.

A realistic phased plan

• •Month 1: Add own delivery as an option in the customer's checkout for the inner zone (0–4 km) only. Keep marketplaces enabled for outer zone. Use existing kitchen staff as drivers during quiet shifts to validate the workflow without a fixed cost.
• •Month 2: If inner-zone direct volume hits 15+ orders per shift, hire one on-call driver scheduled to the actual order-volume peak. Switch marketplace settings so they don't accept inner-zone orders.
• •Month 3: Add dispatch — even a basic interface that batches nearby tickets cuts per-delivery time by 20%. Start tracking per-driver per-hour delivery count as the headline KPI.
• •Month 4–6: If volume justifies it, add a second driver for the weekend peak. By month 6, the venue should be doing 50–70% of its delivery volume in-house, with the remainder routed through marketplaces for the long-tail addresses where their economics still win.

When own delivery doesn't make sense

Three honest signals that you should skip this entirely: total delivery volume under 30 orders/week (you'll never amortise the dispatch overhead), average order value under €15 (commission math doesn't favour you), or you don't have a kitchen team with bandwidth to absorb the operational learning curve of dispatch + driver management.

See the Own Delivery feature page for the dispatch screen, driver app, and zone configuration. The Delivery Zones page covers how to draw your radius. The delivery commission calculator from earlier this month is the easiest way to test the math against your specific volume.