In-House Fleet vs Third-Party Couriers

In-house fleet means the operator recruits, schedules and pays the people who make the deliveries. Third-party couriers means handing each drop to a network — DoorDash Drive, Uber Direct, Shipday or Nash — that supplies the rider and charges per delivery. The choice is usually framed as cost against control, and framed that way it is decided badly, because the two options fail at opposite ends of the same curve.

There are three arrangements, not two

Your own fleet. You hold supply. You also hold the shifts nobody wants, the equipment, the classification question and the wet Tuesday when demand does not arrive but the riders did.

Courier networks only. You hold no supply. Every drop is bought at a per-delivery price, capacity appears without a hiring plan, and you can open a city on a Monday.

Both, with rules. Your own riders cover the hours and zones you can fill predictably; a network absorbs the peaks, the outliers and the postcodes at the edge of the map. This is where most operators end up, and it is worth reaching deliberately rather than by accumulation.

What each actually costs

The comparison people make is per-drop price. That is the smallest part of it.

A fleet costs money when it is idle, which is most of the time in a business with two sharp peaks a day. The cost that matters is not the delivery fee but the paid hour with no order in it — and utilisation, not headcount, is the number to manage.

A network costs money when you are busy, which is exactly when you are also most price-sensitive. Per-drop pricing is steady; your margin is not, and a surge on a Friday lands on the orders you most wanted to be profitable.

A fleet costs management. Recruitment, scheduling, onboarding, equipment, disputes, and a rota that somebody owns. This is a real operating function and it does not appear in a per-drop comparison at all.

A network costs you the last mile of the experience. The rider is not yours. Their behaviour at the door is your brand and not your process, and when a delivery goes wrong the customer blames you while the resolution sits with somebody else’s support team.

Why the hybrid is the default rather than the compromise

Demand is spiky and supply is not. That single mismatch is what makes the pure versions uncomfortable.

Size a fleet for your peak and you pay for idle riders all week. Size it for the average and you fail every Friday. There is no headcount that solves both, which is why the sensible shape is a base plus overflow: own the predictable floor, buy the volatile ceiling.

That also gives you the thing neither pure model has — a fallback. A fleet with no network has nothing when six riders call in sick; a network-only operation has nothing when the network prices a surge you cannot pass on. Running both means each is the other’s contingency, and the routing rule that decides which one gets a given order is a real piece of configuration, not an afterthought. The scoring that sits behind it is set out in how auto-dispatch scores a driver.

What to work out before choosing

1. What share of your week is genuinely predictable? That share is the honest size of a fleet. Everything above it wants to be bought.
2. What does a network cost you at peak, not at list? Get the surge behaviour in writing before you model anything.
3. Who employs the rider, and under what classification where you operate? This is a legal question with different answers by country, and it is the one that turns a fleet from an operating decision into a structural one.
4. What happens to a failed delivery in each model? Specifically: who refunds, who investigates, and how long the customer waits. This is the difference customers actually feel.
5. Can your platform route to both, per order, on rules you control? If switching between them is a migration rather than a setting, you do not really have the hybrid available.

When each one is right

Start on networks when you are opening a city, testing a vertical, or cannot yet predict a week. Capacity you can turn off is worth paying a premium for while you are still learning the shape of demand, and the alternative — hiring against a forecast you do not have — is the more expensive mistake.

Build a fleet once a floor of demand is genuinely reliable, the density is there to batch drops, and the experience at the door has become something you compete on. Groceries and pharmacy tend to reach this sooner than restaurants, because the handover matters more.

Run both in any city that has moved past the first phase — which, in practice, is nearly all of them.

The one position not worth defending is a fleet built for pride rather than arithmetic. Owning riders is not a strategic asset in itself; it is an asset when utilisation is high and the door experience is part of why customers choose you. Where neither is true yet, a network is not a compromise — it is the cheaper way to be uncertain.

For the sequence this sits inside when opening a new market, see launching a regional aggregator city by city; for how the delivery line interacts with the rest of the revenue model, marketplace business models, explained.