FlowsBook a callSign inCreate free accountSign up
Workflows by industry

How delivery-only brands choose a host restaurant kitchen

A ghost kitchen location does not have to be a new unit. How virtual brands pick partner restaurants with room to cook, close to orders and away from rivals.

Part of the guide Workflow automation

By the xMap team · · 5 min read

For many delivery-only brands, the best ghost kitchen location is a restaurant that already exists. Instead of fitting out a unit, the brand licenses its menu to a well-run independent kitchen that has spare hours, sits close to people who order in, and does not cook the same food as everyone around it. Choosing that host comes down to three questions: who can reach whom, how crowded the cuisine already is, and whether the kitchen is good enough to carry your name.

This guide is for operators of virtual restaurant brands and delivery platforms that match brands with kitchens. It covers what makes a good host, how to measure it around each candidate, a worked example in London, the traps, and how to have an AI employee keep the shortlist current. If you are looking for an empty unit to rent instead, the companion article on site selection for a ghost kitchen covers that path.

What a good host kitchen looks like

  • Independent. Chains rarely let an outside brand into their kitchens; owner-operators decide on the spot.
  • Well rated. Customers will judge your brand by food cooked there. A strong rating with many reviews is the best public proof of a steady kitchen.
  • A different cuisine. A pizzeria can cook burgers; it is harder for it to cook a rival pizza menu without confusing staff and suppliers.
  • In the middle of demand. Delivery radius is short. A kitchen at the edge of town wastes half its reach on parks and warehouses.

Measuring the area around each candidate

Start from the delivery reach rather than a circle. The Travel area block draws the area a rider can cover from a point in a set time or distance by road, by car, bike or on foot. A 10-minute ride from a kitchen on a river or a ring road covers far less than a circle suggests.

Inside that reach, count what matters with Nearby places: restaurants of your cuisine (your rivals for the same order), offices and campuses (lunch demand), and how many restaurants there are in all (a sign of a food street where people already order). Where xMap has population figures, the Demographics block adds the residents inside the reach and their age groups. Population data covers some countries, among them Canada, France, India, Thailand and Australia, but not the United Kingdom or the United States, so in those markets lean on counts of places instead.

A worked example in London

London shows why the host matters more than the unit. xMap's places database lists 3,792 restaurants within 10 km of the centre of London (November 2026). Of the 326 nearest the centre, 276 are independent and 50 belong to a chain; 278 have a rating, 4.4 out of 5 on average. A brand looking for hosts is not short of kitchens: it is short of the right ones.

3,792Restaurants listed
276Independent, of the 326 nearest
4.4Average rating
Restaurants within 10 km of the centre of London, xMap's places database, November 2026; the independent count and the rating are of the 326 nearest the centre.

Say a made-up Korean fried chicken brand wants three hosts in inner London. The steps look like this:

  1. Search restaurants in the boroughs the brand delivers to, keeping independents rated 4.3 or more with at least 100 reviews.
  2. Drop any that already serve chicken or Korean food, using the category and a short xMap LLM check of the name.
  3. Draw a 10-minute cycling reach around each remaining kitchen and count other fried chicken places and offices inside it.
  4. Sort by fewest direct rivals and most offices, and keep the top ten for calls.

The figures in the steps are rules the brand chose, not facts about London. A different brand would pick a different rating line or a longer ride.

Spare capacity is the hard part

No public data says how busy a kitchen is at 9 pm on a Tuesday. You can make a reasonable guess. A restaurant built for lunch crowds in an office district often has quiet evenings; one in a residential street may be quiet at midday. The Mobility Footfall block shows how busy the area around a place is by hour and by day, in people-hours of presence, which hints at when the street is quiet. It does not count diners, it arrives about 10 days late, and its history only goes back to mid-October 2026. Treat it as a clue for which owners to ask about their slow shifts, and let the call confirm it.

Traps to avoid

  • Choosing by rating alone. A star kitchen surrounded by five of your rivals will split the same orders.
  • Drawing circles. A 3 km circle across a river is not 3 km of delivery.
  • Ignoring hygiene and capacity checks. The data shortlists; a visit decides.
  • Signing two hosts whose reaches overlap. You pay twice for the same customers; keep each host's area mostly its own.

Keeping the shortlist current with an AI employee

Restaurants open, close and change ratings every month, so a list made once goes stale. In xMap Flows, a monthly flow does the work: Schedule, an Area of interest with your delivery districts, Nearby places for independent restaurants, xMap LLM to rule out clashing menus and rate kitchen fit, Travel area and a second Nearby places to count rivals within each reach, Sort and limit, and an Email to the partnerships team. Start from Which cuisine is missing in my delivery zone to choose the menu, Well-loved restaurants that likely do not deliver yet to find open-minded owners, and Where to open a delivery kitchen when you do want a unit of your own. The wider method of scoring places is set out in the guide to an AI employee for site selection.

Questions

What is a host kitchen?

An existing restaurant that cooks a delivery-only brand's menu under licence, alongside its own food, usually in its quieter hours.

Is a host kitchen better than renting a ghost kitchen unit?

It is cheaper and faster to start, and it lets you test a district before committing. A unit gives you more control once orders are steady.

Can data show how busy a kitchen is?

Not directly. Footfall shows how busy the street around it is by hour, in people-hours, which hints at slow shifts. Ask the owner to confirm.

How far should a host kitchen deliver?

That is your choice. Measure it as a travel time or road distance, not a circle, because rivers, rail lines and one-way streets change real reach.

Templates that do this

Read next

More on workflows by industry · Every article · Templates

Put it to work

An AI employee in xMap Flows does this on live data, on its own timetable. Start from a template or describe it in plain words.