FlowsBook a callSign inCreate free accountSign up
Site selection and expansion

Site selection for a ghost kitchen: where delivery demand meets cheap rent

A ghost kitchen needs no shop front, only orders it can reach quickly. How to weigh residents, rival restaurants, rider travel times and evening roads when you choose a unit.

Part of the guide AI employee for site selection

By the xMap team · · 4 min read

A ghost kitchen, a kitchen that cooks only for delivery, should sit where the most hungry households and offices fall inside a short ride, where the restaurants already delivering there are fewest for the people they serve, and where the roads still move at dinner time. Unlike a restaurant, it does not need a busy street or a nice front; a back-street unit or an industrial lot is often better, because rent is lower and riders can wait without blocking anyone. The site is chosen by the area it can reach, not by the spot itself.

This is for delivery-only brands, restaurant groups opening delivery kitchens, and property owners letting kitchen units. It covers the reach to measure, how to read demand and competition inside it, a made-up search in Mumbai, the mistakes that leave a kitchen with too few orders, and how an AI employee ranks a whole city for you.

Measure the reach, not the spot

Delivery apps show a kitchen to customers within a certain travel time or distance. A ride of 10 to 15 minutes is a common planning figure, because food that travels longer arrives cold. Travel area draws that reach around a candidate unit, along the real roads, by car or by bike: up to 15 minutes or 10 km by road. Motorbike riders move between the two, so draw both and look at the gap.

Everything else is measured inside that shape: who lives there, which restaurants already deliver there, and how the roads move when orders peak.

Three readings inside the reach

  • Households. Demographics gives residents and nine age groups for small hexagons. If your menu is aimed at young adults, weigh the groups aged 20 to 39; if it is family food, weigh the blocks with many children.
  • Restaurants already there. Nearby places counts restaurants and fast food places in the same area. Divide residents by restaurants plus one for people per restaurant, a rough gauge of room.
  • Roads at dinner time. Traffic Flow reads the current speed and jam factor on the roads around the unit at the moment it runs. Run it at 19:00 or 20:00 on several weekdays: a unit whose roads jam every evening has a smaller real reach than its map suggests.

A worked example in Mumbai

The figures are made up. A delivery brand selling regional thalis wants its third kitchen in Mumbai. The team ranks every block of the city by residents for every restaurant within 2 km, keeps the top 20, and then looks at each with a 15-minute ride by bike drawn around it.

BlockResidents within reachRestaurants within reachPeople per restaurantEvening jam factor
Block 1210,0001401,4896.8
Block 2160,000702,2543.1
Block 3240,0002609204.0

Made-up figures for illustration. Evening jam factor: the average of readings at 20:00 on ten weekdays, from 0 (free) to 10 (standstill).

Block 3 has the most residents but also the most restaurants. Block 1 looks roomy but its roads are heavily jammed at dinner, so riders reach less of it than the shape suggests. Block 2 has fewer residents, twice the people per restaurant of Block 3 and roads that flow in the evening: the team visits units there first.

Mistakes that leave a kitchen short of orders

  • Choosing a street like a restaurant. Paying for a busy frontage you will never use.
  • Drawing circles. A 3 km circle across a river or a rail line includes homes no rider reaches in time.
  • One traffic reading. Read the roads at the hours orders come in, on several days, not at noon on a Sunday.
  • Counting rivals loosely. Decide which kinds of restaurant you compete with and count only those.
  • Planning on population where there is none. In a country without population figures, lean on the restaurants around and on your own order data.

How to run it with an AI employee

The template Where to open a delivery kitchen ranks a city's blocks for a delivery kitchen. The blocks, in order:

  1. Schedule: the 1st of every month.
  2. Area of interest: the city.
  3. Demographics: every hexagon with its residents.
  4. Nearby places: restaurants and fast food within 2 km of each, counted.
  5. Calculated field: people_per_restaurant = round(population / (nearby_count + 1)).
  6. Sort and limit: the best 20.
  7. Traffic Flow: the roads within 1 km of each, run in the evening.
  8. Email: the list and a map.

What lands in the inbox is a ranked table of blocks with residents, restaurants nearby, people per restaurant and the evening speed and jam factor, with a map. Once you have candidate units, compare them side by side or draw a 15-minute ride around each with Travel area. For lunch orders from offices, Office clusters for a lunch subscription finds the densest office blocks instead.

Questions

How far should a ghost kitchen deliver?

Plan around a ride short enough for your food to arrive hot, often 10 to 15 minutes. Draw that reach on real roads with Travel area rather than a circle.

Does a ghost kitchen need footfall?

Not walk-in footfall: nobody comes in. What matters is the people within reach and how fast riders can get to them.

Can I see how many delivery orders an area places?

No dataset here counts orders. Residents, their ages and the restaurants already serving the area are the proxies; your own order data is the best check once you open.

Why read traffic in the evening?

Because that is when most dinner orders come. Traffic Flow measures the roads at the moment it runs, so schedule it for the hours that matter.

Templates that do this

Read next

More on site selection and expansion · 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.