FlowsBook a callSign inCreate free accountSign up
Business and places data

How to spot a new cluster of businesses before everyone else does

When several businesses of a kind open close together, a street is changing. How to find these clusters in place data, tell a real one from noise, and watch for the next.

Part of the guide Location intelligence automation

By the xMap team · · 5 min read

A new cluster of businesses is several places of the same kind, or of kinds that feed each other, opening within a short walk of one another in a short time: four specialty coffee bars on one street in a year, a run of physiotherapy and sports clinics around a new stadium, a strip of trade counters near a fresh housing estate. You spot one by asking a places database for businesses that opened recently, putting them on a map, and looking for places where the new ones bunch up more than the old ones do. Then you check that the bunch is real and not an artefact of how the data was collected.

This article is for anyone who gains from seeing a street change early: landlords and property investors, franchise developers, suppliers looking for a new patch, local media and economic development teams. It explains what a cluster is and is not, how to find one with xMap's places data, how to check it, a made-up example in Lisbon, and an AI employee that watches for the next one.

What counts as a cluster, and what does not

A cluster is about change, not density. A city centre always has many restaurants; that is a concentration, and it has been there for years. A cluster is a part of the city where the new openings run ahead of what was there before. Three tests separate the two:

  • Recent: the places opened within the same year or two.
  • Close: they sit within a few hundred metres, a distance people walk between them.
  • Above its base: the share of new places in that patch is clearly higher than in the rest of the city for the same kind of business.

Kinds that feed each other count too. A wine bar, a natural wine shop and a small plates restaurant opening on one street are a cluster of the evening economy even though no two are the same type.

Finding candidates in the data

The Nearby places block searches xMap's places database (more than 300 million places, refreshed every month) for the types you name around a point or across an Area of interest. Its Opened setting keeps only places that opened in the last 6 months, 12 months or 2 years. Put the result on a map with Show on a map in the heat view, and patches of new openings stand out. Run the same search without the Opened filter to see the base: if the heat follows the old places too, you are looking at a concentration, not a cluster.

How much openings vary between cities is visible in xMap's market figures. Of the 414 beauty salons nearest the centre of Toronto with an opening date, 31 opened in the twelve months before November 2026; of the 421 nearest the centre of Sydney with an opening date, 6 did (xMap's places database, November 2026). A city where openings are common needs a stricter test before you call a patch special.

414Toronto: beauty salons with an opening date
31Of those, opened in the last year
421Sydney: beauty salons with an opening date
6Of those, opened in the last year
The beauty salons nearest the centre of each city, xMap's places database, November 2026.

Checking that a cluster is real

Data about openings has blind spots, and a cluster can be one of them.

  • Missing dates. Not every place has an opening date. If one street's places are well dated and the next street's are not, the first will look busier with openings than it is.
  • New to the data, not new to the street. A business can appear in a database long after it opened, for example when a whole market hall is listed at once. Before acting, check a few of the places: a shop that has a long history of reviews did not open last month.
  • A single developer. Six units let in a new building at once are one decision, not a trend. They still matter, but they say less about the street.
  • Closures alongside. A patch where many open and many close is churn. The What changed block, on a flow that runs every month, lets through the places whose details changed, closures among them.

A made-up example in Lisbon

Say a bakery chain in Lisbon wants to know where its next branch would ride a change rather than fight an established street. It asks for cafes, bakeries, brunch places and specialty food shops opened in the last 12 months across the city, and for the same types with no date filter.

In this invented example, the heat map of all places glows over the historic centre, as expected. The map of new places glows over two smaller patches: a riverside street near new offices, and a residential area a little further out where seven of the nine new places sit within 400 m of a recently opened metro entrance. The first patch is mostly one building's ground floor, let at once. The second is spread across several owners and blocks. The chain visits the second patch, finds the newest places busy at weekends, and starts looking for a unit there.

An AI employee that watches for the next cluster

Finding one cluster is a project; noticing the next is a habit, and suits an AI employee.

  1. A Schedule runs on the first Monday of each month.
  2. Area of interest holds the city, and Nearby places looks for your types, opened in the last 12 months.
  3. Only new lets through only the places this flow has not found before.
  4. xMap LLM, working on the whole list, groups them by neighbourhood and says which neighbourhoods gained the most new places this month and which kinds.
  5. Show on a map and an Email send the map and the short report to you.

What lands in the inbox each month is a map of the new places, a paragraph naming the patches that gained most, and the list itself. If you already run other AI employees that find places, the Team findings block reads what they found this period against the last and reports openings by city when they rise sharply, which is another way to catch a cluster forming. Templates that start close to this: Where the city is growing fastest, Rival openings and closures every month and New places every morning, only new.

Questions

How many new places make a cluster?

There is no fixed number. Compare the share of new places in a patch with the share across the whole city for the same kind of business; a patch well above the city's share is worth a visit.

Why do some places have no opening date?

Opening dates are only known for some places, so a filter on Opened leaves the others out. Read a cluster as a lead to check, not a count of every opening.

Can it tell me why a cluster formed?

Not directly. Look at what else changed nearby, such as a new station, a large new employer or new housing, and check the news for the area.

How often should I look?

Monthly suits most uses: the places database is refreshed every month, so running more often mostly finds the same places.

Templates that do this

Read next

More on business and places data · 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.