How to label places with custom business categories your team already uses
Map data names places in thousands of ways. Give every place one label from your own short list, such as sales channels or store formats, and keep it consistent run to run.
Part of the guide Location intelligence automation
To put custom business categories on a list of places, write your categories down as a short fixed list, then have each place assigned exactly one of them by reading its name, its map category and its city. Your list should be the words your team already uses in reports and targets, such as on-trade and off-trade, or flagship, standard and kiosk, not the thousands of types map data comes with. Once every place carries one of your labels, the list can be counted, filtered and compared from month to month without anyone re-sorting it by hand.
This matters because raw categories do not line up with how businesses plan. A beverage company does not care that a place is an Izakaya, a Gastropub or a Wine bar; it cares that all three are on-trade accounts served by the same team. This article covers how to design the list, how the labelling works in xMap Flows, a made-up example, and when a different AI block suits better.
Designing a list that survives contact with the data
- Keep it short. Four to ten labels is usually right. Up to 60 are allowed, but a list nobody can remember will be applied inconsistently by people, and the same is true of AI.
- Make labels exclusive. Every place should fit one label clearly. "Premium" and "Restaurant" overlap; "Restaurant" and "Bar" mostly do not.
- Add Other. Places that fit none of your labels need somewhere to go, or they get forced into the nearest wrong one.
- Use the words of your reports. If finance reports by channel, label by channel. The point is that the output drops straight into a target sheet.
How the labelling works
Place classifier takes the places that reach it and gives each one a field called clean_category. Set its list to Custom list and write your labels one per line. xMap AI reads each place's name, its map category and its city, and picks one label. Any answer that is not on your list is dropped, so the field always holds one of your words, never a new one invented on the fly.
Two practical details. The block works in batches of 25 and labels up to 500 places a run; places past that keep their fields but get no label, and the step says so. And it can merge places that came in twice, either the same place ID or the same name within 60 metres, while it labels, which saves a separate clean-up step.
A made-up example: channels for a drinks company in Lagos
This example is invented. A drinks company in Lagos plans by four channels: On-trade (bars, restaurants, hotels), Modern trade (supermarkets and convenience chains), Traditional trade (kiosks and neighbourhood shops) and Other. Its flow searches the city's districts with Nearby places for a dozen types, then sends the results through Place classifier with those four labels.
Say 900 places come back across forty-odd database types. After labelling, the monthly email reads as four numbers per district, the same four the regional managers are measured on. When a manager disputes a label, the team looks at the place's name and type, and either accepts it or renames a label so the rule is clearer, for example splitting Traditional trade into Kiosk and Shop.
When to use a different block
Place classifier gives one label from one list. Sometimes you need more.
| Need | Better block | Why |
|---|---|---|
| One label from up to 50 options, with a confidence | Jev decision | Its Pick one question returns the option and how sure it is |
| Several labels or a label plus a reason | xMap LLM | You name the fields it writes, such as channel and why |
| A rule that never needs judgement | Calculated field | A formula on the type or brand is free and exact |
Choosing the block for labelling places.
For a rule as simple as "anything with a brand is a chain account", skip AI entirely: Calculated field can write a label from the chain flag. Use AI where judgement is really needed, such as telling a cocktail bar from a hotel bar by its name.
Putting it on a timetable
- Search your area with Nearby places or Map business scan.
- Add Place classifier with your list, and merging on.
- Add Only new if you want only newly found places each run, or leave it out for a full census.
- Send the labelled list to a sheet, or to Make a report, whose charts then show your own labels instead of database types.
Labels in your own words are what turn place data into planning data, and they are a small, early step in most location intelligence automation work. The templates below use them for mall tenants, grocery price tiers and outlet censuses.
Questions
How many custom categories can I use?
Up to 60 in Place classifier. Fewer labels give more consistent results, so start with the handful your reports use.
What happens to places that fit none of my categories?
Add Other to your list and they go there. Without it, an answer outside your list is dropped and the place gets no label.
Will the same place get the same label every month?
Usually, because the label comes from its name, type and city, which rarely change. Clear, exclusive label names make it more stable.
Can I label my own customer list, not just map data?
Yes. Upload it with My data and send it through Place classifier; it reads whatever name and category fields the rows carry.
Templates that do this
Read next
More on business and places data · Every article · Templates