FlowsBook a callSign inCreate free accountSign up
Business and places data

How to geocode a list of business addresses, and check the result

Geocode business addresses into map positions, see which rows landed on the right building and which only on a city centre, and fix the ones that fail.

Part of the guide Location intelligence automation

By the xMap team · · 6 min read

To geocode business addresses, you send each address (street, number, city and country) to a geocoder, which returns a latitude and longitude for it; then you check the rows that landed on a city or street centre instead of the building, and fix those before you use the list. Upload a spreadsheet of addresses and most rows come back with a usable position in minutes. The work is in the last part: knowing which positions to trust.

This guide is for anyone holding a customer, supplier or prospect list as text addresses and wanting it on a map or inside a location workflow: sales operations, franchise teams, distributors and analysts. It covers how geocoding works, the levels of accuracy, why rows fail, a made-up example with a list from Mexico City, a cleaning routine, and how an AI employee does it for every new row.

How geocoding works

A geocoder reads the address, splits it into parts (number, street, district, city, postcode, country) and looks for the best match in its own reference of places and streets. When it finds the building, you get a position on the building. When it finds only the street, you get a point somewhere on the street. When it finds only the city, you get the city's centre. The answer always looks like a pair of numbers, which is the trap: a city centre and a front door are written the same way.

It helps to think of three levels:

LevelWhat the position meansGood enough for
BuildingThe address itself, or the entranceRoutes, visits, counting what is around a site
StreetSomewhere along the street, often its middleCity-level maps, territories
City or postcodeThe centre of the city or postcode areaCounting customers per city, nothing local

How precise a geocoded position can be, and what each level is fit for.

Why some addresses fail or land in the wrong place

  • Missing city or country. A street name alone exists in hundreds of towns. A list kept for one country often leaves the country out, and the geocoder guesses.
  • Suite and floor numbers mixed in. Unit 4B, 3rd floor, rear entrance confuse the matching. Keep them in a separate column.
  • Post office boxes. They have no position worth having. Treat them as a city-level row.
  • Abbreviations and local forms. Av., Col., Blvd and district names written several ways can match the wrong place.
  • New buildings and estates. Very new streets may not be in the reference yet.
  • Two columns swapped. A city typed into the street column sends the row to the city centre.

A worked example from Mexico City

This example is made up. A distributor holds 400 shops it supplies across Mexico City as addresses in a spreadsheet, with the colonia in the same column as the street and no country.

  1. The team adds a country column and moves the colonia into its own column. This one change fixes a large share of the doubtful rows before anything runs.
  2. The sheet is uploaded to My data. Rows without latitude and longitude get a position from their address and city columns.
  3. A first map shows a tight cluster of points stacked on one spot in the city centre. Those are rows that matched only the city.
  4. For the stacked rows, the team looks each shop up by its name near the city with Map business scan's look-up of each place itself, which returns the address and position the online map shows for that business, or says it was not found.
  5. The few still unmatched are fixed by hand or dropped. The list is ready for routes and territories.

The lesson is that half the job is tidying the columns before geocoding, and the other half is looking at a map afterwards.

A cleaning routine that catches most problems

  1. Put number, street, district, city, postcode and country in separate columns, and fill the country on every row.
  2. Geocode, then put every row on a map at once.
  3. Look for stacks: many rows on exactly the same point usually means a city or postcode centre.
  4. Look for strays: points in the sea, in another country or far outside your sales area.
  5. Fix those rows by searching the business name near its city, or by hand, and geocode again.
  6. Keep a column that says how each position was found, so the next person knows which rows are rough.

Mistakes to avoid

  • Trusting every pair of numbers. A city-centre point used to plan a visit sends a rep to the wrong side of town.
  • Counting around rough points. Asking how many competitors are within 500 m of a city-centre point measures the city centre, not your customer.
  • Geocoding the same list again and again. Store the positions once and only geocode new rows.
  • Dropping failed rows silently. Keep them in a list to fix; they are often your newest customers.

How to do it with an AI employee

In xMap Flows you can geocode once by hand or let an AI employee place every new row as it arrives. For a one-off, upload the CSV on the My data page, or use Enrich a list, which places the rows from their address or city and then looks each one up for phone, website and rating. For a list that keeps growing, the blocks are:

  1. New rows: runs when rows are added to a sheet in Google Sheets or a table in My data.
  2. Map business scan, around each incoming place, looking for each place itself: the business's address, position, phone and status as the online map shows them.
  3. Filter or Branch: rows marked not found go one way, matched rows the other.
  4. Show on a map, so stacks and strays are easy to spot.
  5. An update to a sheet in Google Sheets for the matched rows and an Email listing the rows to fix by hand.

What lands in the inbox after each batch is a short list of addresses that could not be placed, with the reason, and a map of the rest. The enrich a lead sheet and push it to HubSpot template is a good start for a sales list, and what opens near my stores shows what to do with your own sites once they have positions. The workflow automation guide covers how flows like this run on their own.

Questions

What is the difference between geocoding and reverse geocoding?

Geocoding turns an address into a position. Reverse geocoding turns a position into an address, for example to name the street of a road incident.

How accurate is a geocoded position?

It depends on what was matched: the building, the street or only the city. The numbers look the same, so check stacks and strays on a map.

How many rows can I upload at once?

A table in My data takes up to 5,000 rows and 40 columns. Rows that already have latitude and longitude keep them.

Why do many of my points sit on one spot?

They matched only the city or postcode, not the building. Add the missing parts of the address or look the business up by its name.

Should I geocode the same list every time a flow runs?

No. Store the positions once and only place new rows, which is faster and keeps your list stable.

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.