This is a step-by-step guide to planning a Google Maps search for a whole US niche with free Census data.
We’ll go through one example: every company that pumps septic tanks or rents out portable toilets in Wisconsin, Minnesota and Iowa.
The Census numbers in this guide are real. I counted them on the official files.
The Census county file shows septic businesses in 509 US counties. But 1,302 of the 4,054 septic businesses with employees sit in counties the file doesn’t list. A search plan built from that file never looks there.
This guide is for founders and list builders who sell to local operators spread across whole states. Every Census download in it is free. If you already have a county grid, skip to step 5.
ZIP codes or a Census grid: which one to use
My TAM playbook searches Google Maps ZIP code by ZIP code: one search for every ZIP code and every search term. That’s the simplest complete method when a search costs a fraction of a cent. The Maps Data API on RapidAPI, for example, has a plan with 300,000 searches for $25 a month.
The Census grid fits better in three cases:
- You use Google’s own Places API. You can limit each search to a rectangle on the map, and each search stops at 60 results. The grid tells you where to use small rectangles.
- You pay per result. Apify’s Maps scraper charges per business it returns. A grid searches densely only where the niche lives and keeps overlapping searches low.
- You want to know what to expect. The Census tells you how many businesses with employees each state has. That gives you a number to compare your results with.
The two also combine well: ZIP codes in the dense counties, one search per county everywhere else. Either way, the Census check in steps 1 and 2 takes ten minutes and shows you where the niche lives.
The running example: say you work at Sumplark, a company that sells tank-level sensors for portable toilets and holding tanks. Sumplark wants every septic and portable toilet operator in the three states, and later only the ones with commercial customers.
Septic tanks are the underground tanks of homes and farms without a sewer connection. So these companies work in small towns and rural counties, the places a list of big cities never searches.
1. Find your industry code and download two Census files
Find your industry code first. NAICS is the US system of industry codes. Every industry has a 6-digit code, and its first 4 digits name the family it belongs to. NAICS 562991, Septic Tank and Related Services, covers pumping septic tanks and cesspools and renting or servicing portable toilets. Its family, 5629, also holds remediation, recycling facilities and work like sewer and catch basin cleaning.
Then download County Business Patterns. County Business Patterns (CBP) is a free Census file that counts businesses with paid employees by county and industry code. It counts establishments, which means business locations: a company with two yards counts twice. The latest year is 2023. You need two files from the CBP datasets page on census.gov: the county file and the state file. No account or key.
Businesses without employees are not in CBP. A one-truck operator never shows up. So you’ll always find more businesses on Maps than the Census counts.
Two details in the files cause mistakes. In the county file, shorter codes are filled up with slashes, so the 5629 family appears as “5629//”. In the state file, keep only the rows that count all legal forms together, or you count the same businesses twice.
2. Check how much the 6-digit file hides
Since 2017, CBP publishes a number only if it covers three or more businesses. Smaller numbers, like the septic businesses in a county with two of them, are left out of the file entirely. So a county with two septic businesses has no row at all.
For a niche full of small businesses, that hides a lot. I counted it on the 2023 files for NAICS 562991:
- The state files add up to 4,054 septic businesses with employees.
- County rows exist for 509 counties and add up to 2,752.
- The other 1,302, a third of the total, sit in counties without a row.
- The 4-digit family 5629 has a row in 978 counties.
Some states almost disappear. Kansas has 24 septic businesses in its state total and not one county row at 6 digits.
The Census Bureau’s methodology page says it’s no longer current while the Bureau reworks how it protects small cells. Check the rule again when the 2024 files come out.
3. Sort every county into three groups
Family rows show up in more counties because they combine septic with related industries. That makes it a hint, never a count. Nationally, septic is 4,054 of the family’s 12,847 businesses. Where the family shows up and the 6-digit code doesn’t, expect one or two septic businesses the file can’t show.
So I’d sort every county into three groups, which I call tiers:
- Tier A: has a 6-digit row. Dense enough to split into smaller search areas.
- Tier B: only the 4-digit family shows. Search it as one area.
- Tier C: neither shows. Search it once anyway, because one or two businesses and every operator without employees are invisible in both files.
Here’s Sumplark’s territory, counted on the same files:
| State | Counties | Tier A | Tier B | Tier C | State total | In county rows | Not in a county row |
|---|---|---|---|---|---|---|---|
| Wisconsin | 72 | 22 | 14 | 36 | 157 | 100 | 57 |
| Minnesota | 87 | 12 | 13 | 62 | 96 | 51 | 45 |
| Iowa | 99 | 4 | 11 | 84 | 53 | 17 | 36 |
| Total | 258 | 38 | 38 | 182 | 306 | 168 | 138 |
A grid built only from 6-digit rows would search 38 of 258 counties. It would miss the counties that hold 138 of the 306 septic businesses with employees.
4. Build the grid
Take the list of counties from somewhere other than CBP, or tier C never appears. The Census Gazetteer file is a plain list of every county with its code, and the Census county boundary file gives you each county’s shape. Both are free downloads on census.gov.
Here’s the prompt I’d give Claude Code:
Build tools/build_grid.py. No API keys needed.
Download into .tmp/census/ from census.gov:
- County Business Patterns 2023, county file and state file
- the latest Gazetteer county file
- the latest cartographic boundary file for counties (500k)
Parameters: NAICS6 = "562991", STATES = ["WI", "MN", "IA"]
FAMILY = first four digits of NAICS6 + "//"
1. The county list comes from the Gazetteer. CBP only adds counts to it.
2. From the CBP county file, take the establishment count (est) for NAICS6
and for FAMILY in every county. Skip the statewide rows (county 999).
3. From the CBP state file, rows for all legal forms only (lfo == "-"),
take the state total for NAICS6.
4. Tier A = has a NAICS6 row. Tier B = only a FAMILY row. Tier C = neither.
5. Print per state: counties per tier, state total, sum of county rows
and the difference. Stop here and show me the table.
6. Take each county's bounding box from the boundary file. Split tier A
boxes into 2 x 2 areas. Tiers B and C stay one area each.
7. Write .tmp/grid.csv with: area_id, county_fips, county, state, tier,
est6, est4, low_lat, low_lng, high_lat, high_lng.
Point 5 is there on purpose. Open the files yourself for one state and compare the numbers. If they differ, a filter broke, and you want to know that before anything gets searched.
For Sumplark, that’s 38 × 4 + 38 + 182 = 372 search areas.
5. Pick reference places before you search
You need a way to tell whether the search found everything. A percentage of the Census count doesn’t work. CBP counts businesses with employees, while Maps also lists one-truck operators, and sometimes the same business twice. A search can beat the Census number in one county and miss real businesses in the next.
Use reference places instead: 30 to 50 businesses you know exist, from sources other than Google. Pick them before the search, so you can’t lean toward the ones you already found.
- State license lookups. Minnesota, for example, has a public search of licensed septic businesses. Use it to check your search, not as a mailing list.
- Association directories. The Portable Sanitation Association International (PSAI) has a public operator directory you can filter by ZIP code, city, state and county. Some listings show only to logged-in members.
- Your own customers in the niche.
- Overture, a free map dataset built from sources other than Google.
Spread them across all three tiers, with a few from tier C.
6. Choose the tool for the search
Three common options return Google Maps listings. They differ in how you search and how you pay.
| Google Places API | Maps Data API on RapidAPI | Apify Google Maps Scraper | |
|---|---|---|---|
| What it is | Google’s official API | An independent provider’s Maps API | A scraper of public Maps pages, run on Apify |
| How you search | Search term inside a rectangle, up to 60 results | Search term plus a place or ZIP code, up to 200 results | Search term plus a county, state or drawn area, splits the area itself |
| How you pay | Searches that return only IDs are free. Contact details cost $20 per 1,000 after 1,000 free a month | Monthly plans: 1,000 searches free, 30,000 for $5, 300,000 for $25 | Per business returned: $4.00 per 1,000 on the free plan, $3.00 on Starter, $2.00 on Scale |
For Sumplark, the example walks through Google’s Places API, because its 60-result cap is where a grid matters most. You need a Google Cloud account and an API key, the password your script sends with every search. Claude Code can walk you through creating both. The same grid works with the other two. With the Maps Data API, a search returns up to 200 businesses, so fewer areas fill up.
Read the terms for what you plan to do. Google’s Places policies say you may store place IDs indefinitely, but must not store other Places content beyond the allowed exceptions. Scrapers collect what Maps shows publicly, and Apify tells you to consider Google’s terms and personal data rules yourself. Get a yes from whoever handles legal questions at your company before you build a list you plan to keep.
7. Write the search matrix
Every search is one area and one search term. Write all of them into a file before you run anything. Then a crashed run can continue where it stopped, and every result can be traced back to its search.
Pick distinct terms, not synonyms. Two terms that mean the same thing find the same businesses and double your searches. Sumplark starts with four: “septic service”, “septic tank pumping”, “portable toilet rental” and “holding tank pumping”.
The template, with example rows:
search_matrix.csv (one row per area x term)
area_id,county_fips,state,tier,term,low_lat,low_lng,high_lat,high_lng,requests,results,status
55009-1,55009,WI,A,septic tank pumping,44.24,-88.25,44.43,-87.89,3,60,full
55009-1a,55009,WI,A,septic tank pumping,44.24,-88.25,44.34,-88.07,2,23,done
55009-1,55009,WI,A,portable toilet rental,44.24,-88.25,44.43,-87.89,2,31,done
19001-1,19001,IA,C,septic service,41.16,-94.70,41.51,-94.24,1,4,done
27031-1,27031,MN,C,holding tank pumping,47.69,-90.88,48.09,-89.48,1,2,done
Split per term, not per area. When an area and a term return the full 60 results, split that area into four for that term only. In the rows above, “portable toilet rental” stops at 31 in the same area, so splitting it too would only cost requests.
Ask only for place IDs during the search. The place ID is Google’s unique ID for each business. With the Places API, text searches that return only IDs are free. You fetch the details later, once per business.
For Sumplark, 372 areas × 4 terms = 1,488 rows. Say 27 of them come back full and get split. Google returns 20 results per page, so a full search takes 3 requests. With the extra pages, that makes 2,100 requests. Cost so far: $0.
8. Remove duplicates, then check your reference places
The search returns the same business many times: once for each matching term, once for each overlapping area, and again on a rerun.
First, merge by place ID. Reduce every result to unique place IDs before you pay for anything. Say Sumplark’s 2,100 requests return 6,900 results and 1,240 unique place IDs.
Then fetch the details once per ID. Name, coordinates, category, website and phone come from one details call. Website and phone are in Google’s most expensive details tier: $20 per 1,000 calls after 1,000 free calls a month. For 1,240 IDs, that’s $4.80.
Put each business in the county its coordinates fall in, never the county of the area that found it. The areas are rectangles, and county lines aren’t.
Then merge by phone and website domain. A septic business with a separate listing for its portable toilets is one business with two IDs. So is one listed at its yard and at the owner’s house. Say the 1,240 listings become 1,060 businesses.
Now check your reference places. For every reference place the search missed, write down why: no Maps listing, a category or name your terms didn’t reach, an area that was full, or your own merge step removing it. Fix the reason and run the search again. Never paste the missing business in by hand. The fix to the search is what finds the businesses you don’t know about.
Say Sumplark finds 33 of its 40 reference places. Three misses are listed as excavating contractors. Two sit in an area that was still full after the first split. One has no listing, and one is listed under the owner’s name. A fifth search term, “septic excavating”, and one more split bring it to 38 of 40.
9. Separate commercial from residential
A Maps category says what a business does, never who it does it for. “Septic system service” covers the pumper who empties the tanks of homeowners and the operator with 300 portable toilets on construction sites. And Google’s own list of place types has no septic type at all. Plumber is the closest.
Association membership is good evidence. PSAI describes its operators as serving construction, agriculture, events and emergency services. So a match in its operator directory points to commercial work. Match on phone number and website domain, never on the name alone. A company website that names construction sites, events or campgrounds counts too.
A license doesn’t tell you. Wisconsin, for example, requires the same license from every business that pumps septic tanks, holding tanks or portable toilets.
Let AI read each website for the rest. A script opens every website and gives its text to a model with one question. Here’s the prompt I’d use:
You are checking one septic or portable toilet business.
Question: Does this business serve commercial customers, like
construction sites, events, farms, campgrounds or businesses?
Answer NO if: it only serves homeowners.
Answer NOT_OPERATOR if: it is an excavator, a treatment plant,
a supply store or a plumber that doesn't pump tanks.
Input: name, Maps category, website text (first 4,000 characters)
Return: answer (yes / no / not_operator / unclear) and the exact
sentence from the website behind the answer.
Use only the input. If the text doesn't say it, answer unclear.
Before the full run, label 20 businesses yourself and fix the prompt until the model agrees with you on two rounds of 10 in a row. Then run it on every business.
Say 640 of Sumplark’s 1,060 businesses are real operators once the check removes excavators, treatment plants and supply stores. That’s about twice the 306 businesses in the Census files, and both can be right.
The Census counts only businesses with employees whose main work is septic. Maps also lists one-truck operators and plumbers who pump tanks on the side. The Census does count businesses without employees in a separate file, but only for waste management as a whole, so it can’t give you a septic number.
Of the 640, 70 match the PSAI directory and 120 more name commercial work on their own website. That’s 190 commercial operators, 330 with evidence of home customers only, and 120 marked unclear because nothing points either way.
What you hand over: one row per business with place ID, name, county, tier, the terms that found it, website, phone, the commercial evidence with its source link, and a status (commercial, home only, unclear or excluded). Keep the reference check with the list, so whoever uses it can see what was tested.
The next step is finding the owners. There’s a separate guide on that for companies that aren’t on LinkedIn.
Where to start today
You can map your niche today in three steps:
- Download the two Census files for your industry code and check how much the 6-digit file hides.
- Build the grid and write the search matrix (steps 3 to 7).
- Run the searches, remove the duplicates and separate commercial from residential. What’s left is your niche.


