Nobody has a census of business websites. We are building one.
A census counts everyone. Not a sample, not the ones who answered, everyone in a named place. Nobody has one for business websites. There are samples, guesses and lists people sell. There is no count.
We are building it. This page is where it starts, and it will change as the count grows. Every count and share on it is read from our own records when the page is built.
85,095 addresses gave us a page, over 560,895 tries. The count is smaller because it leaves out 4,160 chains in the United States, one website over many storefronts, and 1,658 addresses that a later reading found were no longer the business. It also leaves out 470 sites outside the United States, 82 of them chains.
Newest reading October 16, 2026. A shift between two reads may mean new sites, changed sites, or a change in our own measuring software, not a change in the trade.
What we hold today
We built a machine that reads a business website from the outside, the way the server first sends it to a stranger, before any scripts run. It reads one page, the front page, in a few seconds. It grades that page on 40 checks.
It now holds 78,807 business websites in the count. The middle one scored 83 out of 100. About 8 in every 100 earned an A. About 8 in every 100 earned a D or an F.
An A starts at 92, a B at 82, a C at 70 and a D at 55; below that is an F. Those cut-offs are read from the engine's own code when this page is built, so a letter here means the same thing as a letter on a report.
Not one site is named here. We read them without asking, the way a customer would. Naming them would turn a count into a shaming. The pattern is the point.
Why this is not a census yet
Three things make this a sample, and a shaped one. Know them before you trust any share on this page.
Where the addresses came from
Nearly all of the addresses came from one place: OpenStreetMap, a map that volunteers keep and the OpenStreetMap Foundation publishes under the Open Database License. A business is on it because someone put it there and wrote its website next to it. A business with no website is not in it. A business nobody mapped is not in it. So the count leans toward the people who need us least.
What kind of business
By the kind of business each site is listed as, about 42 in every 100 are places to eat or drink: restaurants, cafes, bars and the like. A middle number taken across all of them is in large part a number about them. Read every blended share with that in mind.
Where they are
The map has more volunteers in some states than others. So the count follows the mappers, not the people. We also count a site by the search area that found it, not its mailing address, and a search area can cross a state line. A search area near a national border can reach across it too. We label a site outside the United States when its map entry gives an address there. Some entries name no country, state, province or postal code: often just a street, or nothing. We label one of those when its point lies inside Canada on Canada's own boundary map and inside no state on the Census Bureau's state boundary map. We keep it and leave it out of every state, every county and the count: 609 addresses so far. A website with any map entry inside a state, or with a US address, is counted as a US site, even when another of its entries is across the border. Mexico has no boundary map here, so a site in Mexico is placed only by an address that names Mexico. Where the two boundary maps do not agree on the line, or where both countries claim the water, a site with no address could be placed on the wrong side. And on September 30, 2026, we had not yet run this check on 6,528 addresses from searches in 14 border states, most of them in Minnesota, New Hampshire and Maine. So some sites near a border may still be counted in a state. Canada's boundary map: Adapted from Statistics Canada, Province and Territory 2021, Digital Boundary Files (DBF), January 1, 2021. This does not constitute an endorsement by Statistics Canada of this product. Licence: Open Government Licence - Canada. Florida, our own state, has 7,572 sites in the count, because we started here. New Jersey is next with 8,994. Texas, the second most populous state, has 1,326. The count reaches 50 states and the District of Columbia, and it is thin in most of them.
About 27 in every 100 addresses never gave us a page
We tried 117,229 addresses. 32,134 of them, about 27 in every 100, never gave us a page we could count. Here is what they did instead, in the machine's own words for each kind.
| What the address did | Addresses |
|---|---|
| Its name no longer points to any server | 7,256 |
| Its security certificate failed our check | 4,360 |
| A server answers to the name but refused or dropped the connection | 1,818 |
| Did not answer within our time limit | 2,241 |
| Could not be reached, kind not recorded | 883 |
| Answered 403, blocking automated visitors | 6,695 |
| Answered 429, asking us to slow down | 160 |
| Answered 404, the front page is gone | 1,017 |
| Answered 410, gone for good | 24 |
| Put up a security challenge | 1,038 |
| Server error (a code in the 500s) | 418 |
| Asked every crawler to stay out, so we did | 1,114 |
| Gave a page that was not the business's own site | 432 |
| Sends visitors to another address | 4,522 |
| A page on an ordering, booking or listing platform, not a site of its own | 12 |
| Something else | 144 |
The addresses that gave us no page fall into four groups. 8,729 of the 32,134, about 27 in every 100, point to a site that is gone, broken or taken over. That means the name points to no server, the front page is gone or answered 410, or the page is parked, hijacked or a placeholder.
9,007, about 28 in every 100, exist and refuse automated visitors. That says nothing about the business, so they carry no grade and sit in no share of graded sites; they are counted as addresses tried.
4,534 are not a site of their own: the address sends visitors to another address, which may be the same business on a new one, or it is a platform's page for the business. The rest, 9,864, timed out, dropped the connection, gave an error, failed our check of their security certificate, or answered in a way we have no group for. A security certificate is the proof of identity a browser checks; our check refusing one does not show the site is gone. We knock on every address again on a schedule, and a certificate refusal stays in this group however often it repeats. Until October 10, 2026, we counted those refusals as gone; a sample of 60 found about four in ten still serve their own page.
Dead is a finding too. Of the 8,729 addresses gone, broken or taken over, 8,729 have been knocked on again on a later day and still gave us no page we could count. 8,729 of them were knocked on three or more separate days. 0 have been asked on one day only so far. A map full of website addresses that lead nowhere is part of what a census would show.
Where the dead addresses are
The share of addresses tried that lead to a site gone, broken or taken over, by trade and by state: in each table, Tried is the addresses tried and Gone is that share.
Every address tried counts toward the share, the ones that refuse automated visitors included. A share is stated only where at least 200 addresses were tried; the rest are folded into one row. This measures how stale the map's addresses are as much as how businesses fare, because a website address mapped years ago rots either way.
Next to each share is its chance margin, in points. 95 times in 100, chance alone would not move the share by more than that many points either way. It is worked out from that row's own count of addresses, so a row with fewer addresses has a wider margin.
Chance can reach further on one side than the other, so the margin takes the wider side, rounded up to a whole point. The motel row, at 18% with a margin of 6, shows how: chance alone could put it anywhere from 13% to 24%, and 12% to 24% covers all of that.
To compare two rows, take each share plus and minus its margin. When the two ranges do not overlap, the gap between the rows is more than chance. When they overlap, the ranges alone do not settle it: the gap may be chance, or it may still be more than chance.
| Trade | Tried | Gone | Chance margin, points |
|---|---|---|---|
| motel | 228 | 18% | ± 6 |
| copyshop | 246 | 17% | ± 6 |
| dry cleaning | 290 | 16% | ± 5 |
| massage | 485 | 15% | ± 4 |
| tax advisor | 200 | 15% | ± 6 |
| salons and spas | 5,699 | 14% | ± 1 |
| auto shops | 2,680 | 14% | ± 2 |
| antiques | 222 | 14% | ± 6 |
| pet grooming | 333 | 13% | ± 5 |
| laundry | 286 | 13% | ± 5 |
| caterer | 213 | 12% | ± 6 |
| alcohol | 777 | 11% | ± 3 |
| tattoo | 519 | 11% | ± 3 |
| childcare | 488 | 11% | ± 4 |
| dance | 281 | 11% | ± 5 |
| second hand | 269 | 11% | ± 5 |
| photographer | 257 | 11% | ± 5 |
| kindergarten | 727 | 10% | ± 3 |
| contractors and remodelers | 675 | 10% | ± 3 |
| financial | 334 | 10% | ± 4 |
| accountant | 328 | 10% | ± 4 |
| interior decoration | 277 | 10% | ± 5 |
| insurance | 999 | 9% | ± 2 |
| jewelry | 849 | 9% | ± 3 |
| car wash | 426 | 9% | ± 4 |
| fast food | 3,581 | 8% | ± 1 |
| clothes | 2,525 | 8% | ± 2 |
| real estate businesses | 1,327 | 8% | ± 2 |
| gift | 721 | 8% | ± 3 |
| pet | 472 | 8% | ± 3 |
| storage rental | 427 | 8% | ± 4 |
| pharmacy | 366 | 8% | ± 4 |
| art | 332 | 8% | ± 4 |
| electronics repair | 259 | 8% | ± 5 |
| car parts | 239 | 8% | ± 5 |
| restaurants | 32,023 | 7% | ± 1 |
| cafe | 5,284 | 7% | ± 1 |
| bar | 3,587 | 7% | ± 1 |
| fitness centre | 2,130 | 7% | ± 2 |
| sports centre | 1,459 | 7% | ± 2 |
| alternative | 1,127 | 7% | ± 2 |
| furniture | 793 | 7% | ± 3 |
| optician | 501 | 7% | ± 3 |
| sports | 423 | 7% | ± 3 |
| shoes | 295 | 7% | ± 4 |
| psychotherapist | 240 | 7% | ± 4 |
| kitchen | 228 | 7% | ± 4 |
| car | 2,183 | 6% | ± 2 |
| law practices | 2,137 | 6% | ± 2 |
| pub | 1,824 | 6% | ± 2 |
| hotel | 1,710 | 6% | ± 2 |
| bakery | 1,355 | 6% | ± 2 |
| ice cream | 767 | 6% | ± 2 |
| air conditioning companies | 509 | 6% | ± 3 |
| it | 466 | 6% | ± 3 |
| advertising agency | 385 | 6% | ± 3 |
| plumbers | 356 | 6% | ± 3 |
| hardware | 307 | 6% | ± 4 |
| deli | 294 | 6% | ± 4 |
| animal boarding | 223 | 6% | ± 4 |
| medical and dental practices | 5,304 | 5% | ± 1 |
| golf course | 2,182 | 5% | ± 1 |
| clinic | 1,641 | 5% | ± 2 |
| bicycle | 781 | 5% | ± 2 |
| florist | 696 | 5% | ± 2 |
| studio | 434 | 5% | ± 3 |
| marina | 331 | 5% | ± 3 |
| architect | 288 | 5% | ± 4 |
| music school | 203 | 5% | ± 5 |
| garden centre | 341 | 4% | ± 3 |
| trade | 317 | 4% | ± 3 |
| wine | 252 | 4% | ± 4 |
| musical instrument | 235 | 4% | ± 4 |
| confectionery | 224 | 4% | ± 4 |
| rehabilitation | 219 | 4% | ± 4 |
| veterinary | 1,221 | 3% | ± 2 |
| books | 680 | 3% | ± 2 |
| brewery | 647 | 3% | ± 2 |
| roofers | 405 | 3% | ± 3 |
| physiotherapist | 374 | 3% | ± 3 |
| optometrist | 314 | 3% | ± 3 |
| camp site | 229 | 3% | ± 4 |
| financial advisor | 216 | 3% | ± 4 |
| consulting | 204 | 3% | ± 4 |
| farm | 201 | 3% | ± 4 |
| funeral directors | 396 | 2% | ± 2 |
| property management | 380 | 2% | ± 2 |
| everything under 200 addresses tried, together | 9,571 | 7% | ± 1 |
| State searched | Tried | Gone | Chance margin, points |
|---|---|---|---|
| New Hampshire | 1,324 | 18% | ± 3 |
| New Mexico | 604 | 13% | ± 3 |
| Pennsylvania | 1,588 | 11% | ± 2 |
| Oregon | 6,129 | 10% | ± 1 |
| Florida | 12,073 | 9% | ± 1 |
| Massachusetts | 5,469 | 9% | ± 1 |
| District of Columbia | 2,184 | 9% | ± 2 |
| Rhode Island | 995 | 9% | ± 2 |
| Louisiana | 900 | 9% | ± 3 |
| Nebraska | 687 | 9% | ± 3 |
| South Dakota | 222 | 9% | ± 5 |
| New Jersey | 13,101 | 8% | ± 1 |
| Maryland | 6,267 | 8% | ± 1 |
| Colorado | 4,448 | 8% | ± 1 |
| Michigan | 2,555 | 8% | ± 2 |
| Nevada | 1,437 | 8% | ± 2 |
| Ohio | 1,436 | 8% | ± 2 |
| Connecticut | 1,397 | 8% | ± 2 |
| Iowa | 1,066 | 8% | ± 2 |
| Indiana | 815 | 8% | ± 3 |
| Arizona | 2,419 | 7% | ± 2 |
| Texas | 2,049 | 7% | ± 2 |
| Tennessee | 890 | 7% | ± 2 |
| Maine | 748 | 7% | ± 3 |
| Kentucky | 740 | 7% | ± 3 |
| Idaho | 727 | 7% | ± 3 |
| Oklahoma | 619 | 7% | ± 3 |
| California | 6,819 | 6% | ± 1 |
| Illinois | 5,533 | 6% | ± 1 |
| Georgia | 3,788 | 6% | ± 1 |
| Minnesota | 3,738 | 6% | ± 1 |
| Delaware | 2,834 | 6% | ± 1 |
| Missouri | 1,843 | 6% | ± 2 |
| Wisconsin | 1,617 | 6% | ± 2 |
| Virginia | 1,440 | 6% | ± 2 |
| North Carolina | 1,229 | 6% | ± 2 |
| New York | 1,207 | 6% | ± 2 |
| Alabama | 506 | 6% | ± 3 |
| Washington | 6,586 | 5% | ± 1 |
| Vermont | 616 | 5% | ± 3 |
| Arkansas | 330 | 5% | ± 4 |
| Kansas | 231 | 5% | ± 4 |
| Utah | 2,429 | 4% | ± 1 |
| Hawaii | 934 | 4% | ± 2 |
| Alaska | 260 | 4% | ± 4 |
| South Carolina | 1,235 | 3% | ± 2 |
| everything under 200 addresses tried, together | 556 | 10% | ± 3 |
| outside the United States | 609 | 8% | ± 3 |
By county, the same share is stated for the 89 counties, of the 386 we reached, where at least 200 addresses were tried. In Florida the county is the one we searched, so a business near a county line can be filed under its neighbor. Elsewhere it is the county our table of towns lists beside the business's town. Neither is a street address.
About 28 in every 100 of the addresses tried, 32,257, cannot be placed in a county. 16,762 are in towns our table lists under more than one county, such as Portland, Denver and Minneapolis, so we leave them out rather than guess. 15,468 are in towns our table does not list. 27 carry no town or state we can read, or a town our table lists without a county. They have their own row at the foot of the table.
| County | Tried | Gone | Chance margin, points |
|---|---|---|---|
| Rockingham County, New Hampshire | 414 | 26% | ± 5 |
| Hillsborough County, New Hampshire | 665 | 16% | ± 3 |
| Leon County, Florida | 413 | 14% | ± 4 |
| Bernalillo County, New Mexico | 536 | 13% | ± 4 |
| Dakota County, Minnesota | 220 | 13% | ± 6 |
| Pinellas County, Florida | 663 | 12% | ± 3 |
| Volusia County, Florida | 459 | 12% | ± 4 |
| Polk County, Florida | 425 | 12% | ± 4 |
| Washington County, Oregon | 856 | 11% | ± 3 |
| Plymouth County, Massachusetts | 256 | 11% | ± 5 |
| Jefferson Parish, Louisiana | 242 | 11% | ± 5 |
| Hamilton County, Indiana | 237 | 11% | ± 5 |
| New Haven County, Connecticut | 210 | 11% | ± 6 |
| Bergen County, New Jersey | 1,466 | 10% | ± 2 |
| Allegheny County, Pennsylvania | 1,168 | 10% | ± 2 |
| Baltimore County, Maryland | 587 | 10% | ± 3 |
| Prince George's County, Maryland | 405 | 10% | ± 4 |
| Escambia County, Florida | 237 | 10% | ± 5 |
| DeKalb County, Georgia | 231 | 10% | ± 5 |
| Middlesex County, Massachusetts | 2,130 | 9% | ± 2 |
| Hillsborough County, Florida | 1,018 | 9% | ± 2 |
| Wayne County, Michigan | 934 | 9% | ± 2 |
| Harford County, Maryland | 836 | 9% | ± 3 |
| Clackamas County, Oregon | 716 | 9% | ± 3 |
| Oakland County, Michigan | 689 | 9% | ± 3 |
| Norfolk County, Massachusetts | 615 | 9% | ± 3 |
| Lee County, Florida | 598 | 9% | ± 3 |
| Orange County, Florida | 543 | 9% | ± 3 |
| Snohomish County, Washington | 521 | 9% | ± 3 |
| Gwinnett County, Georgia | 210 | 9% | ± 5 |
| Hudson County, New Jersey | 5,718 | 8% | ± 1 |
| Miami-Dade County, Florida | 1,922 | 8% | ± 2 |
| Clark County, Nevada | 1,422 | 8% | ± 2 |
| Orange County, California | 1,063 | 8% | ± 2 |
| Polk County, Iowa | 808 | 8% | ± 3 |
| District of Columbia | 644 | 8% | ± 3 |
| Broward County, Florida | 566 | 8% | ± 3 |
| Middlesex County, New Jersey | 551 | 8% | ± 3 |
| Lake County, Florida | 443 | 8% | ± 3 |
| Jefferson County, Kentucky | 428 | 8% | ± 4 |
| Pierce County, Washington | 362 | 8% | ± 4 |
| Manatee County, Florida | 328 | 8% | ± 4 |
| Monmouth County, New Jersey | 279 | 8% | ± 4 |
| Morris County, New Jersey | 235 | 8% | ± 5 |
| Maricopa County, Arizona | 2,352 | 7% | ± 2 |
| Erie County, New York | 964 | 7% | ± 2 |
| Boulder County, Colorado | 784 | 7% | ± 2 |
| Palm Beach County, Florida | 768 | 7% | ± 3 |
| Hartford County, Connecticut | 760 | 7% | ± 3 |
| Cumberland County, Maine | 586 | 7% | ± 3 |
| Essex County, Massachusetts | 559 | 7% | ± 3 |
| Providence County, Rhode Island | 378 | 7% | ± 4 |
| Fulton County, Georgia | 320 | 7% | ± 4 |
| DuPage County, Illinois | 258 | 7% | ± 4 |
| Los Angeles County, California | 5,659 | 6% | ± 1 |
| Cook County, Illinois | 4,711 | 6% | ± 1 |
| Suffolk County, Massachusetts | 1,341 | 6% | ± 2 |
| New Castle County, Delaware | 819 | 6% | ± 2 |
| Milwaukee County, Wisconsin | 767 | 6% | ± 2 |
| Anne Arundel County, Maryland | 648 | 6% | ± 3 |
| Davidson County, Tennessee | 596 | 6% | ± 3 |
| Ada County, Idaho | 588 | 6% | ± 3 |
| Montgomery County, Maryland | 556 | 6% | ± 3 |
| Brevard County, Florida | 442 | 6% | ± 3 |
| Harris County, Texas | 396 | 6% | ± 3 |
| Cobb County, Georgia | 1,284 | 5% | ± 2 |
| Mecklenburg County, North Carolina | 750 | 5% | ± 2 |
| Macomb County, Michigan | 689 | 5% | ± 2 |
| Essex County, New Jersey | 659 | 5% | ± 2 |
| Hennepin County, Minnesota | 442 | 5% | ± 3 |
| Howard County, Maryland | 366 | 5% | ± 3 |
| Duval County, Florida | 281 | 5% | ± 4 |
| Davis County, Utah | 263 | 5% | ± 4 |
| Sedgwick County, Kansas | 214 | 5% | ± 4 |
| Franklin County, Ohio | 200 | 5% | ± 4 |
| King County, Washington | 4,839 | 4% | ± 1 |
| Salt Lake County, Utah | 1,174 | 4% | ± 2 |
| Honolulu County, Hawaii | 896 | 4% | ± 2 |
| Chittenden County, Vermont | 475 | 4% | ± 3 |
| Cherokee County, Georgia | 290 | 4% | ± 3 |
| Anchorage Municipality, Alaska | 238 | 4% | ± 4 |
| Douglas County, Colorado | 225 | 4% | ± 4 |
| Collier County, Florida | 703 | 3% | ± 2 |
| Kitsap County, Washington | 543 | 3% | ± 2 |
| Utah County, Utah | 358 | 3% | ± 3 |
| Jefferson County, Colorado | 344 | 3% | ± 3 |
| Charleston County, South Carolina | 263 | 3% | ± 3 |
| Waukesha County, Wisconsin | 229 | 3% | ± 4 |
| Monroe County, Florida | 224 | 3% | ± 4 |
| everything under 200 addresses tried, together | 10,893 | 9% | ± 1 |
| outside the United States | 609 | 8% | ± 3 |
| could not be placed in a county | 32,257 | 8% | ± 1 |
How the machine behaves
A census means reading a great many sites that did not ask to be read. So the machine has manners, and every one of them is read from the code that keeps it, not from a promise.
- It says its name on every request:
Mozilla/5.0 (compatible; TrueMeasure-Inspector/2.0; +https://truemeasurewebdesign.com/inspector). That address leads to a page for server owners. It tells them how to slow us down, block us, or refuse us for good. - One inspection makes at most 9 requests to a site, plus one more each time the site forwards us to another address. That limit is read from the code. The first request asks for robots.txt, a small file on a website that tells automated visitors where they may go.
- Only one request reads the front page. The rest ask for small files around it, such as the sitemap (a list of the site's pages), the site icon, and one address we expect to be missing. When its requests are done, it leaves. It gives up on a page after 12 seconds and stops reading at 3 megabytes.
- It reads robots.txt first. A site whose robots.txt shuts out all automated visitors (crawlers), or shuts us out by name, is not read at all.
- Inside one inspection its requests go out together, within a second or two. A new visit to the same website starts at least 2 seconds after the last one began, no matter how many sites it is reading. If the site's robots.txt asks for a longer delay, that delay is the least it waits, up to 60 seconds. Ask for longer and we take it as a no.
- It never runs a site's scripts (the code that draws parts of a page after it arrives), fills a form or signs in, and on a business website it never follows a link. It reads what the server sends and closes.
- To find business websites, it also reads one public business directory, a chamber of commerce member list, a few times a year. There it follows links, but only to that directory's own pages. It sends at most one request a second, obeys every rule in its robots.txt, and holds each request to the limits above.
- Opting out costs nothing. Neither the census nor the free inspection starts reading a site on our do-not-measure list, robots.txt included. A different site that forwards to it, or borrows its image, icon or sitemap, can still send a request its way. No one has asked us to add a site yet. The address to ask is on the inspector page.
- It keeps a reading of the page. It keeps whether a business shows a phone number, with a fingerprint of each number. The fingerprint is a code made with a secret key kept only on our census machine. It can be matched against a public register, and without that key it cannot be turned back into the number. Readings taken before this change reached the census still hold the number itself. It keeps license numbers as printed.
- Copies of front pages are kept only for a named study, privately, and never published. Our rule is to delete them when the study ends. Copies from one finished study are still on our computer and are waiting to be deleted.
The machine has a version now
Two readings can only be compared if the same machine took them. Since September 23, 2026 the engine carries a version, 1.8.1, sealed with a fingerprint, 0ff07078bb262f37.
We keep a fixed set of test pages. If a change to the machine moves any result on those pages, the version number must change. The census will be read under one version. Then a change in the count means the web changed, not the ruler.
The readings we hold so far came from 40 builds of the engine, and 361 of the 78,807 sites in the count were read before the version existed. No single version has yet read every site. Different versions of the machine read different sites, so every share here shows today, not a change over time.
The count needs a register, not a map
A census has to know how many businesses exist before it can say how many have a website. A map cannot tell us that. The state can. Florida publishes who holds a license. We read the terms at the source, and they set no limit on using the file.
Florida's license files name 71,654 building trades businesses, 30,043 salon businesses and 48,127 food service businesses. That is 149,824 named businesses in three kinds of trade. In the 18 Florida counties we searched first, the map gave us 5,510 businesses with websites in the trades we score.
We do not join those license records to any website count on this page yet, on purpose. A license is not a website. Mixing the two would make every number on this page a lie. Joining them needs something both sides print: a phone number, or the license number itself. The machine finds both on a page. Since September 22 it keeps the license number as printed. It keeps the phone as well: as the number itself at first, and now as the fingerprint described above.
And do not subtract one number from the other. A business missing from our website count tells you how the map was harvested, not whether it has a website. The map only recorded a business when somebody had already written its website next to it. We tried the name match against businesses whose sites we have read ourselves, and it missed most of them.
We measured what widening the frame does. In Lee County, our home county, the map held 299 businesses with websites in the trades we score when we read it on September 3, 2026. Widened to every kind of business, hotels and golf courses and clinics and all, it held 1,054 on September 14, 2026. That is 3.5 times more, in one county.
Read again, not just once
A website that answers us is not done. It joins a line to be read again, the oldest reading first. Of the 85,095 addresses that have given us a page, 82,454 have been read on two or more different days, and nine in ten were looked at within the last 6 days. Over the last seven full days on our record, the census made about 19,129 tries a day, and 100 in 100 of them were a second look at an address already in the record.
An address that never answered is knocked on again too, on a widening schedule: two days after the first miss, then doubling, out to once a quarter. So far that comes to about 6.6 tries for every address that gave us a page.
One reading is a photograph. Reading the same sites again is what will let the census show what changed, and a reading nobody took last month cannot be taken later.
What we will not claim
- Nothing here shows that a better score earns more money. We hold scores, not sales.
- One page per site. A score describes a front page, not a business.
- A share is counted only over the sites that were asked the question. A check the machine asks of restaurants alone is never blended with one it asks of everyone.
- The machine reads without a browser. What it cannot see is written down, and how much of each market we have read is published beside the count.
What happens next, in order
- Fill the queue wide: every kind of business, Florida first, fewer restaurants.
- Read a thousand sites twice under the same version, to learn how steady the ruler is.
- Get the government's own count of businesses by county, so the count has a bottom line that is not ours.
- Join the license register to the readings, and publish the first county census: Lee County, every licensed business, with or without a website.
Take the numbers
Two files carry these counts, and they are yours to use under the Open Database License. Download them as a spreadsheet file, census-counts.csv, or as data, census-counts.json. Both are built with this page, from the same records. Both carry its version, 8a841c0e39.
They hold totals only: the count by the state we searched, by trade and by grade, and why addresses never gave us a page we could count. No site is named. No address is in them. Any group under 30 sites is folded into other. Read them the way this page asks you to read it: as a sample, shaped by the map.
Contains information from OpenStreetMap, which is made available here under the Open Database License (ODbL). If you use the counts, credit OpenStreetMap and True Measure Inspector, and say they are under the Open Database License. Keep the notice in the files as it is, including its line for Canada's boundary map and that map's license. A database you build from them and publish goes under that license too.
Both files include the search template we sent OpenStreetMap and the table we use to turn each map label into a trade. We do not yet publish the list of search areas or the date each one was last read; write to hello@truemeasurewebdesign.com if you need either.
If your business is in the count
You can read your own result. Our free inspection runs the same 40 checks on your site in a few seconds, and prints what it looked at next to every finding. If your site is one of the good ones, the report says so.