Deep research, first pass: verified local supply

Verdict, one page

Three things are now clear enough to act on, and one of them is bad.

First, the shape we have been designing is a graveyard shape. Every consumer-facing attempt to sell local discovery itself is dead or shrinking, and the survivors all monetise a transaction instead: a voucher, a booking, a delivered order, an ad auction tied to purchase intent. Foursquare, which raised over four hundred million dollars and reached a hundred million in annual revenue on enterprise data, was sold to Infillion in September 2026 and its equity holders reportedly received nothing.

Second, verification is real but very expensive per record, and the only sensible way to sell it is per record rather than per query. A human field visit prices at ten to fifteen dollars to the shopper and forty to one hundred fifty to the client. At a plausible wholesale value of two cents per query, a twenty-dollar verification needs about a thousand paid lookups to break even, which the long tail of small businesses will never produce.

Third, the one thing genuinely unowned is what we set out to own: dated, owner-confirmed freshness. Google's verification confirms that a business exists and that you control it, and then it stops. The best-resourced commercial places dataset publishes a real-and-open rate of seventy percent. Nobody publishes a validation rate, a re-check cadence or a remedy.

Our recommendation is therefore not to launch the product we described. It is to spend ninety days measuring three unknowns with real money changing hands, and to stop if any of them fails.

0published prices for owner-confirmed dated local data, anywhere

70%real-and-open rate published by the best commercial places dataset

$10-15paid to a human per verification visit, $40-150 charged to the client

1,000paid queries needed to repay one $20 verification at 2 cents each

Question	Answer
Run the pilot?	Yes, but as a measurement, not a launch. No consumer product before the numbers exist.
What the pilot tests	Owner re-confirmation without a nudge; whether an owner pays for a decision; whether any business buyer pays per record.
What it must not test	Whether consumers like a nicer search. That question has a graveyard behind it and does not decide the business.
Money in the pilot	A real price asked from at least one owner and one business buyer. Free pilots teach nothing about willingness to pay.
Fundable or small?	On this evidence, the fundable version is verified supply plus a transaction or a licence, not verified discovery.

1. The dead, and the pattern across them

Every consumer-facing attempt to own local discovery died, shrank or was absorbed, including the ones with the best funding and the best brand. The causes cluster into three: discovery is not what a merchant pays for, the per-town cost of keeping information good never fell, and the value that did accumulate was in enterprise data rather than in consumers.

Company	Years	Money	What happened	Cause
Foursquare	2009 to 2026	Over $411M raised; enterprise revenue above $100M in consecutive years	App sunset December 2024; acquired by Infillion September 2026, undisclosed value. Founder states equity holders received no return.	Consumer discovery never monetised; the value was enterprise location data, and it was sold on. Verified by us.
Google Neighbourly	2018 to 2020	Internal	Shut down; lessons folded into Maps Local Guides	Not enough growth. Single source category, official help page.
Zenreach to Adentro	2012 onward	About $107M over six rounds; $210M valuation in 2018	Rebranded Adentro; now sells guest-WiFi capture and in-store measurement	Capture commoditised; discovery abandoned for measurement. Funding figure from an aggregator, treat as estimate.
LivingSocial	2007 to 2016	Valued at $6B at peak	Bought by Groupon in 2016 for approximately zero; office closed	The deals model collapsed and the assets were worthless.
Groupon	2008 onward, dying	Public	FY2025 revenue $498M, net loss $81M, 16.2M customers	Deals margin eroded; flat revenue and deepening losses.
Patch	2009 onward	AOL investment estimated above $300M, unverified	Majority sold to Hale Global in 2014; now AI-generated newsletters	Hyperlocal ad revenue could not cover per-town editorial cost. The rescue cut cost, it did not raise rates.
Belly	2011 to 2018	Total not found	Assets acquired by Mobivity	Tablet loyalty for small business was replaced by loyalty inside the point of sale.
Nextdoor	2011 onward, shrinking	Public	FY2025 revenue $258M, plus 4 percent; weekly actives down 5 percent to 21.0M; net loss $54M	Neighbourhood adjacency does not convert into enough revenue per user.

The pattern we would defend: the platforms that monetised attention or footfall are gone or shrinking; the ones that monetise a transaction, a booking, a voucher or an ad auction tied to intent are alive and priced. Patch is the clearest tell: its rescue did not raise local ad rates, it removed editorial cost with automation.

2. The living, and what each actually sells

Player	What it sells	Who pays	Scale, dated
Yelp	Local ads and services leads; reservation product	Small business advertisers, pay per click	FY2025 revenue $1.46B, net income $146M
Google Maps and Business Profile	Free listings; auctions on top; places API sold to developers	Advertisers and developers	Place data $5 to $25 per thousand calls; grounding with Maps $25 per thousand prompts
Meituan and Dianping	Vouchers, in-store listings, delivery, field-army onboarding	Consumers and merchants	FY2024 revenue RMB337.6B, 14.5M active merchants, commission roughly 15 to 25 percent
Too Good To Go	Surplus food bags; business subscription	Consumers and about 200k businesses	DKK 725M revenue 2024; 120M+ registered users
Reservations	Bookings and guest CRM	Restaurants and hotels	Amex bought Tock for $400M, owns Resy; DoorDash bought SevenRooms for $1.2B
Loyalty	Chain loyalty, now inside point of sale	Restaurant chains	PAR bought Punchh for $500M; Fivestars absorbed by SumUp
Local newsletters	Sponsor-supported daily email	Local and national advertisers	Axios Local: 43 cities, 2M subscribers, still not profitable; 6AM City pivoted to selling newsletter revenue for others
Adentro	Guest WiFi capture, audiences, paid media, walk-through measurement, venue sponsorship	Chains, malls, stadiums; resold through Verizon and Toast	6,000+ locations claimed; case studies at 40 to 91 locations

Two things stand out. The companies that touch real transactions can afford to give data and analytics away, which is why they do. And the only company in this list selling something close to our trust claim, Adentro, reached its scale through a telephone company's bill, a point-of-sale ecosystem and a shopping-centre landlord, not by walking into shops one at a time.

3. What one live verified record costs

The verification industry prices by visit, and the human gets a small share of the invoice. The discipline around it is a supervisory layer: one coordinator covering hundreds of shoppers, plus validators who treat a report as fiction until it is checked.

Unit	Price	Source type	Label
Mystery shop, one store visit, United States	$40 to $150 charged; $10 to $15 paid to the shopper	Vendor price lists and job ads, 2026	Estimate
Retail audit, one store visit	$20 to $50	Audit vendor guide	Estimate, single source
Computer-vision audit per store per month	$10 to $30	Audit vendor price list, 2026	Estimate
Menu data upkeep, one restaurant	about £12.50 per month	Digitisation vendor target	Estimate, single source
Field data collection app, one seat	$43 per user per month	Field tooling vendor	Estimate, single source
Full certification audit, food safety	$8,000 to $20,000 initial, 2 to 3 days on site	Certification scheme and vendor material	Estimate
Owner-confirmed, human re-confirmed local offer data	No published price anywhere	Google, Yelp, Yext publish none	Not found

The arithmetic that decides the business model. At a realistic wholesale value of about two cents per query, a twenty-dollar verification of one record needs roughly a thousand paid lookups before it repays itself. That is about eighty-three paid queries per month for a year, which the long tail of small places will never reach. Verification can therefore only be funded per record, sold once and used many times, or by the owner supplying it at no charge. our estimate

4. The wholesale layer: real prices, closed doors

The layer exists and is priced. What it is not is open to a small entrant.

Provider	Published price	Blocking term
Google Places API	$5 per thousand (essentials), $17 (pro), $20 to $35 (enterprise)	Verified by us: the terms prohibit exporting, extracting or scraping Maps content for use outside the services, and prohibit building a product whose features substantially recreate theirs
Gemini grounding with Google Maps	$25 per thousand grounded prompts	The de facto per-query price for local facts sold to developers, and it is Google's own data
Yelp Places API	$229 to $643 per month plus $5.91 to $14.13 per thousand calls	Terms forbid use on behalf of a third party and any sublicensing of content; liability capped at $100
Foursquare Places API	$9 to $36 per thousand	Same vendor has released 100 million places free under an open licence, which sets the commodity floor at zero
SafeGraph	Custom, enterprise; all-US places about $150,000 per year	Enterprise sales only; accuracy self-published at 70 percent real-and-open (verified by us)
Overture Maps, OpenStreetMap	Free	The floor that any paid place data now competes with

The one real payment for local facts by an AI company. Yelp licensed reviews, ratings, photos and business data to OpenAI, announced 23 July 2026, non-exclusive, financial terms undisclosed, with a quote-request flow inside ChatGPT. Verified by us across four independent outlets. It is a bulk licence, not a per-query market with a price list we can plug into.

5. Freshness: the gap is real, the willingness to pay is unproven

Google's verification confirms that a business exists and that you control it, and stops there: prices, availability and opening hours carry no freshness guarantee at all. Microsoft, Yelp and others do not publish a validation rate either. So a dated, owner-confirmed record is genuinely unowned.

What is missing is any evidence that anyone pays more for freshness on its own. No major vendor publishes a tariff where the same fields cost more purely for a faster refresh. Richness is monetised indirectly: Google's enterprise tier costs four to five times its essentials tier. And the free releases by Foursquare, Overture and OpenStreetMap push basic place data to zero. opinion Freshness is the only defensible paid axis, and nobody has demonstrated the premium.

6. The merchant: who can actually pay

Two thirds of American small businesses spend under a thousand dollars a year on marketing, and about 85 percent spend under ten thousand. The paying segment is a minority, and it is the established, busy business, not the marginal one.
The money that moves goes to customers and to proof that advertising worked. This is what Adentro sells, at chains of forty-plus locations, through a telephone company's invoice.
An owner will not adopt another login. Consolidation is what they buy: about 43 percent of small businesses say they would pay more for a tool that reduces the number of tools they use.
We found no published price for owner-confirmed local data sold to a merchant by anyone. That is either the opening or the warning.

7. The place: San Francisco, concretely

Fifty neighbourhood commercial districts exist and are enumerated by the city planning department, with a permitted-use map. The choice of corridor is real work, not a formality.
Citywide retail vacancy is about 7.6 percent, but the damage is concentrated: Van Ness above 50 percent, Union Square above 20 percent, downtown office vacancy at a record 35 percent.
Our reading: choose a residential corridor with a high count of independent businesses in one category, low tourist share, and neighbours who live there. Avoid downtown entirely, because vacancy is a symptom of the problem we cannot fix.
Not found: any published density threshold, from any source, at which a local product becomes useful. Any number we adopt is our own invention and should be labelled as such.

8. Premise scorecard

Our claim	Status	What the evidence says
Verified, owner-confirmed, dated supply is the defensible asset because it costs human hours	Partly survives	Nothing commercial holds dated owner confirmation: Google verifies ownership once, and the best commercial places dataset publishes a 70 percent real-and-open rate. But the commodity floor for basic place data is now zero, so only a narrow, freshness-sensitive slice can carry a price.
Owners will re-confirm weekly	Dubious, must be measured	We found no operator anywhere doing weekly owner re-confirmation. The nearest comparable, a mystery shop, is priced per visit and is not weekly. This is an assumption with no support in evidence.
Consumers return without a discount	Against	The entire dead list monetised attention or footfall and failed. Every survivor sells a transaction, a booking or a voucher.
An owner pays for a decision, not a dashboard	Half right	Owners demonstrably pay for customers and for proof that advertising worked. A decision is only sellable when it produces one of those. Two thirds of American small businesses spend under a thousand dollars a year on marketing at all.
The same record can be sold up-market, wholesale and to agents	Blocked as described	Place platforms forbid directory use and re-creation of their products. The only money that has changed hands for local facts with an AI company is a bulk, non-exclusive, undisclosed licence from Yelp to OpenAI.
One district and two categories is a sufficient launch scope	Untested	San Francisco has 50 neighbourhood commercial districts, so the choice is real, but no source publishes the density threshold at which a local product becomes useful. Any threshold we adopt is our own invention.
Labelled promotion is acceptable and paid position is not	Untested, and expensive to hold	Essentially all the money in this market is promotion of one kind or another: auctions, sponsorships, sponsored wifi. Refusing it removes the only line that has ever reliably paid for local operations.
Our own reach is enough early distribution	Against	Comparables reached scale through a utility bill, a point-of-sale ecosystem, a landlord, or a field army. None of them grew by individual conversations.
Running the company on our own agent system is a selling point	Untested, probably invisible	No evidence found that a shop owner values it. It may matter to investors and to hires, not to merchants.
The bare word cannot be owned as a brand	Confirmed	Three live registrations in the software class, 103 marks in the relevant classes, and the obvious domain held by a working platform since the 1990s.

9. Three observations that should stop us

Owner re-confirmation without a nudge falls below one in three by the third week, which would mean the supply chain cannot exist at two-person scale.
No owner, at any price we are willing to ask, will pay for a weekly decision, which would mean merchants are not the buyer.
No business buyer pays per record for dated facts, which would close the wholesale path that the per-query arithmetic already suggests is the only route to a margin.

10. What we could not find

Any published price for owner-confirmed, dated local facts. Google, Yelp and the directory vendors publish none.
Any density threshold for a local product. No source publishes one.
Any operator doing weekly owner re-confirmation. Our core supply assumption has no precedent we could find.
Any freshness premium in a published tariff. The theory exists; the market does not show it.
Any evidence that a shop owner values the fact that we run our company on agents.
Churn benchmarks for field verification teams. Not found in primary sources.

11. What only field work can settle

Whether owners confirm again without being chased. Whether an owner pays for a decision rather than for customers. Whether a business buyer pays per record. Whether one corridor and two categories produce a useful density. These four cannot be researched from outside, and they are the only questions that decide the next ninety days.

Method: this pass ran 27 searches and 12 full-page extractions directly, plus three parallel delegated passes on the graveyard, verification operations and the wholesale market, whose load-bearing claims we re-checked against primary sources (Google Maps Platform terms, SafeGraph accuracy documentation, TechCrunch on Foursquare and Infillion, four independent reports on the Yelp and OpenAI licence, Google Places pricing).
Figures marked as vendor or aggregator claims are not audited; they are named so they can be re-checked. Compiled 27 October 2026.

Verified local supply:what the evidence says, including where it argues against us