Earlier this year I built what looked like a clean little data product: coffee shops in New York and Chicago that have no website. The pitch writes itself. A café with no site is a web designer's best prospect, and the data came from public map listings, so pulling it felt like an afternoon of work.
Then I spot checked it. 51 rows sampled across the two cities, each one searched by hand. 28 of the 51 businesses that were supposedly siteless had an official website. One had a phone number with two digits transposed. A few were just closed. The list wasn't a scam, it was worse than that, it was plausible. It looked right and wasn't.
So the whole thing got rebuilt, and the rebuild is the actual product now.
The candidates come from public map data (OpenStreetMap, ODbL, attribution in every file). Across 17 cities that data flagged 612 cafés as having no website. Every one of the 612 got rechecked against live sources: the business's own domain if one exists, current directories, review pages. A Facebook page doesn't count as a website. A DoorDash page doesn't count. A chain's location page on the chain's own domain does.
Here is what the 612 turned into:
- 288 had an official website after all
- 59 were permanently closed, including one café inside a grocery store that shut its doors
- 33 couldn't be confirmed either way, and unconfirmed rows don't ship
- 84 phone numbers got corrected against current listings on the way through
231 cafés survived. New York has 97 of them out of a full list of 1,446 cafés. Brooklyn alone has 42. San Francisco, of all places, has 31 out of 377 checked. The full lists ride along in every file, 5,308 cafés total, with a has_website column so buyers can cut the data their own way.
Pricing was the easy part. The tools that rent this kind of data charge $35 to $49 a month, and the pay-as-you-go scrapers charge per record for data nobody verified. A freelancer who wants cafés in one metro doesn't need a subscription. They need a file that's right. So it's $19 per city, one time, or $49 for all 17 cities. The sample is free because after the first version of this list, "trust me" isn't a pitch I get to make.
Store: https://tidyrows.gumroad.com
Free NYC sample, the full 97-row format in miniature: https://tidyrows.gumroad.com/l/jjrflz
What I'd actually like feedback on: if you prospect local businesses, what kills a list for you? Stale rows, missing emails (this data is phone-first, emails are rare in the source and I say so on the listing), wrong category? The verification pass can check almost anything a second time. It just can't check everything.
Top comments (1)
plausible but wrong is the worst kind of bad data, nobody catches it till a customer does. are you planning to recheck on a schedule? cafes close and add sites all the time so a list thats right today drifts in a month