Earlier this year, our data science team published a technical write-up about the car wash market in Dallas–Fort Worth. It was meant as a demonstration of what becomes possible when you combine modern AI with granular geospatial data — not a product, just an example.
We had a lot of interest from private equity funds and operators after the post went out, and decided to turn it into a product.
Today we're launching WashIndex v1.0. It covers every car wash in every U.S. state — about 80,000 locations and 12 million customer reviews.
What's New in v1.0
A lot more went into the second build than the first.
Every review now runs through an LLM that scores it across 55 separate dimensions — wait-time pain, damage rate, membership value, staff quality, and so on — rather than just collapsing to an overall sentiment number.
We scrape every operator's website to pull menus, pricing, and format signals, so a location can be classified as an Express Tunnel or a Self-Serve Bay rather than lumped together.
We built chain detection that links sister locations together across messy, inconsistent listings, which turns out to be harder than it sounds once operators rebrand, franchise, or use slightly different names from one listing to the next.
What We Learned Along the Way
Going from one metro to a country isn't really a multiplier problem; most of the engineering went into the long tail — sites without websites, formats that don't fit clean categories, addresses that don't geocode. The shift from demo to product was harder than the scaling.
The most interesting pattern we found in the data wasn't between brands, but within them: customer experience varies more across locations of the same operator than it does between operators.
Bespoke Analysis on Demand
An unintended outcome of the build: now that the dataset and the pipeline exist, questions that used to take weeks of analyst time come together in a fraction of that. Specific market studies, acquisition target deep-dives, portfolio comparisons across geographies — most of it falls out of what we've already built.
If you're working on something the standard views don't quite cover, send us a note.