How the DogDnB Score — and the census — work

Every ranking on this site is produced by one formula, applied identically to all 1,350 facilities. No facility has ever paid to affect its position, and none can. Here is the whole thing.

The formula

This is a standard Bayesian average. In plain English: a rating has to earn its sample size. A facility with 6 reviews is mostly scored at the market average; a facility with 900 reviews is scored almost entirely on its own rating. That’s why a 5.0★ place with a handful of reviews ranks below a 4.8★ place with a thousand — one is a promising guess, the other is a track record.

Where the data comes from

• Listing details — ratings, review counts, rating distributions, categories, hours, contact info, and review topics — are live-scraped from public Google Business profiles via Apify’s Google Maps Scraper. Every facility is re-scraped directly by its place ID on each refresh, so ratings reflect the profile as it stood on the refresh date, not a cached copy. Last full refresh: July 2026.
• Facilities marked permanently or temporarily closed on their live profile are removed at each refresh instead of lingering with stale data.
• Every listing links to its live Google Maps profile so you can verify anything yourself. We show review counts next to every rating on every page.
• Facilities are indexed within roughly 25 miles (40 km) of each metro’s downtown, in the categories pet boarding service, dog day care center, and kennel — including veterinary practices and pet stores whose public profiles list those services. New facilities are discovered through periodic Maps search sweeps per metro and added as they appear.

What we deliberately don’t do

• No pay-to-rank. There are no sponsored positions mixed into any ranked list. If we ever introduce clearly-labeled featured slots, they will be visually separate from rankings and this page will say so.
• No hidden edits. Operators can claim listings and correct factual details (hours, services, contact info) for free — but nobody can edit their rating, reviews, or score.
• No fake precision. Facilities without enough public reviews show “—” instead of a score.

Known limitations (read these)

• Google reviews aren’t ground truth. They skew positive, and motivated parties occasionally game them. Aggregates over hundreds of reviews are robust; single data points aren’t.
• A score is not an inspection. We have not visited these facilities. Use the score to build a shortlist, then tour like an operator.
• Data has a timestamp. Ratings shown reflect the July 2026 pull. A facility’s live profile may have moved since; that’s why we link to it.
• Coverage is bounded. ~50 top facilities per metro by review volume. A missing facility isn’t a judgment — smaller and newer places enter the index as their review base grows, and operators can request inclusion for free.

The disclosure census

A rating tells you how a facility’s customers felt. It does not tell you whether anyone is in the building at 3am, what shots your dog needs, or what it costs. Facilities publish some of that on their own websites and stay silent on the rest — and both halves are information. The census records them.

What we check

Twenty-seven policies in two groups. Admission: vaccines required, Bordetella lead time, fecal test, spay/neuter, temperament evaluation, whether records are verified. Operations: staff on site overnight, staff-to-dog ratio, playgroup cap, staff certification, play model, how groups are separated, time out of the kennel, solo option, housing, outdoor access, capacity, climate control, live cameras, photo updates, tours, injury policy, veterinary access, medication, feeding, availability, online booking. Plus published rates.

How a value gets on the page

1. A crawler fetches each facility’s website (homepage plus linked policy, FAQ and rate pages) and flags candidate passages for each field.
2. A person reads every candidate. Marketing adjectives, vendor boilerplate, customer testimonials and unrelated services are rejected and the rejection is logged.
3. A surviving fact is recorded with the facility’s verbatim sentence and the URL it was published on. Nothing is paraphrased into existence.
4. A field with candidates that all failed review — or with no candidates at all on a site we reached — is recorded as not published. That is a finding about a website on a date.

Four states, not two

• Published — the operator states it on their own site. Shown with the quote.
• Described by reviewers — customers report it (in review topics, or in a testimonial the facility quotes) and the operator does not say it. Weaker evidence, marked as such, never counted as published.
• Not published — we reached the site, a person read the candidates, nothing states it. Counted in the census denominator.
• Unknown — we never checked. Never shown as a blank; the row is simply absent, and the city is not in the census.

What the numbers mean

“13 of 97 Dallas facility websites state whether a person is on site overnight” is a census: a finite universe (websites we reached and read), one binary test, a fraction. The denominator is always websites read, never listings in the index, because a site we could not reach says nothing about that operator. Every figure on the site carries its denominator and date.

When a city goes live

A city is held out of search indexing until its census clears five gates: at least 40% of the proprietary fields resolved across its facilities, at least 8 published rates, at least 15 facilities with admission requirements recorded, the state licensing question answered, and a written market analysis whose every figure is recomputed from the data. A city that fails a gate on a later refresh drops back out on its own. Index-only cities remain browsable; they are simply not yet part of the census.

Change log

• September 7, 2026 — Site renamed DogDnB. Disclosure census published on facility and city pages for the markets that have cleared the gate.
• August 2026 — Disclosure census curated for Austin, Denver, Colorado Springs, Dallas and Chicago. Publish gate introduced; non-census cities set to noindex.
• July 7, 2026 — Live re-scrape of the full index via place IDs (Apify Google Maps Scraper): 35 closed facilities removed, 87 new Austin facilities added from a search sweep. Index: 1,350 facilities, global mean C=4.647.
• July 6, 2026 — Initial index: 945 facilities, 19 metros, prior weight m=25.