The listing schema, in full.

A small mobile service business, fully described.

Bob runs a plumbing shop out of a Dallas suburb. He has a Google Business Profile and a Yelp page, but no website worth scraping. An agent that wants to dispatch a plumber tonight at 2am needs to know, verifiably, that Bob is open, licensed, in range, and capable of fielding the call. The UBI listing below is what that agent reads. The same shape works for a Manhattan dentist, a Topeka feed store, or a global SaaS. Only the capabilities block differs.

facet_id: ubi:us:tx:dallas:75002:bobs-plumbing  # global stable ID
nap:
  name: "Bob's Plumbing"
  address: { street, city, state, postal, country, lat, lng }
  service_area: [{ type: radius_miles, value: 25 }]  # for mobile/service biz
  phone, email, website, social: { instagram, fb, tiktok, ... }
hours:
  regular: { mon: [09:00-17:00], tue: ..., ... }
  exceptions: [{ date: 2026-12-25, closed: true }]
  seasonal: { open_from, open_until }
  status: open_now | closed_now | seasonal | permanently_closed
category:
  naics: 238220
  google_place_type: plumber
  facet_taxonomy: [home_services, plumbing, emergency_24_7]
identity:
  claim_status: unclaimed | claimed_unverified | verified_owner | kyc_verified
  verification_methods: [postcard, phone_otp, dns_txt, gmb_oauth, stripe_kyc]
  licenses: [{ jurisdiction, type, number, verified_via, expires }]
reputation:
  facet_score: 4.6
  agent_review_count: 142
  human_review_count: 1283  # imported from Google/Yelp/BBB if licensed
  dispute_rate: 0.4%
  response_time_p50_hours: 2.3
capabilities:  # which primitives this listing supports
  - view_only      # always (Bucket 5 to 7)
  - click_to_call  # most (Bucket 4)
  - book_appointment   # subset (Bucket 2)
  - request_quote      # subset (Bucket 3)
  - purchase_catalog   # subset (Bucket 1)
provenance:
  sources: [google_places, yelp_fusion, owner_claim, web_crawl, state_license_registry]
  last_verified: 2026-04-30T18:22Z
  signed_by: kid:facet-prod-2026-04

Pseudo-syntax note. The address, social, hours.regular, hours.seasonal, and licenses blocks above use a compact key-list shorthand ({ street, city, state, ... }) for readability. Live UBI documents inflate each shorthand to fully typed sub-objects per the JSON Schema referenced below. A fully inflated JSON version of this exact example is available at ubi-sample-listing.json.

Eight blocks. One document. Globally addressable.

The schema is intentionally narrow. NAP, hours, category, and provenance are the universal Maps/Yelp baseline; identity, reputation, and capabilities are what makes a UBI listing machine-actionable rather than scrape-fodder.

Four tiers. Each unlocks one more thing the agent can do.

Listings progress monotonically through four claim_status values. Each tier is a strict superset of the prior tier's capabilities. No listing ever loses primitives short of an explicit revoke. Most businesses spend their entire lifetime in unclaimed; the ones that show up in agent search results don't.

The schema lives upstream of this page.

This page renders the listing example in marketing-friendly form. The authoritative schema, ER diagram, and JSON Schema validator live in the Facet protocol repo and version independently of the marketing site.

UBI is Facet's proprietary knowledge graph in document form.

Each listing is a flat projection of nodes and edges that live in Facet's proprietary knowledge-graph layer. nap, licenses, and category decompose into nodes; capabilities, verification_methods, and provenance.sources become typed edges. Agents that need traversal queries, "find me a kyc_verified plumber within 25 miles who hasn't had a dispute in the last 90 days", query the graph directly. The flat document is just one of many views.