Verified assets

Established by direct inspection of the Glow repository on 16 September 2026. Each item states what it does not prove.

Whitefield business graph

• 1,139 canonical beauty businesses
• 821 directory-ready · 9 needs-review · 309 rejected
• 692 carry a phone number (84.3%)
• 687 independent, not chain branches
• 614 in core Whitefield · 242 greater · 1,025 brands · 222 duplicate pairs classified
• Median 52 Google reviews; 299 with fewer than 25

Does not prove: that any of these businesses want a listing, will accept a call, or will pay. No other Bangalore area has been mapped.

Discovery pipeline

• 510 discovery sweeps logged across 12 zones × 13 categories
• 994 / 994 place-detail fetches succeeded — zero failures, 16 retries absorbed
• Final build pass: 1,138 classification cache hits against 1 live model call
• Derived cost to build a new area: ₹3,622 (~$38) — 510 text searches at $35/1,000 plus 994 detail calls at $20/1,000
• Self-test: 12/12 pipeline checks passing

Does not prove: that paying advertisers can be acquired cheaply. Constructing a graph and monetising it are different funnels.

Product & data integrity

• 711 tests passing across 13 suites, including live RLS and migration tests
• 32 database migrations; 3,541 lines of application code
• Postgres trigger fn_guard_glow_authorship plus RLS make fabricated community activity structurally impossible
• 46 seed conversation items (30 polls, 16 discussions); 85 generated visual assets
• Only the Google place ID is stored indefinitely; all other provider data expires on a 30-day TTL

What is absent

Zero registered users, zero MAU, zero votes, zero comments, zero revenue, zero paying advertisers, zero business outreach performed, no public deployment. Fashion and jewellery supply is effectively absent (7 and 1 listings) — the built graph is beauty services.