Brand onboarding — facts, products and collection

Verification status — 6 October 2026

The inventory below is historical. In particular, its automatic entity-creation description must be reconciled with the approved review requirement; it is not a new instruction to create entities automatically. The program audit tracks that conflict. The guideline audit also identifies source, draft-review and regeneration risks in local code; production behavior has not been verified. Audience research, product and feature discovery, source approval and ongoing collection still need a connected workflow and acceptance evidence.

Onboarding should establish a brand’s facts, official links, products, features and guidelines, then connect it to ongoing collection. The following capability inventory was recorded on 17 August 2026. Its statuses and counts are historical observations.

Earlier implementation inventory — 17 August 2026

Capability	Status	What it does
Entity enrichment	built	pulls a brand's basic facts — founding year, headquarters, industry, ticker, website, logo — from public sources, with a person-level pass for founders and executives that stages uncertain matches for review
Brand-guidelines pipeline	built, and it has run	a five-stage path — discover the guidelines online, scrape them, store the images, structure the sections with a model, seed them into the database. Ninety-four guideline sections across twelve brands are live from it
Entity auto-create	built	a brand mentioned in analyzed content gets an entity row and a sector tag automatically
Products & features harvest	missing	product data today is a byproduct of per-piece analysis; nothing deliberately walks a brand's site to build its product and feature map
"Onboard this brand" routing step	missing	the ruled onboarding step — decide what a newly-added brand needs and kick it off — does not exist; each piece above is triggered by hand

What a build would decide first

Understanding audiences belongs in the research as well. Joe's earlier notes call for understanding a company, its products and the audiences they address, and connecting that knowledge to the work we collect. He suggested an agent taking a first pass for a person to review. The audience requirement applies across companies, products, features and content; it does not settle where that research runs or which steps happen automatically.

The implementation inventory above is the August snapshot. The recovered requirement is not evidence that an audience-research workflow or schedule is already running.

• The order of operations — enrichment, then links, then guidelines, then products, then collection: one orchestrated run, or independent steps a person triggers?
• Verification — the guidelines pipeline captures what it finds; the ruled bar ("find it, scrape it, verify it") needs a review surface before seeded sections are treated as truth.
• Where product discovery writes — the products layer has two connected homes; the harvest must respect the ruled shape rather than invent a third.

Where to look next

• The campaign pipeline — what should light up once a brand is onboarded.
• The Products chapter — the shape the harvest must fill.
• The entities chapter — what an entity is and carries.