PaidAcquisition — Paid-Acquisition Expertise System (foundation)

Created: 2026-07-17 · Owner: Marita (Marketing Ops) · Status: foundation pass complete (research + canon + first knowledge pack + RAG/agent design + first real audit). Nothing deployed — the netcup datacenter box is mid-cutover.

What this is

A method-first knowledge base for paid acquisition: design from real decisions → curate an authoritative canon → assemble a knowledge pack → design a hybrid RAG + specialized agent → prove it on a real case. Built to serve MSAI (B2B industrial) first, then ecommerce clients (TFC Cali Extrax, TGH Houston).

Read in this order

  1. 00-DECISIONS.md — the money decisions the system must get right (D1-D7). Start here; everything serves these.
  2. 01-CANON.md — the curated authoritative sources, mapped to sub-areas, each with why it matters + date.
  3. KB/ — the first knowledge pack (current best practice, sourced + dated):
    • KB/01-conversion-tracking-server-side.md — why a channel reads "zero" (D1, D2)
    • KB/02-utm-to-crm-attribution.md — click→UTM→HubSpot plumbing (D3, D4)
    • KB/03-b2b-budget-allocation.md — where the budget goes + decision gate (D5, D6)
    • KB/04-measurement-benchmarks.md — benchmark quick-reference + authorities (D6, D7)
  4. RAG-AND-AGENT-DESIGN.md — the hybrid RAG + specialized agent design + post-cutover deploy plan.
  5. First real audit (the proof): Z:\Knowledge\Clients\Multisensors AI\Context\PAID-ACQ-AUDIT.md.

Rules baked in

Read-only on live ad data · draft-and-approve · verify-and-explain · per-client isolation · every claim cited with source + date · LLM only via the Max bridge (no metered keys) · deterministic-first ("the LLM builds it, math runs it") · visual-first surfacing (dashboard tile, not a wall).

Provenance

Internal MSAI facts are sourced from read-only NAS/SharePoint pulls (the Lead Source Attribution Cleanup Tracker, Alecia's board deck, Fathom ads-meeting transcripts). External best-practice is sourced from the canon (official platform docs + measurement authorities + dated B2B benchmark reports), researched July 2026.