The exact objectives, learning-phase thresholds, budget floors, creative specs and policy rules the agent works from — rendered from the same versioned knowledge module, nothing added.
Google best practice · knowledge v2026-07-13.1, verified 2026-07-13
Not connected yet — no Google OAuth in the product today. This module powers knowledge, tips and generation only; publishing states 'not connected / coming'.
Every campaign starts from a goal. Each goal maps to one Google objective with its own optimization events, required setup and learning phase.
Awareness maps to Video reach or Display campaigns, not Search.
Demand Gen is the closest native equivalent to a social engagement objective — re-verify surface availability.
| Field | Hard max | Recommended |
|---|---|---|
| Headline | 30 chars | — |
| Description | 90 chars | — |
| Path | 15 chars | — |
| Long headline· check docs | 90 chars | — |
Headline: RSA: up to 15 headlines. Same constant validateRsa() enforces.
Description: RSA: up to 4 descriptions. Same constant validateRsa() enforces.
Path: Two optional display-URL path segments.
Long headline: PMax/Demand Gen long headline.
Ship the full ratio set per placement — serving one stretched asset everywhere is penalized by delivery ranking.
Personalized-advertising policy prohibits targeting or copy implying personal hardship or identity (health conditions, financial status, etc.).
Healthcare advertising is restricted by market; many treatments/claims require certification, and prescription terms are limited.
Editorial policy blocks gimmicky capitalization, repeated punctuation, and unverifiable superlatives ('best', '#1') without third-party proof on the landing page.
Fifteen rephrasings of one message cap Ad Strength; cover distinct angles: benefit, proof, offer, brand, urgency, question. (validateRsa flags exact duplicates.)
Pinning most assets defeats RSA combination learning; pin at most a compliance-required headline.
Setting tCPA/tROAS before ~30 conversions/30 days starves delivery; start on maximize conversions.
Broad match plus smart bidding works, but only with routine search-terms review and a negative-keyword habit.
Intent-matching and literal: mirror the search query's language, lead with the concrete benefit or proof point, sentence case, no ellipses or gimmick punctuation.
Traffic: Search 'hooks' are query mirrors: put the searched noun in headline 1, the differentiator in headline 2, the offer or proof in headline 3.
Source: Google RSA best-practice docs.
Sales: PMax works when conversion values flow and feeds are clean; brand traffic will be absorbed into its numbers, so judge it on incremental conversions, not blended ROAS.
Source: Google PMax guidance; independent PMax measurement literature.
Ladder: maximize clicks (no signal) → maximize conversions (some signal) → tCPA/tROAS (stable signal). Skipping rungs is the most common self-inflicted delivery problem.
Source: Google bidding guidance.
Search fatigue is slower than social; review RSA asset performance monthly and replace 'Low'-rated assets rather than rewriting whole ads.
Source: Google asset reporting docs.
Last verified against official Google documentation: 2026-07-13 (module v2026-07-13.1). Values marked “check docs” are guidance not yet confirmed against the platform's current published specs.
Questions? hello@dotstack.io