Marketing Factory KB

The knowledge foundation for a marketing factory that agents can run. Operational reference on modern SaaS marketing — channels, positioning, content, growth, measurement, and how agents execute each area. Built for factory operators configuring, guiding, and auditing AI marketing agents.

49 sources · 95 concepts — browse the full All Pages index or the Sources catalog.

Working an actual task? Start at the Task Router — job-to-be-done → minimal reading list, gates, and where to log results.

Start Here

  • marketing-factory — the organizing concept: systems over campaigns, product as distribution, data loops
  • agent-orchestration — the conductor layer that turns per-channel agents into a factory
  • workflow-vs-agent — the restraint principle: use the least autonomy that works
  • agent-ownership-boundary — what agents own end-to-end vs. what humans keep
  • experiment-loop — the decision protocol: every output is an experiment with a hypothesis, metric, deadline

The Layers

Execution — agent-workflow-pattern · content-machine · cold-email-sequence · geo-content-pipeline · social-media-automation · autonomous-marketing-agent

Decision — channel-selection · marketing-budget-allocation · marketing-benchmark-prior · bayesian-decision-rule · sequential-testing · minimum-detectable-effect

Measurement — marketing-attribution · closed-loop-attribution · incrementality-testing · geo-citation-measurement · brand-measurement · net-revenue-retention

Positioning & growth — positioning · category-design · message-market-fit · ideal-customer-profile · product-led-growth · growth-flywheel · bowtie-model

Data substrate — event-tracking-schema · identity-resolution · privacy-resilient-tracking · shared-agent-memory

Safety & governance — human-review-gate · agentic-failure-modes · brand-voice-drift · prompt-injection · marketing-compliance

Pre-testing — synthetic-consumer-panels · semantic-similarity-rating · calibration-loop · synthetic-pricing-research

LLM search / GEO — llm-search · geo-prompt-research · content-authority · llms-txt · schema-markup

Field Evidence

Maintained by the OpenClaw factory. Build log: compiled/log.md. Every ingest is linted (kb_shared/tools/lint.py) and logged — no pipeline run counts if it isn't logged.