Inputs
- A target definition — service and geography — and its conversion definition
- Live search, video-platform and community-discussion data for demand signal collection
- Verified facts and claims a business can actually cite
- Curated geo and service adjacency data for candidate expansion
Processing
Each of the seven phases above is its own deterministic process boundary with a
documented node contract, automated steps and a completion test, not a
single monolithic script. A candidate only proceeds from one phase to the next
once its own phase's quality gates are satisfied.
Outputs
- Generated content and video assets, each traceable to a real demand signal and verified fact
- Distributed, published assets across the configured channels
- Captured and attributed leads with full lineage back to the originating signal
- A cost and performance ledger once real paid actions are wired in
Intelligence
Every classification carries a commercial-intent score and a service-relevance
check computed directly against live fetched results, not a static
keyword list, so a candidate is excluded by evidence, not assumption.
Measurement
- Node acceptance status per phase, tracked on the project's own live workflow map
- Test-suite pass rate on every change (over 650 tests passing as of the last full run)
- Real quality-gate exclusions, logged per stopped candidate with a reason
Feedback loop
When the learning loop (nodes 32–37) flags a winner, that result can
trigger winner-replication (re-running the proven chain with a new format) and
one-hop candidate clustering into adjacent geographies or services, each
new candidate still has to earn its own real signal and facts before proceeding,
not inherit the winner's.