# GTM workflow spec: - Status: draft | approved for pilot | live. Version: . Date: . Author: . Reviewers: - Decision: When happens for , gets within , and returns to . - Business owner: . Operator: . Stop-control owner: - Unit: . Roll-up rule if both: - Success outcome and window: . Baseline today: - Weekly capacity: x / = ## Systems and placement - Executing system per decision: identity , exclusions , enrichment , score , route , enrollment , writeback - Durable state (episodes, enrollment keys, run ledger): , because - Read-only design access per system: ## Signals One block per signal (at most three in a first workflow): - Signal , class - Source and event key: : (fallback hash: ) - Identity basis: - Clocks: occurred_at , observed_at , actioned_at - Expiry: . Unknown source date: research only - Evidence kept: . Episode rule: - Action ceiling: - Check frequency and cost: - Rule version: ## Identity and exclusions - Match order: . Ambiguous: hold with both IDs. Lookup error: retried, never "no match" - Exclusions before any paid step, with reason codes and source of truth: suppressed , customer , open opportunity , active conversation , internal or test or competitor or partner , excluded regions - Snapshot age at run time: - Reconciliation per stage: input = excluded by reason + held + passed ## Enrichment - Data points and fields, each with its consumer and freshness: - Waterfall per data point: stage , route , unit cost , charged on misses - Acceptance policy: - Stop and retry rules: - Spend controls and where each is enforced: run conditions , auto-run , approved view or record count , credit limits - Per-record and batch limits that no tool enforces: - AI research outputs and evaluation result: ## Scoring - Gates outside the points: - Fit conditions and window: . Engagement conditions and window: . Excluded inputs: email opens, unidentified page views - Unknown handling: - Proactive band: . Weekly band size against capacity: . Tie-breaker: - Score version stored with each decision: ## Routing - Precedence with rule IDs: - Pools and distribution: - Availability source and covering rule: - Response targets and escalation: - Fallback owner and deadline: - Policy file: . Dry run on , episodes (at least 200): . Rules that never fired and their test rows: ## Outbound - Plays: - Gate fields re-checked at enrollment: - Approval scope and expiry: - Enrollment key: ; unknown outcome handling: - Suppression path per tool and client scope, with test result: - Stop rules: - Reply ownership: - Sender paths with header evidence: - Pause and resume rule: ## CRM writeback - Contract: . Field classes: never , enrichment-owned , shared with candidates - Update gate: . Create gate: - Blank and clear handling per writer and the tool setting that enforces it: - Loop prevention: - check_writeback.py summary line: - Change log location and rollback by run: ## Cost - Meters and unit prices: - Fixed allocation and review time: - Cost per returned candidate: . Marginal per accepted by stage: . Blended per accepted: - Ceiling and its source: - Billed reconciliation: ## Monitoring - Runbook: . Stop control: - Weekly review: . Expansion rule: after four reviews, zero safety FAILs, acceptance at or above , cost within model ## Pilot and acceptance - Pilot job: records , fields , providers , per-record limit and batch limit checked after the run, no enrollment, no sends, approver - Acceptance cases run, with evidence links: . Live tests pending, with owners: ## Open decisions - : ## Proposed martech-context updates (not written) - stack.md: - decisions.md: - data-dictionary.md: