Crew Lifecycle Metrics

Crew has no product code or analytics yet, so this is a forward-looking instrumentation spec and a stage-health model — what to measure when Crew is built, not an audit of existing events. It leads with the Platform Eng Lead (cross-tool rollup, showback→chargeback, Finance's "who spent this?") and carries the Startup CTO as a lighter self-serve track.

Five stages drive the structure: Onboarding/Activation → Conversion → Transaction → Retention → Expansion. The pivotal moment across the whole lifecycle is the Cross-Tool Reveal within 5 minutes — connect billing, surface a per-developer/per-team fact no single vendor dashboard can produce. That reveal is the activation event and the seed of every later stage.

Three signals run across stages rather than inside one:

• Attribution coverage % — share of connected spend mapped to a developer or team. A first-class health metric, not a back-office data-quality number, because Crew's value collapses when 20–30% of spend sits unattributed. It gates activation (was the reveal real?), retention (is coverage degrading?), and chargeback-readiness (≥80% mapped).
• Native-convergence watch — Crew's primary falsifier (GitHub budget controls + cost centers, June 2026; Cohrint's expanding coverage). Tracked as a competitive metric.
• Cross-sell readiness (Crew→Trace) — a behavioral trigger event (a user asks a production-cost question), never a forced-upsell counter.

Hard numeric targets and business metrics are deliberately deferred to /metrics; this doc gives evidence-anchored directional ranges only. Measurement is owned by the founder/engineer today, with a noted future split (Growth owns funnel, Eng owns the data pipeline and attribution quality, Founder owns expansion).

These decisions were confirmed in the interview and govern the whole deliverable.

How well-supported each stage's metric definitions are by existing Crew research (journey map, ICP, expansion map). Lower bars are not weaknesses in the model — they mark where a downstream artifact (packaging, /metrics, onboarding map) must supply firm numbers.

Each stage lists entry/exit events, the headline metric for the Platform Lead (primary), the CTO (secondary) variant, leading vs. lagging signals, evidence quality, and the notional owner. Directional ranges are anchored to journey-map / ICP evidence; firm targets → /metrics.

Stage 1 — Onboarding / Activation

• Entry → Exit: account_created → cross_tool_reveal_viewed (first dashboard showing a per-developer/per-team cost fact spanning ≥2 tools).
• Headline (Platform Lead): time_to_first_cross_tool_reveal; % connecting ≥2 providers in first session. Directional: reveal within the first session.
• CTO variant: time_to_first_dashboard — directional <5 minutes self-serve.
• Leading: providers connected; connect success rate; attribution coverage % at first load. Lagging: activation rate; session-1 return.
• Failure modes: no reveal within 5 min; connect fails on aggregate-only billing APIs. Evidence: High (definition) / Unvalidated (threshold). Owner: Founder/eng (future Growth).

Stage 2 — Conversion

• Entry → Exit: activated → converted (Platform Lead: paid/team rollout; CTO: free→paid).
• Headline (Platform Lead): cfo_question_answered ("Now I can answer the CFO") + activation→team-rollout rate. Proof is answering Finance, not "saving money."
• CTO variant: aha_surprise_logged ("Developer A is $800/mo") → keep/upgrade decision.
• Leading: export/share actions; second provider added; invite-teammate. Lagging: activation→paid rate; cycle time (Platform Lead 2–6 wk; CTO 1–3 days). Evidence: High (triggers) / Unvalidated (rates). Owner: Founder/eng (future Growth).

Stage 3 — Transaction (Payment)

• Entry → Exit: checkout_started → payment_succeeded / payment_failed.
• Headline (Platform Lead): payment success rate; involuntary churn (dunning); invoice_or_po_requested (enterprise signal). CTO variant: self-serve card success; free→paid completion.
• Note: no Crew monetization/packaging artifact exists yet — seat-vs-usage undefined; these are structural placeholders. Evidence: Low (inferred). Owner: Founder/eng (future Eng owns billing pipeline).

Stage 4 — Retention

• Entry → Exit: subscription_active → renewed / churned.
• Headline (Platform Lead): monthly_budget_review_completed (the "source of truth for the finance meeting" ritual — the retention signal); anomaly_alert_action_rate (acted within one business day); attribution-coverage trend.
• CTO variant: monthly_cost_visibility_return; cost-per-output viewed ("once they see cost-per-PR, can't unsee").
• Leading (churn risk): attribution coverage dropping; alerts ignored; dashboard viewed only at renewal/crisis. Lagging: logo + net revenue retention; reactivation. Evidence: High (ritual + indicators). Owner: Founder/eng (future Eng + Growth).

Stage 5 — Expansion

• Entry: retained. Expansion events: seat_or_team_added, provider_added, showback_to_chargeback_converted, role_added, cross_sell_readiness_triggered.
• Headline (Platform Lead): seat/team/cost-center growth; showback→chargeback conversion (visibility becomes governance); role-expansion count.
• CTO variant: team growth; tools added; subscription→usage-billing transition.
• Cross-sell signal (both): cross_sell_readiness_triggered fires when a Crew user asks a production-cost question (the Crew→Trace trigger); surface the sibling only on a concrete missing answer.
• Evidence: Medium-high (triggers) / Unvalidated (rates); cross-sell has no customer interviews. Owner: Founder.

Three signals run across all stages rather than inside one.

Proposed canonical events for when Crew is instrumented. Names are snake_case; properties are suggestions, not final schema.

Standard properties on every event: account_id, user_id, persona, timestamp, plan, provider_set.

Each major claim mapped to source evidence, inference, confidence, assumption status, and decision impact.

Directional anchors only; commit firm numbers in /metrics research/crew.

Future, non-blocking measurement work. Written to tasks/record-todo.md on approval.

• Define the manual native-convergence sweep cadence + vendor source list feeding native_vendor_feature_observed.
• Validate per-provider attribution granularity (Claude Code, Cursor, Copilot, Devin/Windsurf) and record max achievable coverage % per provider.
• After monetization/packaging research, finalize the Transaction + Expansion event schema (seat vs. usage).
• Customer discovery to validate cross_sell_readiness_triggered semantics before in-product cross-sell prompts.
• Define the measurement-owner split trigger (Growth / Eng / Founder) and at what headcount it activates.

This page is confirmed. The approved post-approval route is /onboarding-map research/crew; the agent continues into it directly. The remaining items below are open follow-ups.

1. /onboarding-map research/crew — approved route, in progress. The activation events here assume a first-30-minutes flow that isn't yet mapped.
2. /metrics research/crew — commit firm baselines/targets and the broader success/business-metric framework for the "→ /metrics" and "BASELINE NEEDED" items.
3. Run monetization/packaging research before finalizing Transaction + Expansion schema.
4. Record-Todo Candidates added to tasks/record-todo.md.

Final compiled response YAML, approval_status: ready-for-agent-review. All five required gates answered, no unresolved negative feedback. Decisions are preserved here read-only.