Crew Onboarding Map
The first-30-minutes plan for Crew, the AI dev-tool cost product path under research/crew/. A forward-looking target-state flow — Crew has no product yet — designed so every step drives toward one activation win: the Cross-Tool Reveal. It leads with the Platform Engineering Lead (primary) and carries the Startup CTO as a lighter self-serve track.
Pre-approval stop: the next action is review of this page. The flow below was confirmed in the in-conversation interview; this page is the structured approval record. Use the gates to approve, request edits, or flag concerns. Canonical markdown and downstream routing are withheld until a compiled response approves the page.
Proposed canonical outputs (written on approval): research/crew/onboarding-map.md and research/crew/onboarding-map-interview.md. Builds on the confirmed research/crew/journey-map.md and research/crew/lifecycle-metrics.md.
Summary
deliverableCrew's onboarding has exactly one job: get a new account to the Cross-Tool Reveal — a per-developer or per-team cost fact spanning ≥2 tools that no single vendor dashboard can produce — in the first session, and under 5 minutes for the self-serve CTO. Everything else (team mapping, invites, sharing, chargeback) is deferred until after that aha.
The design philosophy is reveal-first, degrade gracefully, defer heavy setup:
- Reveal-first: the flow is judged solely by time-to-first-reveal. Setup steps that don't serve the reveal are pushed past activation.
- Degrade gracefully: the connect step accepts even aggregate-only billing providers, shows an attribution-coverage % meter, and still delivers a team/aggregate-level reveal rather than blocking.
- Defer heavy setup: cost-center / team mapping — the Platform Lead's heaviest step — happens after the reveal, as an optional second step with auto-infer-and-confirm.
Both personas self-serve, with an optional sales-assist path for larger orgs (security blocker, 2–6 week cycle). The single biggest failure mode — and the one with the strongest designed recovery — is no reveal in the first session.
Scope & Decisions
approval requiredThese decisions were confirmed in the interview and govern the whole onboarding map.
Crew only (research/crew/); both personas, Platform Lead primary / CTO secondary.
Forward-looking target-state — no product exists yet; this designs the ideal first 30 minutes.
The Cross-Tool Reveal itself (the confirmed activation event). Deeper value is post-activation.
Connect degrades gracefully: accept aggregate-only providers, show a coverage % meter, don't block.
Deferred — reveal first on raw connected data; mapping is an optional second step.
Solo through the reveal; invites & Finance sharing are share-triggered right after the aha.
No reveal in first session — the activation-killer, gets the strongest recovery path.
Self-serve for both personas, with an optional sales-assist path for larger orgs.
Personas & Entry Points
deliverable| Dimension | Platform Eng Lead (primary) | Startup CTO (secondary) |
|---|---|---|
| Trigger to arrive | The Attribution Gap — Finance asks for a cross-tool breakdown; budget shock; tool sprawl | Personal bill shock ($500–$2,000/mo); the $6,000 overnight runaway; investor question |
| Discovery channel | Platform Eng Slack, Rands, PlatformCon; Vantage / CloudZero / DX content; FinOps community | Twitter/X, YC Slack, Hacker News, Google comparison articles |
| Entry model | Self-serve, with optional "talk to us" for larger orgs (security blocker) | Pure self-serve, no sales touch |
| What they need first | Cross-tool, team-level rollup answering "which team spent what?" | Per-developer reveal in <5 min: "Developer A is $800/mo" |
| Decision speed | 2–6 weeks (champion + economic buyer + Finance + Security) | 1–3 days, CTO alone |
| Does NOT need at onboarding | Chargeback enforcement on day one (showback first) | SSO, RBAC, compliance, chargeback |
Signup & Setup Flow
deliverableThe shared path to activation. Steps 1–4 are the critical line to the reveal; step 5 onward is deferred, post-activation work.
- Sign up — email or SSO. Capture
persona(platform_lead | cto) and company size with one optional question; never block on it. Firesaccount_created. - Connect ≥1 billing provider — Claude Code / Cursor / Copilot / Devin·Windsurf. Per-provider OAuth or API key. Show expected granularity (per-seat | per-key | aggregate-only) during connect so expectations are set. Fires
provider_connected. - Coverage-aware ingest — pull and normalize billing; display a live attribution-coverage % meter as data lands. Aggregate-only providers count toward spend but are flagged as "team-level only." No team mapping yet. Fires
attribution_coverage_sampled. - Cross-Tool Reveal (ACTIVATION) — the first dashboard showing a per-developer/per-team cost fact spanning ≥2 tools. CTO target <5 min; Platform Lead within first session. Fires
cross_tool_reveal_viewedwithtime_since_signup_sec,providers_count,attribution_coverage_pct. - Share / invite (post-aha, optional) — share-triggered: the first export or "send to Finance" prompts teammate/Finance invites. Fires
report_exported,teammate_invited. - Team / cost-center mapping (deferred, optional) — auto-infer groupings from billing metadata (email domains, seat assignments), present a draft the lead confirms/edits. Unlocks team-segmented dashboards and showback.
- Alerts & ritual setup (deferred) — showback-first anomaly alerts and the monthly budget-review ritual; seeds retention. Chargeback is a later expansion step.
Time-to-Value Budget
visualAn illustrative target allocation of the CTO's <5-minute (300-second) self-serve path to the reveal — a design budget, not measured data. It shows where the flow can and cannot afford friction: the connect step is the largest, riskiest slice.
| Step | Target seconds | Note |
|---|---|---|
| Sign up | 30 | Email/SSO; persona captured, never blocking |
| Connect provider | 120 | Largest, riskiest slice — OAuth/API key per provider |
| Coverage-aware ingest | 90 | Pull + normalize; coverage meter renders live |
| Cross-Tool Reveal | 60 | First per-developer/per-team cross-tool fact |
Activation Path
approval requiredActivation is a single, sharp event: cross_tool_reveal_viewed. The path and its guardrails:
The activation chain
- connect → ingest → reveal. The reveal must span ≥2 tools — a single-tool view is not the Crew wedge.
- Reveal content: a concrete, surprising fact — "Developer A: $800/mo," "test automation is 40% of Claude spend," or (Platform Lead) "Team X spent 2× Team Y on Cursor."
- Coverage honesty: the reveal always shows the attribution-coverage % so the user trusts the fact and sees what's still unattributed.
Single-provider guardrail
If a user connects only one provider, there is no cross-tool reveal. The flow surfaces a "add a second provider to unlock the cross-tool view" nudge rather than showing a single-vendor dashboard that any vendor already provides.
First Success
deliverableFirst success = the Cross-Tool Reveal. Onboarding is "done" the moment the user sees their first cross-tool per-developer/per-team fact. This deliberately stops short of "answered the CFO" or "shared with Finance" — those are conversion/retention milestones, not the onboarding bar.
Why stop here: the reveal is the first irreversible value moment (journey-map Critical Moments #1/#2) and the confirmed activation metric. Setting first success deeper would conflate onboarding with conversion and hide onboarding drop-off behind slower buying cycles — especially the Platform Lead's 2–6 week cycle.
Post-success momentum (not required for activation, but the immediate next nudges): add a second provider if not already; export/share the reveal; for the Platform Lead, map teams to unlock segmented dashboards.
Drop-Offs & Recovery
approval requiredThree onboarding failure modes, ordered by priority. "No reveal in first session" gets the strongest designed recovery.
| Drop-off | Priority | Detection signal | Recovery path |
|---|---|---|---|
| No reveal in first session | 1 (highest) | provider_connected with no cross_tool_reveal_viewed in session; thin or aggregate-only data | Walk to a guaranteed team/aggregate-level reveal; prompt a second provider; offer CSV upload; follow-up email with "your reveal is ready" once enough data lands |
| Connect failure | 2 | OAuth/API auth fails; provider unsupported; provider_connected with connect_status=failed | Inline retry; CSV/export upload fallback; provider-status guidance; "we'll email you when this provider is supported" |
| Single-provider stall | 3 | Exactly one provider_connected, no second within session | Persistent "add a second provider to unlock the cross-tool view" nudge; list the user's likely other tools |
Instrumentation Needs
deliverableOnboarding reuses the confirmed lifecycle-metrics event inventory; no new canonical events are introduced. The onboarding-specific funnel and signals:
| Signal | Built from (confirmed events) | Why it matters |
|---|---|---|
| Onboarding funnel (signup → connect → reveal) | account_created → provider_connected → cross_tool_reveal_viewed | The core activation funnel; locates the drop-off step |
| Time-to-first-reveal | cross_tool_reveal_viewed.time_since_signup_sec | The headline onboarding metric; CTO <5 min target |
| Providers-connected distribution | provider_connected count per account | Single-provider stall detection; cross-tool wedge health |
| Attribution coverage at first reveal | cross_tool_reveal_viewed.attribution_coverage_pct, attribution_coverage_sampled | Was the reveal real? Gates activation quality |
| Connect success rate | provider_connected.connect_status, granularity | Connect-failure drop-off; per-provider API reality |
| Post-activation momentum | report_exported, teammate_invited | Share-triggered expansion into multiplayer |
Product Gaps
open| Gap | Why It Matters | Resolution Path |
|---|---|---|
| No product or analytics layer exists | The entire flow is a target-state design | Build the connect + ingest + dashboard MVP; this map is the flow seed |
| Per-provider billing-API granularity unvalidated | Determines whether step 2→4 can produce a per-developer reveal at all | Technical spike per provider (Claude Code, Cursor, Copilot, Devin/Windsurf) |
| Aggregate-only degradation UX undefined | The graceful-degrade promise needs a real coverage-meter + team-level reveal design | UX variation pass on the coverage-aware ingest + reveal screens |
| Team auto-infer heuristic unproven | Deferred mapping relies on inferring teams from billing metadata | Validate email-domain / seat-assignment grouping against real billing exports |
| Second-provider nudge timing | Too early annoys; too late loses the cross-tool aha | Design + later A/B once instrumented |
Next Steps
routingRouting is proposed; it activates only after this page is approved.
/conversion-map research/crew— the activation→paid / team-rollout path that begins right where onboarding's first success ends./metrics research/crew— commit firm onboarding-funnel targets (time-to-reveal, activation rate, connect success) left as BASELINE NEEDED in lifecycle-metrics.- Per-provider billing-API granularity spike before the connect step can be specced.
Compile Responses
Answer one or more gates, or select section feedback, then compile YAML. A complete response with every required gate answered and no unresolved negative feedback can be pasted back to the agent to approve this page and write the canonical artifacts.