Second step and first checkpoint. Determine the agency's stage 0–5 and where the gap is in the client's economy. The stage governs the depth of all subsequent steps (“the earlier the stage — the less of everything”).
Read ${CLAUDE_PLUGIN_ROOT}/reference/agency-stages.md (sections 1–5). Input — Company Snapshot
from step 1 (from Notion or from context).
Process
- Stage (cold-audit hypothesis). Match the proxy signals from the Snapshot against the stage table and
cold-audit signals (section 1 of the reference). Determine the most probable stage 0–5 and give
confidence level: low / med / high with 2–4 reasons “why exactly this stage”.
- If the signals are contradictory — name two adjacent stages and what would distinguish them.
- ARR is not directly visible → do not invent a number; work with proxies (presence, cases, team, channels, offers).
- Team composition from LinkedIn (section 11 of the Snapshot) — the main proxy for stage: 0 visible Sales/Marketing/SDR/AM = founder-led → Stage 0-1; appearance of separate bizdev roles → 2; function heads (Head of Sales/Marketing/CS) → 3-4; C-level with separated functions (CRO/CMO/CCO) + practice leads → 5. This overrides the site's claims: if the site says “20+ years / 200+ projects”, but LinkedIn shows 1-2 people and the entity is new — the stage is based on the real team, and mark the claim as “the founder's personal experience / a different legal entity”.
- Headcount as a proxy (entire staff, not just BizDev): up to 5 → 0 · 5–15 → 1 · 15–45 → 2 · 45–140 → 3 · 140–400 → 4 · 400+ → 5. Cross-check via revenue: headcount × $70K ≈ ARR (corridor $60–90K per person for EE IT services). If the stated ARR and headcount give different stages — take the lower one and name the discrepancy in the assumptions: much above $90K/person = US/UK rates or a product share, much below $60K = bench / subcontracting / cheap outstaffing. ⚠️ Thousands of people — this is Stage 5 (a portfolio of practices), not “outside the model”. But first check whether the LinkedIn entity is correct (the same brand/slug) — global groups inflate headcount.
- Growth trend (6m/1y/2y) = momentum: hiring spike → active scaling (more channels can be enabled); flat/decline → caution with investments. Median tenure = team stability (low tenure + rapid growth = young org; high tenure = maturity).
- Tie-breaker Stage 0 ↔ 1 (common mistake): presence artifacts (landing page, logos) DO NOT raise the stage — a landing page can exist at 0 too. 6+ vague services = "everything for everyone" (0), and not deliberate 2-4 test offers with metrics (1). What decides is traction + team: Stage 1 = 1-2 stable paying clients + 2-3 active channels + ARR >$300K. Otherwise (0 bizdev · delivery = 1-2 engineers · 0 cases · "everything for everyone") → Stage 0. Don't confuse "a new brand with a landing page" with Stage 1.
- SMF lens. According to section 2, identify the most likely hole (L / Cr / $ / #) for this stage and argue from the site's signals (e.g., no cases with $ → likely Cr/$ hole; empty company page + only Upwork → L hole). Assign the cheapest hole to close first.
- AGA focus. According to section 4, assign the recommended modules (M1 / M2–M3 / M4 / M5) for the stage. At Stages 4–5 we don't offer AGA: we come in as a Fractional CRO, GTM audit, or RevOps for one practice/vertical, and the buyer is a CRO/CMO or practice leader, not the CEO.
- Overshoot/undershoot (quick). Given the Snapshot, mark 1–2 clear overshoots (material/asset ahead of the stage) and 1–2 undershoots (lagging) — the full audit will be in steps 9–11.
- 3 discovery questions. Output 3 questions from section 1 of the guide to confirm the stage with the client.
GATE 1 — checkpoint
Return to the orchestrator a concise block: Stage X (confidence) · SMF hole · AGA focus · 3 questions · key cold-audit assumptions. The orchestrator will show this to the user and wait for confirmation/edits before proceeding. Do not start ICP until the stage is confirmed.
Output (Notion)
Subpage «02 · Stage Diagnostic»: stage hypothesis + reasons + confidence; SMF gap with
reasoning; AGA focus; quick overshoot/undershoot; 3 discovery questions; "Sources/assumptions" block.
Update the client page properties: Stage, Stage confidence, SMF gap, AGA focus,
Status = Diagnosed.
Definition of Done
- Stage 0–5 with confidence level and justification (ARR bracket + headcount aligned).
- SMF gap identified + "close first".
- AGA focus + 1–2 overshoot/undershoot each.
- 3 discovery questions derived for GATE 1.
- Notion updated (sub-page + properties).