Third step and second checkpoint. We build the ICP top-down in 3 levels. The main idea: at Stage 0-1 the agency doesn't yet know where to knock — so step 1 is not "describe the client", but "choose the market". Cold audit → this is a hypothesis from offer + client language + tech signal, which we confirm at discovery.
Read ${CLAUDE_PLUGIN_ROOT}/reference/agency-stages.md. Input — Snapshot (01) + Stage (02) from Notion.
GTM enrichment (${CLAUDE_PLUGIN_ROOT}/reference/gtm-frameworks.md §1-2): ECP-first (whom we
monetize in 1-2 q) vs ICP · beachhead criteria (winnable ≤18 months, ~100 companies, 30-50%, painkiller) ·
Market-Problem Map (Pain × Ease-of-sale × Ease-of-impl) · GTM segmentation (JTBD/pains/alternatives/values) ·
DMU + Anti-persona · validation confidence-ladder (interviews→presale).
Format of persona fields — according to Viktor's ICP template (Ideal Company / DM / User / Anti-ICP).
(This step absorbed the former 04-personas: personas are level 3 here.)
Reuse
icp-builder(outbound-engine-skills) /account-dossier(account-management-skills) for synthesis;market-research-edp(this pack) /niche-data-finder(outbound-engine-skills) — for market assessment and sizing. Everything — if installed. Named account lists — later indata-research/account-sourcing(outbound-engine-skills), not here.
Level 1 · Target markets (FIRST)
Don't jump straight to "3 markets" — first show how they made it onto the list:
1a · Wedge + required conditions (pre-criteria, BEFORE scoring). Extract the agency's real wedge from the Snapshot (strength + tech signal, not marketing claims). Formulate 5-7 mandatory gate conditions, which a candidate market MUST have for the wedge to deliver obvious ROI (examples: high volume of repetitive phone/intake/workflow tasks · already has systems for integration · fragmented market with many SMB/mid operators · clear measurable $-pain · reachable owner/ops-DM · required geo · can launch a pilot without a regulatory block). These are qualification gates, not scores.
1b · Longlist. 8-12 verticals that pass the gates.
1c · Shortlist → 3. Choose 3 that best pass the conditions + fit the wedge most precisely; explain why exactly these 3 and why the others were filtered out (one line per rejected one). And only then — scoring.
Shortlist scoring (3 markets) 1-5 on 6 criteria:
- Fit — how well it fits the agency's strength/offer.
- Pain/Demand — the acuteness and frequency of pain in the market.
- Accessibility — whether it's easy to build a list and reach them (reachable, identifiable).
- Deal $ — budget/deal size.
- Competition — how saturated it is (less = better; higher score = less competition).
- Size — market size / TAM (how many targeted accounts exist).
Σ-score (max 30) → recommend 1 primary market + 1-2 lines on "why exactly it fits this stage" (Stage 0 → choose the one that gives the fastest first 3-5 deals for case material: shorter cycle, simpler committee, direct $-narrative).
Level 2 · Account tiers
Within the primary market — 3 tiers: Tier 1 / 2 / 3 (= A / B / C), where Tier 1 (A) is the largest, Tier 3 (C) — the smallest. For each: profile, size (employees + revenue), pilot check, cycle, DM, effort. Fields from the "Ideal Company" template (size, revenue, geo, tech stack, deal size). Assign a starting tier for the stage (0-1 → usually Tier 2-3 (B/C) for fast deals; defer Tier 1 (A) — slow committee).
Level 3 · Personas
For the starting tier describe the personas using the template:
- Decision-Maker (required) and Champion/User (min. briefly). Fields: Persona profile (titles, dept, success metrics) · Transformation/JTBD (jobs → existing way → ideal solution → pains → implications → success criteria) · Buying process (awareness / consideration / conversion) · Buying context (buying committee · checklist · preferred channels · who sells complementary) · Triggers (3).
- Depth by stage: Stage 0-1 — DM (per template) + champion (briefly). From Stage 2 inclusive — full template for both personas + Consumption behaviour (top-5 sites/newsletters/communities/events/influencers). Don't invent fake top-5 at early stages — mark it "to be clarified during discovery".
- Anti-ICP: 3-5 disqualification criteria.
Stage-adaptive depth
- Stage 0-1: 3 markets → 1 primary → 3 tiers (start Tier 2-3) → 1 DM + 1 champion (briefly) + Anti-ICP. Hypothesis-grade, explicitly "validate during discovery".
- Stage 2: 1-2 validated markets; full personas (DM + User) using the full template + consumption.
- Stage 3-4: multi-market, per-segment personas + TAM hints (link to market-sizing).
GATE 2 — checkpoint
Return to the orchestrator: top-3 markets (with Σ) + recommended primary + starting tier + DM persona + confidence. The orchestrator will show the user and wait for confirmation/edits. Do not start VP/offers until the market+ICP is confirmed.
Output (Notion)
Sub-page «03 · Market → ICP → Persona»: Level 1 (table of 3 markets with scoring on 6 criteria + recommendation) · Level 2 (Tier 1/2/3 table + starting) · Level 3 (personas by template) · Anti-ICP · "Sources / assumptions" block. In Ukrainian; titles/verbatim — in the original language.
Definition of Done
- Recorded wedge + condition gates (pre-criteria), longlist 8-12 і why the shortlist is exactly these 3 (with those filtered out).
- 3 markets evaluated by 6 criteria (incl. Size) + recommended primary with justification for the stage.
- 3 tiers (Tier 1/2/3, where 1/A is the largest) with size/check/cycle + designate the starting one.
- DM persona by template (JTBD + buying process + triggers); champion — at minimum briefly.
- Anti-ICP ≥ 3 disqualifiers.
- Pains tied to verbatim from the Snapshot; confidence + discovery questions recorded.
- Notion page created; summary returned to the orchestrator.