--- name: Cold Email ICP and Signal Design description: Child skill for defining the right person, right moment, segment, buying signal, timing thesis, buying committee map, and disqualifiers before cold outreach. Returns a persona×signal×reason-now row, a graded signal scorecard, disqualifier kill-list results, and a committee map ready for the outreach compiler. parent_id: cold_email_engagement_first_outreach depth: 1 preserve_markdown: true legacy_paths: - cold_email_outreach.icp_signal - cold_email_outreach.signal_design reference_modules: - id: cold_email_icp_signal_design.public_campaign_gate name: Public Campaign Gate summary: Portable public checklist for approving or rejecting a cold-outreach segment before any email is written. when_to_load: - When using the portable bundle outside BuildOS. - When a target segment, timing signal, or buying committee map needs a fast public-facing approval gate. path: references/public-campaign-gate.md visibility: public - id: cold_email_icp_signal_design.signal_scoring_rubric name: Signal Taxonomy Deep-Dive summary: Trigger sub-classifiers behind the inline scorecard — Holland inbound/postbound/bridgebound sub-types, Elias A/B/C trigger families, Maurya switching-trigger types, the three-taxonomy cross-validation rule, and compound-relevance multipliers. when_to_load: - When the primary signal is trigger-based and needs sub-classification before scoring. - When two candidate signals on the same segment compete and the inline scorecard ties. - When a trigger is ambiguous and the Holland/Elias/Maurya cross-check must be run. path: references/signal-scoring-rubric.md visibility: internal - id: cold_email_icp_signal_design.buying_committee_map name: Buying Committee Map summary: Consolidated role map table (Champion, Economic Buyer, User, Blocker × what each email leads with), Mobilizer/Talker/Blocker behavioral cut, the six 30MPC multithreading rules, Golden Path sequencing, and mode-specific committee shapes. when_to_load: - When approving a multi-stakeholder B2B segment that needs a stakeholder map (skip for single-stakeholder modes — founder-to-founder, recruiting, PR, customer research). - When choosing who to write first in a committee. - When classifying a reply by role and deciding the next action. path: references/buying-committee-map.md visibility: internal - id: cold_email_icp_signal_design.segment_examples name: Segment Examples summary: Weak / acceptable / strong worked segment examples in full output-contract shape, each annotated with why it passes or fails. when_to_load: - When the user provides a candidate segment and wants it compared against graded examples. - When showing what good and bad segments look like. path: references/segment-examples.md visibility: internal path: apps/web/src/lib/services/agentic-chat/tools/skills/definitions/cold_email_icp_signal_design/SKILL.md --- # Cold Email ICP and Signal Design Use this child skill when the root cold email workflow has a weak "right person" or "right moment." The job: turn a broad target idea into one persona × one narrowing signal with a verified timing thesis, an MVS-passing segment, a buying committee map, and a clean disqualifier run — before any anchor, offer, or draft work begins. Every approved segment ships the four artifacts named in the output contract. ## When to Use - The target list is broad, mixed, or based only on job titles. - The user cannot explain why this person should receive outreach now. - Reply data would be hard to interpret because the segment is too noisy. - The offer may be good, but the recipient or timing thesis is weak. - A B2B segment has no committee map and a single-contact strategy. - The signal is a static filter (firmographic / demographic only). ## Workflow 1. **Fill one row of the `## Persona × Signal × Reason-Now Schema` below.** One persona, one narrowing signal, one reason-now sentence. Reject mixed-persona lists at this step — a second persona or second signal is a second row, which is a separate campaign. 2. **Score the signal with the inline `## Signal Scoring Rubric`.** If the signal is trigger-based, or two candidate signals compete, load `cold_email_icp_signal_design.signal_scoring_rubric` to sub-classify (Holland sub-type, Elias family, Maurya switching type) before scoring. Do not start list-building below a total of 5. 3. **Run the MVS check** — Common Needs, Dominability, Viability with MVP (Underscore VC / Skok). Reject segments that fail any of the three. 4. **Map the buying committee.** For multi-stakeholder B2B, load `cold_email_icp_signal_design.buying_committee_map` and return the role map table: Champion, Economic Buyer, User, Blocker, Adamson class tags, Golden Path, first outreach target. For single-stakeholder modes (founder-to-founder, recruiting, PR, customer research) use the collapsed shapes in that reference's mode table — no full map needed. 5. **Verify the timing thesis.** Cross-check the trigger with the three-taxonomy rule in the signal taxonomy reference (Holland sub-type vs Elias family vs Maurya switching type). If they disagree, the trigger is ambiguous — re-test before approving. 6. **Name the buyer-progress thesis** (Moesta): what struggle, current workaround, or desired progress makes this signal matter now. This is the substance behind the reason-now column. 7. **Tier the segment** (Mark Roberge): Green (PMF measured), Yellow (experiment only), Red (do not write). Reject Red segments. Run Yellow segments only as one-variable experiments. 8. **Run the inline `## Disqualifier Kill-List`.** Any NO in the hard-kill table blocks campaign mode and sends the segment back to research. Any NO in the downgrade table converts the campaign to a low-volume one-variable experiment. 9. **Return the four artifacts and the verdict** per the output contract: persona×signal×reason-now row, signal scorecard, kill-list results, committee map. ## Persona × Signal × Reason-Now Schema The skill's core deliverable. Fill in one row per campaign — one row = one campaign, always. Template (every cell is required; an empty cell blocks the segment): | Column | Fill in | Validity rule | | ---------- | ------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------- | | Persona | One role + seniority + tenure band + company filters | More than one persona → split into separate rows/campaigns. | | Signal | One observable event + at least one verifiable external source + date observed | No verifiable source → row invalid. Static filters (industry, title) are not signals. | | Reason-Now | Complete the sentence: "[signal] means [pain] means [action] is on the table this quarter." | If the sentence cannot be completed honestly, the segment is not approved. No stacked inferences (one hop maximum). | Completed example row: | Persona | Signal | Reason-Now | | ------------------------------------------------------------------------------- | -------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ | | CTO, hired in last 30–60 days, at Series A B2B SaaS, 30–150 employees, US-based | CTO joined within the last 30–60 days — source: LinkedIn job-change notification + press release, observed this week | New-in-role exec window (~70% of budget spent in first 100 days — Florin Tatulea, practitioner data) means they are auditing the stack with a political mandate, so a tooling change is on the table this quarter. | ## Signal Scoring Rubric Score the row's signal before any list-building. Two layers: type strength, then the four-dimension scorecard. **Layer 1 — type strength** (Becc Holland's four-type relevance taxonomy; rank order is practitioner consensus, not experimental): | Signal type | Strength | Rule | | -------------------------- | -------- | ------------------------------------------------------------------------------ | | Firmographic / Demographic | Weak | A filter, not a signal. Alone it never passes — add a layer or do not send. | | Technographic | Medium | Implies a current job-to-be-done; still static. | | Trigger-based | Strong | Time-sensitive. Sub-classify via the signal taxonomy reference before scoring. | **Layer 2 — scorecard.** Score each dimension 0 / 1 / 2; total is 0–8. (Score bands and day windows are internal defaults, not sourced thresholds.) | Dimension | 0 — weak | 1 — acceptable | 2 — strong | | --------------- | ----------------------------------------- | ----------------------------------- | ------------------------------------------------------------------ | | Freshness | >90 days old, or static | Within 30–90 days | Within 30 days, or actively unfolding | | Reliability | Inferred or auto-generated; unverifiable | Single source, manually verified | Multiple sources confirm, or the prospect themselves stated it | | Pain likelihood | Speculative connection to a buying motive | Plausible link to a recognized pain | Buyer language or behavior directly confirms the pain | | Testability | Reply data uninterpretable; mixed segment | Readable on a 25–50 person batch | Readable on a sub-25 person batch (high-conviction micro-campaign) | **Threshold rule: total must be ≥ 5 before list-building begins** (internal default). Verdict bands: - 0–2 → `do_not_send`. Back to research. - 3–4 → `low_volume_test_only`. One-variable experiment, no list-building beyond the test batch. - 5–6 → `volume_approved`. List-building may begin. - 7–8 → `high_conviction`. Manual per-account outreach and a Mafia-Offer-quality hook; do not waste it on volume automation. ## Disqualifier Kill-List Binary kill-questions. Run every candidate segment through both tables after scoring. **Hard kills — any NO blocks campaign mode and returns the segment to research:** | # | Group | Kill question (answer YES to survive) | | --- | --------- | ----------------------------------------------------------------------------------------------------------------------- | | 1 | Fit | Does their tech stack support the product (or an integration path)? (Murphy: technical fit) | | 2 | Fit | Does the product solve their actual job-to-be-done? (functional fit) | | 3 | Fit | Do they have the time, headcount, and budget to deploy? (resource fit) | | 4 | Fit | Does their team have the skills to use the product correctly? (competence fit) | | 5 | Fit | Do they know what good looks like in this category — not a first-time buyer judging on wrong criteria? (experience fit) | | 6 | Fit | Does how they work match how the product works? (cultural fit) | | 7 | MVS | Do the segment's customers share one single dominant need? | | 8 | MVS | Is the segment small enough to dominate, not just serve? | | 9 | MVS | Can the current product serve the segment without divergent customization? | | 10 | Signal | Is the signal more than a firmographic or demographic filter? | | 11 | Signal | Is the signal fresh — within 90 days (internal default)? | | 12 | Signal | Is the signal verifiable from at least one external source? | | 13 | Signal | Does the signal have a plausible link to a recognized pain? | | 14 | Signal | Will reply data from this segment be interpretable on a 25–50 person batch? | | 15 | Committee | Is the committee size estimable from segment data? (B2B only) | | 16 | Committee | Is there at least one identifiable champion role? | | 17 | Committee | Is McMahon's three-test (power + personal win + will fight for you) plausible for that champion role? | | 18 | Committee | Is the likely Blocker named, not ignored? | | 19 | Committee | Is a mode-appropriate Golden Path chosen (Top-Down or Bottom-Up)? | **Downgrades — any NO converts the campaign to a low-volume, one-variable experiment with explicit notes:** | # | Group | Question (answer YES to keep campaign mode) | | --- | ------- | ----------------------------------------------------------------------------------------------------------- | | 20 | Stage | Is an LIR target defined for the segment — "P% of customers achieve E events every T days" (Roberge)? | | 21 | Stage | Is the segment tier named, and is it Green? (Yellow → experiment only; Red → hard kill, return to PMF work) | | 22 | Abel | Is the segment cleanly SMB-shaped or enterprise-shaped — no "mid-market" framing? | | 23 | Abel | If enterprise ACV, is the pricing assumption at least $75–150k landed (Jen Abel)? | | 24 | Persona | Does the segment lead with a switching trigger and desired outcome rather than demographics (Maurya)? | | 25 | Persona | Are the current solution and its pet peeves / workarounds / struggling moments named? | ## Output Contract Deliverables — every approved segment returns all four artifacts: (1) the completed persona×signal×reason-now row, (2) the signal scorecard with total and verdict, (3) the disqualifier kill-list results, (4) the buying committee map (full for multi-stakeholder B2B; collapsed shape otherwise). ```yaml persona_signal_reason_now: persona: { role: string, seniority: string, tenure_band: string } signal: one observable event + source + date reason_now: '[signal] means [pain] means [action] is on the table this quarter' segment_name: string company_filters: list of strings signal: description: string source: at least one verifiable external source type: firmographic | demographic | technographic | trigger_based trigger_sub_type: inbound | postbound | bridgebound | null trigger_family: A_awareness | B_bad_experience | C_change | null switching_trigger_type: bad_experience | change_in_circumstance | awareness | null freshness_days: integer why_now: one-sentence thesis linking signal → pain → buying motion buyer_progress_thesis: current struggle or workaround this signal suggests signal_scorecard: freshness: 0 | 1 | 2 reliability: 0 | 1 | 2 pain_likelihood: 0 | 1 | 2 testability: 0 | 1 | 2 total: 0-8 (must be >= 5 before list-building) kill_list: hard_kill_fails: list of failed question numbers (empty = pass) downgrade_fails: list of failed question numbers mvs_check: common_needs: pass | fail dominability: pass | fail viability: pass | fail buying_committee: expected_size: integer roles: list of role objects (champion | economic_buyer | user | blocker, with adamson_class tags) golden_path: top_down | bottom_up first_outreach_target: role mobilizer_probability: high | medium | low segment_tier: green | yellow | red verdict: do_not_send | low_volume_test_only | volume_approved | high_conviction open_questions: list of unresolved items offerlab_prompt: optional follow-up if the offer is still unclear ``` Stop conditions: stop and return immediately when (a) the persona×signal×reason-now row cannot be completed, (b) the scorecard total is ≤ 2, or (c) any hard-kill question fails — do not continue polishing a dead segment. ## Guardrails - Do not approve a mixed-persona list. - Do not treat title match as timing. - Do not invent buying signals or stack inferred triggers. - Do not start list-building below a scorecard total of 5. - Do not declare a segment ready if reply data would be illegible on a 25-50 person batch. - Do not approve a multi-stakeholder B2B segment without a committee map. - Do not approve a segment with a stale (>90 days) trigger as a "now" pitch. - Do not call decorative AI-generated openers "personalization." Grade them as relevance with the four-type taxonomy. - Do not approve a Red-tier segment for campaign mode. Send back to PMF work. - Do not run a campaign-mode plan on a Yellow segment — convert to a one-variable experiment. - Do not skip the McMahon three-test on the champion role for high-value strategic outreach. ## Routing Back to the Root When this skill returns, the root cold email workflow should pick up at the next stage: - If the segment passes with `volume_approved` or `high_conviction`: route to `cold_email_offer_lab` (if offer is weak) or `cold_email_outreach_compiler` (if offer is ready). - If the segment passes with `low_volume_test_only`: route to `cold_email_outreach_compiler` with mode set to single-variable experiment. - If the segment fails: return the kill-list results and open questions to the user. Do not pretend the segment is shippable. ## Examples ### User has a list with mixed personas - The persona×signal×reason-now schema rejects the row at step 1; MVS fails Common Needs. - Return a split proposal: separate Persona A and Persona B into two rows / two segments. - Recommend running the strongest sub-segment first. ### User has a clear ICP but no signal - Run the inline scorecard. Score 0-2 on every dimension. - Verdict: `do_not_send`. - Recommend a research pass to find a Bridgebound trigger before campaigning. ### User has a strong trigger but no committee map for a B2B segment - Approve the signal. - Block on kill-list questions 15-19. Return the role map table from [references/buying-committee-map.md](references/buying-committee-map.md). - Request the committee shape before approving campaign-mode outreach. ### User has a high-conviction segment - Scorecard totals 7-8; approve as `high_conviction`. - Recommend manual outreach per account, not volume automation. - Suggest a Mafia-Offer-quality hook and Top-Down Golden Path. ## Source Lineage This skill draws from: - **Ash Maurya** — switching-trigger taxonomy (bad experience / change in circumstance / awareness event), job-based ICP (trigger × outcome × current solution), three-bucket interview extraction. - **Craig Elias** — A/B/C trigger families, first-in 75% win advantage, three-event buying model, monitoring surfaces. - **Becc Holland** — four-type relevance taxonomy (firmographic / demographic / technographic / trigger-based), three trigger sub-types (inbound / postbound / bridgebound), "send to two people" test. - **Lincoln Murphy** — seven ICP dimensions (Ready / Willing / Able / Success Potential / Acquisition / Ascension / Advocacy), six fit types. - **Mark Roberge** — Leading Indicator of Retention formula, Green/Yellow/Red segment tiering, Quality × Engagement grid. - **Underscore VC / Michael Skok** — Minimum Viable Segment (Common Needs / Dominability / Viability). - **30MPC + Brent Adamson + John McMahon + Gartner** — buying committee map, Golden Path, Mobilizer/Talker/Blocker, McMahon Champion test, 6.8-buyer committee data. - **Jen Abel** — no-mid-market rule, tier-1 logos as early adopters, discounting as inverted disqualifier signal. - **April Dunford** — alternatives-naming, trade-off-led positioning as input to segment definition. - **Bob Moesta** — buyer progress, struggling moments, current workaround, and demand-side timing. - **Rob Fitzpatrick** — false-positive avoidance and past-behavior evidence for research-mode outreach. Scorecard point bands, the ≥5 list-building threshold, and freshness day-windows are internal defaults (assembled, not sourced). Full provenance is in the source analyses under `apps/web/src/content/blogs/source-analyses/` and the root [source-map.md](../cold_email_engagement_first_outreach/references/source-map.md).