USE_CASE_ID: icp_outbound_prospecting
NAME: ICP-based outbound prospecting
CATEGORY: Outbound prospecting

The user has an Ideal Customer Profile (ICP) defined by firmographics
(industry, headcount, geography, optionally tech stack) and a target role.
They want a list of verified contacts at matching companies, enriched and
pushed into an outreach sequence — starting from zero, with no existing list.

For routing, file structure, and cross-cutting principles, see
https://floqer.com/docs/use-case-catalog.txt. This file does not repeat them.

INDEX:
  1. When to use / when not to use
  2. Inputs
  3. Outputs
  4. Workflow design
  5. Implementation
  6. Best practices
  7. Common variations
  8. Failure modes and mitigations
  9. Related use cases

================================================================================
1. WHEN TO USE / WHEN NOT TO USE
================================================================================

USE WHEN:
  - The user is starting cold (no existing list of accounts or contacts).
  - The ICP can be expressed in firmographic filters (size, industry, geo,
    tech stack).
  - The target persona is a defined role (e.g. "Head of RevOps at Series B
    SaaS in NA").

DO NOT USE WHEN:
  - The user already has a list of companies — skip the discovery stage and
    jump straight to people enrichment.
  - The trigger is an external intent signal (hiring, social activity, web
    visit) — see Intent-driven outreach category.
  - The user wants to enrich existing CRM contacts — see CRM hygiene
    category.

================================================================================
2. INPUTS
================================================================================

  - ICP firmographic filters: industry, headcount range, geography, optional
    tech stack.
  - Target role(s) at the company, with seniority where relevant.
  - Outreach destination: a connected sequencing tool. One of:
    smartlead, instantly, lemlist, outreach, salesloft, reply, reachinbox,
    heyreach, salesforge, plusvibe.
  - Optional: exclusion list (current customers, do-not-contact, churned
    accounts).

================================================================================
3. OUTPUTS
================================================================================

  - A list of verified contacts (email, LinkedIn, role, company) pushed into
    the chosen sequencing tool.
  - A Floqer table snapshot of the same list for audit and re-use.

================================================================================
4. WORKFLOW DESIGN
================================================================================

High-level conceptual stages:

  1. Define candidate companies — translate ICP filters into a company list.
  2. Filter and cap — apply hard filters and trim to a sensible volume
     before any paid enrichment runs.
  3. Find target-role employees at each company.
  4. Enrich verified work email for each contact.
  5. Dedupe against the user's exclusion list and existing CRM.
  6. Push to outreach sequence with the right field mapping.
  7. Persist the final list to a Floqer table for audit and reuse.

================================================================================
5. IMPLEMENTATION
================================================================================

STAGE 1 — Define candidate companies
  Action: floqer_company_firmographics (preferred), or get_employees_by_company_using_apollo
          filtered to companies as a fallback.
  Why: Floqer firmographics is native and cheaper. Apollo is a fallback when
       a needed filter is unsupported.

STAGE 2 — Filter and cap
  Action: filter, then csv_to_structured_array_format with a hard cap (e.g. 500).
  Why: Hard cap before any per-row paid action — see cross-cutting
       principle 2 in the catalog.

STAGE 3 — Find target-role employees
  Action: get_employees_by_company_using_floqer_native (preferred) or get_employees_by_company_using_apollo.
  Why: Apply role + seniority filters here, not later.

STAGE 3 (alternative for under-indexed ICPs) — Web-agent fan-out
  Action: llm_web_agents (Floqer Deep or Parallel Core) returning a
          single-field raw_array of people, piped through
          raw_to_structured_array -> push_data_to_sheet.
  Why: Small, fragmented, or non-tech segments — independent shops,
       family businesses, government contractors, regional service
       providers, anything below ~50 headcount outside tech — have
       thin or absent coverage in Apollo and Floqer Native. Dedicated
       finders return "No data found" for many such companies, which
       errors the action and halts the row's chain (see section 8
       failure modes). The web agent reads company websites,
       Companies House, news, YouTube, and other sources B2B
       databases don't index, and applies size-tiered title logic in
       the prompt itself rather than across multiple chained finder
       actions with mutually exclusive run_if conditions.
  Mission shape: instruct the agent to return up to N most-senior
       matches with full_name, person_title, and a flagship LinkedIn
       URL; embed size branching in natural language ("if headcount <
       50, find founder/owner/MD; if 50+, find sales/BD director").
       See llm_web_agents.txt section 8 for the raw_array output
       pattern and the strict JSON formatting rules required to keep
       raw_to_structured_array parsing.

  Heuristic: prefer this path when the target list is small,
       fragmented, or under-indexed in standard B2B databases —
       independent shops, family businesses, government contractors,
       regional service providers, anything below ~50 headcount in
       non-tech verticals. Reach for
       get_employees_by_company_using_floqer_native /
       get_employees_by_company_using_apollo when targeting tech
       companies, mid-market and up, or any segment with dense
       LinkedIn coverage.

STAGE 4 — Enrich verified work email
  Action: person_work_email_waterfall, then verify with one of
          person_work_email_verification_by_neverbounce,
          person_work_email_verification_by_million_verifier,
          person_work_email_verification_by_zerobounce, or
          person_work_email_verification_by_allegrow.
  Why: Always verify after enrichment — see cross-cutting principle 3.

STAGE 5 — Dedupe
  Action: auto_dedupe_rows against an uploaded exclusion list, or
          lookup_another_floqer_workflow_row against a CRM-backed sheet.
  Why: Idempotency — see cross-cutting principle 6.

STAGE 6 — Push to sequence
  Action: One of smartleads_add_to_campaign, instantly_add_to_campaign,
          lemlist_add_lead_to_campaign, outreach_add_to_sequence,
          salesloft_add_to_cadence, reply_add_and_push_to_campaign,
          reachinbox_add_to_campaign, heyreach_add_leads_to_campaign,
          salesforge_add_to_sequence, plusvibe_add_to_campaign.
  Why: Match to the user's connected tool. Field mapping at minimum:
       email, first_name, company.

STAGE 7 — Persist
  Action: push_data_to_another_floqer_workflow.
  Why: Audit + reuse for follow-up campaigns.

DATA PASSING:
  Carry company_name and company_domain on the row from stage 1 onward so
  personalization works downstream. Reference upstream outputs as
  {{action_instance_id.field_name}} or, for structured arrays,
  {{action_instance_id.list_name.column_name}}. See concepts.txt section 5.

================================================================================
6. BEST PRACTICES
================================================================================

  - Cap before enrich. Always insert a filter + cap step before stages 3
    and 4. Without it, a wide ICP can run thousands of paid calls.
  - Verify after enrichment, every time. Unverified emails harm domain
    reputation more than they help open rates.
  - Dedupe before push, not after. Pushing duplicates into the sequencing
    tool causes silent merges or contact-limit overages.
  - Carry company context through every stage. Keep company_name and
    company_domain on the row so personalization works downstream.
  - Snapshot the list. Persisting to a Floqer table at the end makes
    follow-up campaigns and reporting trivial.

================================================================================
7. COMMON VARIATIONS
================================================================================

  - Add personalization: insert llm_web_agents (Web Agent) for
    research, then llm_models (AI Generate Content) for writing, in
    that order, between stages 5 and 6 to draft a per-contact opener.
    The web agent gathers context only (recent news, LinkedIn
    activity, company signals, mutual connections, recent hires);
    the LLM uses that context to write the opener. Keep these as
    separate actions — never conflate them. Web agent does research,
    not writing. LLM does writing, not research.
  - Add intent gate: insert a hiring-signal or LinkedIn-content-activity
    filter after stage 1 so only companies showing intent advance.
  - Multi-persona: run stage 3 once per role and union with merge_employee_finder_structured_array
    before stage 4.
  - Sample-first: for unfamiliar ICPs, run with a cap of 25 and review
    before scaling up.

================================================================================
8. FAILURE MODES AND MITIGATIONS
================================================================================

  - ICP too broad → cost spike.
    Mitigation: enforce stage 2 cap; warn the user if pre-filter company
    count exceeds 5,000.

  - Low email find rate.
    Mitigation: fall back from person_work_email_waterfall to
    personal_email_by_linkedin_url_waterfall only with explicit user
    approval. Personal email outreach has different deliverability and
    compliance implications.

  - Sequencing tool field mismatch.
    Mitigation: use a data-formatter action to align field names before
    push.

  - CRM dedupe miss.
    Mitigation: dedupe on (email + LinkedIn URL), not email alone —
    different people can share an old work email after job changes.

  - Dedicated finder (Apollo / Floqer Native) returns "No data found"
    -> halts the row's chain, no output for that company.
    Mitigation: for under-indexed segments (sub-50-headcount, non-tech,
    regional service providers, family businesses, government
    contractors), use the web-agent fan-out path described in STAGE 3
    (alternative) instead. The dedicated finders' coverage gap on
    these segments is structural, not a transient error — retrying the
    same finder won't help.

================================================================================
9. RELATED USE CASES
================================================================================

  - Intent-driven outreach — same shape, different trigger.
  - CRM hygiene → enrich missing emails — reuses stages 4 + 5 against an
    existing CRM list.
  - Research and scoring → ICP fit score — can replace stage 2's hard
    filter with a model-derived score threshold.

================================================================================