Champion Move Detection

Phase 3 — On every CONFIRMED move

Task ownership (routing). Every champion-move task or alert this phase produces is assigned to {{MQL_OWNER}}. The prior-relationship owner is not an automatic sender — surface them as a recommendation inside the task (e.g. "consider having [prior deal owner] send — they owned the relationship at [Old Account]"). {{MQL_OWNER}} decides who sends.

A. The old account (company losing a champion):

1. Prepend to account memory: [CHAMPION DEPARTED] [date] — [Name] ([title]) left for [New Company] ([new title]). Was: [qualification reason].
2. Mark the watchlist entry processed (dedup). Update the person's [CHAMPION WATCHLIST] line on this account to MOVED → [New Company], processed [date]. This entry is now a historical record — Phase 2's skip guard will never re-detect it, so the same move can only ever fire ONE MQL. Do not delete it; the record is what makes the run idempotent month over month.
3. If the account is an active paying customer and the departed person was its primary user or champion, flag it as a churn risk in the digest and in the account's state summary.
4. Compound-departure escalation: if TWO OR MORE watchlist people from the same account have left (this run or across prior [CHAMPION DEPARTED] entries), flag a critical churn risk prominently in the digest.

B. The new company:

0. Champion-qualification gate (run BEFORE everything else in Phase 3.B). The champion-move machinery — tier floor, MQL, champion-move tag — only applies to a person who actually realized value at the origin company. A move is NOT a champion move just because the person was a registered user or a named deal contact. Confirm BOTH:

   • Realized value. The origin relationship must be EITHER a Closed Won customer, OR the person must have been a materially active user there (real usage). A registered user of a free/never-activated workspace with negligible usage, or a named contact on a deal that never closed, fails this check — brand recognition ≠ advocacy; the person never saw the product deliver.
   • Origin deal is not a recent loss. If the origin is Closed Lost within the last 90 days, the person is leaving a company that just declined you — that is not a buying signal. Fails this check.

   If EITHER check fails → STOP: no tier floor, no MQL, no champion-move tag, no scoring run. Log the move to both companies' memory and, only if the person is GTM-relevant and the new company is a plausible ICP fit, post a soft FYI to the digest as a potential future warm intro from the prior deal owner — never a sales motion, never an MQL. Still carry the person forward per Phase 3.C (trajectory tracking is independent of this gate). Only when BOTH checks pass does the move proceed to scoring with the tier floor.

1. Competitor / never-prospect check FIRST. If the new company is one of {{COMPETITORS}} (or any direct competitor per your competitive knowledge) → log to memory only. No scoring, no outreach, no MQL. Also respect any standing never-prospect entries in org memory.

2. Verify the new company's employee count and funding with real sources (LinkedIn company page; enrichment for funding). Never guess funding — routing hinges on it. If funding cannot be sourced and employee count is near {{SELF_SERVE_THRESHOLD}}, classify routing as NEEDS-REVIEW in the digest instead of deciding.

3. Score the new company — this trigger behaves like every other signal-driven trigger. Load and run {{SCORING_SKILL}} in full, feeding the champion move in as the triggering high-intent, first-party signal with relationship context. The scoring skill then owns everything downstream: gates, tier, MQL routing, buying-committee enrichment, alerts, and any outreach — exactly as it does for a website visit or a hand-raise. Do NOT hand-craft a separate outreach motion here.

   • Relationship context to pass into the signal: who the person is, which customer they championed, what they did with {{PRODUCT}} there (role, usage, workflows built), their new title + company, and who owned the prior relationship. That owner is surfaced as a recommendation inside {{MQL_OWNER}}'s task — not an automatic assignment.
   • Champion-move tier floor: a confirmed champion move is demonstrated first-party history — the person has already bought or heavily used {{PRODUCT}} (guaranteed by the B.0 gate; a warm-familiarity or recent-loss move never reaches this step). Any result that would land mid-tier is raised to {{TIER_FLOOR}} (does not lift the lowest tier). Pass this floor to the scoring run.
   • ⛔ Gate precedence: the tier floor does NOT override the scoring skill's hard gates. If the self-serve gate fires ({{SELF_SERVE_THRESHOLD}}), there is NO MQL — post an FYI to the digest: a champion joining a small startup is future pipeline and deserves a personal note from the old account's owner, not a sales motion. Same precedence for vendor/partner, government, VC/PE, and competitor gates.

4. When scoring lands at or above {{TIER_FLOOR}} with no gate fired, it produces an MQL through its normal flow. This skill contributes only its provenance markers:

   • Apply the champion-move tag to the new company.
   • Ensure the champion exists as a contact on the new company with their LinkedIn URL and a note capturing the prior relationship.
   • The scoring skill's own steps handle the MQL stage move, {{CRM}} sync, buying committee, and the alert. Do not duplicate them; just make sure the champion-move story and the prior-owner relationship are present in the signal so the alert carries them.

5. Multiple movers to the SAME destination = ONE stronger signal, scored once. Never fire several MQLs for one company.

6. Prepend to the new company's memory: [CHAMPION ARRIVED] [date] — [Name] joined as [new title]. Ex-[Old Account] ([qualification reason]). Scored [tier]. MQL: [yes/no + why].

C. Carry the champion forward (track serial movers forever):

A proven mover — someone who already carried {{PRODUCT}} from one company to another — is your highest-value person to keep watching. Their NEXT move is a hotter signal than this one, not colder. The champion travels with the person, not the account.

1. Add the person to the NEW company's [CHAMPION WATCHLIST] as a live, carried-forward entry: - [Name] | [new title] | [email if known] | [LinkedIn URL] | carried-forward: ex-[Old Company] ([original qualification reason]) | last verified [date] This entry is ACTIVE for detection — next month Phase 2 verifies them against the new company and detects their move to a third company, firing a fresh MQL there. The chain continues indefinitely across every future move.
2. This applies regardless of gate outcome. Even if the move was FYI-routed or gated (no MQL fired), still carry the person forward — the tracking is about the person's trajectory, not this single company's score. A champion at a tiny startup today may be at a perfect-fit company next year, and you only catch it if you keep watching.
3. The champion-move tag on the new company (from B.4) + the active carried-forward watchlist entry are what put this company into next month's scan set. Both are required.
4. Watchlist-rebuild interaction: the {{REFRESH_WINDOW}}-day rebuild in Phase 1 refreshes an account's OWN champions/heavy users from current {{CRM}} + usage. It must NOT delete carried-forward entries — those are tracked people, not native champions of the account, and would never be rediscovered by a {{CRM}}/usage scan. Preserve all carried-forward and MOVED → processed entries across rebuilds; only refresh the natively-sourced lines.

Phase 4 — Monthly digest

End every run with ONE summary message to {{DIGEST_CHANNEL}} containing: accounts scanned, watchlist size, moves confirmed (with tiers + MQL status), FYI-routed moves (self-serve/gated), UNCERTAIN people needing human review, churn-risk flags (compound departures first), NEEDS-REVIEW routing calls (unsourced funding), and coverage gaps (accounts with no identifiable champion; customer domains with no matching workspace).

Guardrails

• Never fire the MQL flow on UNCERTAIN. False champion-move alerts erode trust in the signal.
• Never send outreach from this skill.
• Credit discipline: this fans out a LinkedIn lookup per watched person monthly. Keep watchlists tight (Phase 1 filters + {{REFRESH_WINDOW}}-day reuse) rather than checking every contact on every account.

What good looks like

A monthly run reads like a machine, not a hunch: every active watchlist person got a live LinkedIn check; every MOVED classification survived the experience-section read, the affirmative-evidence bar, and the wrong-person guard; warm familiarity never got promoted to champion; a mover into a competitor or a tiny startup produced a memory note and an FYI, not an MQL; one company with three arriving champions fired exactly one scoring run; the old accounts got churn flags (critical when compound); serial movers stayed on watch at their new company; and the digest let a human review every UNCERTAIN instead of the machine guessing. The failure modes this skill exists to prevent: a same-name stranger becoming a hot MQL, a stale headline hiding a move, the same move firing MQLs two months running, and a free-trial signup being treated like an advocate.