deal-momentum-analyzer（交易动量分析器）

<quick_start> Daily automated run (7am CST): Scheduled task pulls all open deals → scores momentum → delivers prioritized action list

On-demand: "deal momentum" → runs full analysis now "pipeline review" → same, formatted as pipeline review "which deals need attention" → filtered to at-risk only

Trigger phrases:

• "deal health" / "deal momentum"
• "pipeline review" / "stalled deals"
• "which deals need attention"
• "morning brief" / "SOD" (integrates into daily brief) </quick_start>

<success_criteria>

• Every open deal scored 0-100 on momentum
• Deals classified: GREEN (on-track), YELLOW (slowing), RED (stalled)
• Top 3-5 at-risk deals surfaced with specific next-best-action
• Weekly trend: are deals accelerating or decelerating?
• Target: recover 5-10% of stalled deals per month via timely intervention
• Zero false negatives on deals about to slip close date </success_criteria>

Architecture

SCHEDULED (7am CST)          ANALYSIS                    OUTPUT
─────────────────────────────────────────────────────────────────
HubSpot: all open deals  →  Score each deal on     →  Prioritized dashboard
CRM: activity history    →  6 momentum signals     →  Next-best-action per deal
Activity: emails/calls   →  Classify G/Y/R         →  Gmail draft (optional)
                          →  Compare to last run    →  Calendar blocks (optional)

Stage 1: Data Collection

1a. Pull All Open Deals (Owner OR Collaborator)

Use search_crm_objects (HubSpot MCP) with two OR-ed filter groups to capture both owned and collaborated deals:

FilterGroup 1 — Tim is COLLABORATOR:

• hs_all_collaborator_owner_ids CONTAINS_TOKEN 87486452
• hs_is_closed EQ false

FilterGroup 2 — Tim is OWNER:

• hubspot_owner_id EQ 87486452
• hs_is_closed EQ false

Properties to request: dealname, dealstage, amount, hubspot_owner_id, hs_lastmodifieddate, notes_last_updated, closedate, pipeline, createdate, hs_all_collaborator_owner_ids, num_associated_contacts, hs_deal_stage_probability

Sort by hs_lastmodifieddate DESCENDING. Limit 100.

For each deal, capture:

1a-bis. Deal Role Tagging

After pulling deals, tag each deal with Tim's role:

• OWNER — hubspot_owner_id = 87486452 → Tim owns the deal directly
• COLLABORATOR — hs_all_collaborator_owner_ids contains 87486452 but hubspot_owner_id ≠ 87486452 → Tim is helping close (AE-owned or SE-owned deal)

Display both types in the momentum report. Collaborator deals where an AE owns them (owner IDs 82625923 Lex Evans, 423155215 Ron Epstein, 190030668 Phillip Sandler) are deals Tim sourced or is actively supporting — they ARE his pipeline contribution and must be tracked.

Exclude from scoring only deals where Tim is NEITHER owner NOR collaborator (i.e., deals that somehow appeared but have no Tim involvement).

1b. Clari Call Intelligence (Signal 4 Data Source)

Pull recent Clari call data for contacts associated with each deal:

For each deal's associated contacts:

1. Search Clari calls: clari_search_calls(attendeeEmail=contact_email, daysBack=30)
2. For calls found: clari_get_call_summary(callId=call_id)
3. Extract:

- Call sentiment: positive/neutral/negative from AI summary - Discovery completeness: were MEDDIC questions asked? - Competitor mentions: any competitor names in summary - Action items: unresolved action items = stall risk - Last call date: recency signal for Signal 4 scoring

Signal 4 Enhancement (Call Momentum — 15 points):

If no Clari data found, fall back to HubSpot activity recency (existing logic). Never score 0 just because Clari is empty.

1c. Pull Activity History

For each deal's associated company, use ask_agent:

• Query: "Show all activity, notes, and engagement for [company] deals in last 30 days"
• Extracts: call notes, email activity, meeting outcomes, deal stage changes

1c. Pull Contact Engagement

For each deal's associated contacts, use hubspot_search_contacts:

• Last email open/click dates
• Form submissions
• Page views
• Meeting bookings

Stage 2: Momentum Scoring (0-100)

Score each deal on 6 weighted signals:

Signal 1: Days in Current Stage (25 points)

Stage medians (calibrate from Tim's historical data):

Signal 2: Activity Recency (20 points)

Signal 3: Stakeholder Breadth (15 points)

ATL/BTL Validation Required (see CLAUDE.md § ATL/BTL Classification v1.0): For each deal's associated contacts, classify by title before scoring:

• ATL: Chief, VP, Director, Dean, Provost, Superintendent, Court Administrator, City Manager, Senior Pastor, Executive Pastor
• GRAY: Manager (AV/Facilities/IT) — only counts as ATL-equivalent if confirmed budget >$25K
• BTL: Technician, Specialist, Coordinator, Support, Administrator (Systems/Network/Database), Engineer, Operator, Instructor/Faculty, Designer, Assistant, Clerk (non-Court Admin), Volunteer, Intern, Student, Resident, Help Desk
• NEVER ATL: Warehouse Manager, Network Manager, Systems Administrator, AV Technician, Graphic Design Instructor, Program Administrator, Web Designer, Classroom Support, Lab Coordinator, Maintenance, Building Engineer, Multimedia Services Manager, Video Production Specialist, Streaming Crew

Penalty flag: If a deal has 0 ATL-tier contacts, add a ⚠️ flag in the report: "NO ECONOMIC BUYER IDENTIFIED — deal at risk of stalling at Decision Maker stage. Action: ask champion to intro Director+/VP+."

Signal 4: Call Momentum (15 points)

Uses Clari call data as primary signal source (see Stage 1b: Clari Call Intelligence).

Data sources:

• Primary: Clari call summaries (via clari_search_calls + clari_get_call_summary)
• Fallback: HubSpot activity recency if no Clari data available

Signal 5: MEDDIC Completeness (15 points)

Estimate from available data:

Signal 6: Close Date Integrity (10 points)

Classification

Stage 3: Next-Best-Action Engine

For each YELLOW/RED deal, prescribe specific action:

Action Matrix

Action Output Format

For each action, provide:

1. What: Specific action description
2. Why: Which signal triggered this
3. How: Draft email / calendar invite / talking point ready to use
4. When: Today / This week / Before [close date]

Stage 4: Output Format

╔══════════════════════════════════════════════════════════════╗
║  DEAL MOMENTUM REPORT — [Date]                               ║
║  Pipeline: $[total] | Deals: [count] | Weighted: $[weighted] ║
╠══════════════════════════════════════════════════════════════╣

SUMMARY:
🟢 GREEN: [X] deals ($[amount]) — on track
🟡 YELLOW: [X] deals ($[amount]) — slowing, intervene this week
🔴 RED: [X] deals ($[amount]) — stalled, intervene TODAY

TREND vs LAST RUN:
- Deals improved: [X] (moved from RED→YELLOW or YELLOW→GREEN)
- Deals declined: [X] (moved down)
- New deals: [X]
- Pipeline delta: [+/- $amount]

═══════════════════════════════════════════════════════════════

🔴 PRIORITY ACTIONS (do these today):

1. [Deal Name] — $[amount] | Stage: [stage] | Score: [XX/100]
   ⚠️ Signals: [days in stage: 28 | no activity: 18 days | single-threaded]
   → ACTION: [specific next-best-action]
   → DRAFT: [ready-to-send email or talking points]

2. [Deal Name] — $[amount] | Stage: [stage] | Score: [XX/100]
   ⚠️ Signals: [close date overdue | negative call sentiment]
   → ACTION: [specific next-best-action]
   → DRAFT: [ready-to-send email or talking points]

═══════════════════════════════════════════════════════════════

🟡 WATCH LIST (intervene this week):

3. [Deal Name] — $[amount] | Score: [XX/100]
   → [brief action recommendation]

═══════════════════════════════════════════════════════════════

🟢 ON TRACK:

[Deal list with scores, sorted by close date]

╚══════════════════════════════════════════════════════════════╝

<scheduled_automation>

Daily 7am CST Run

This skill is designed to run as a scheduled task:

Schedule: Daily at 7:00 AM CST (13:00 UTC) Task name: "deal-momentum-daily" Scheduling options:

• Session-based: CronCreate "57 6 * * 1-5" with prompt "Run deal-momentum-analyzer" (3-day auto-expire, must re-create each session)
• Persistent: launchd agent at ~/Library/LaunchAgents/com.tim.deal-momentum.plist (future work)
• On-demand: Tim says "deal momentum" or "morning brief"

Flow:

1. Pull all open deals from HubSpot
2. Score each deal on 6 momentum signals
3. Classify GREEN/YELLOW/RED
4. Generate next-best-actions for YELLOW/RED
5. Create Gmail drafts for recommended re-engagement emails
6. Output report (saved to workspace + presented in morning brief)

Integration with SOD/Morning Brief: When Tim says "morning brief" or "SOD", this report is included as the pipeline section. </scheduled_automation>

Sibling Skills Referenced

• hubspot-revops-skill — HubSpot query patterns, pipeline stage definitions
• sales-revenue-skill — MEDDIC framework, pipeline coverage benchmarks
• meddic-call-prep-auto-skill — Generates full call prep when action = "book a call"
• portfolio-artifact-skill — Captures deal recovery metrics for GTME portfolio

Emit Outcome Sidecar

As the final step, write to ~/.claude/skill-analytics/last-outcome-deal-momentum-analyzer.json:

{"ts":"[UTC ISO8601]","skill":"deal-momentum-analyzer","version":"1.0.0","variant":"default",
 "status":"[success|partial|error]","runtime_ms":[estimated ms from start],
 "metrics":{"deals_scored":[n],"green_count":[n],"yellow_count":[n],"red_count":[n],"blockers_flagged":[n],"avg_momentum_score":[n],"drafts_created":[n]},
 "error":null,"session_id":"[YYYY-MM-DD]"}

Use status "partial" if some stages failed but results were produced. Use "error" only if no output was generated.