Higher-Ed Dial Queue Builder (SDR Pod Lead Engine v1)

<quick_start> Trigger: M-F 6:45 AM ET (before morning-brief 7:30, before pod dials). Manual: "Build HE dial queue" / "Refresh pod dial list". Config: reference/verticals/higher_ed.md — lifecycle values, DIALABLE set, suppression (AE IDs, deal rules), tiering, floors, ATL/BTL, CMS signals. All knobs live there. Dependencies: HubSpot (portal 21530819) read; Google Drive (Sheet); Slack (alert). Apollo only on top-up. Output: Google Sheet HE Dial Queue — <YYYY-MM-DD> (Tier 1→2→3) + inventory status + sidecar JSON. Dialer: Nooks (imports the Sheet / HubSpot natively). NOT Salesfinity. </quick_start>

<success_criteria>

• Pull HE callable pool from HubSpot, filtered to the DIALABLE lifecycle set (keeps result < 10k cap), paginated to completion
• Apply suppression: AE-ownership (<90d), open-deal assoc, recently-closed-deal assoc, bad lead_status, Golden Rules
• Classify Tier 1/2/3 using REAL lifecyclestage values + notes_last_contacted recency (TIER1_RECENCY_DAYS)
• Order Tier 1 → Tier 2 → Tier 3; assign SDRs round-robin across the daily target (BLOCKED until roster confirmed)
• Emit Google Sheet (text/csv → auto-converts to Sheet) with contract columns; capture returned file URL
• Inventory check: if Tier 1+2 < floor, Slack-alert + flag capped Apollo top-up (≤50/day)
• (Gated) Stamp he_dial_tier + he_dial_date if the HubSpot property exists
• Emit sidecar JSON; consume ZERO Apollo/Clay credits unless top-up fired </success_criteria>

Stage 0: Load vertical config

Read reference/verticals/higher_ed.md — the single source for lifecycle internal values, DIALABLE set, AE_SUPPRESS_SET, deal-suppression rules, TIER1_RECENCY_DAYS, the inventory floor formula, ATL/BTL, and CMS signals. Keep values there, not here.

Stage 1: Query the HE callable pool (HubSpot)

MCP tool: search_crm_objects (objectType contacts). Properties: firstname, lastname, email, phone, mobilephone, jobtitle, company, lifecyclestage, hs_lead_status, notes_last_contacted, hubspot_owner_id, industry, num_associated_deals, hs_lastmodifieddate.

Filter — OR filterGroups, each AND'd with the DIALABLE lifecycle set so the result stays under HubSpot's 10k search ceiling:

• Phone present: include a group for phone HAS_PROPERTY AND a group for mobilephone HAS_PROPERTY (a contact with only a mobile must still appear).
• HE membership: email CONTAINS_TOKEN "edu" OR industry CONTAINS_TOKEN "education" (also "higher"/"university"/"college").
• Lifecycle: lifecyclestage IN [opportunity, salesqualifiedlead, 970124947, marketingqualifiedlead, 1217869329, lead, 1028712882] — the DIALABLE set. This EXCLUDES customer/Declined/Junk/Nurture at the query layer, cutting ~11.5k raw down to ~3.5k dialable (well under the 10k cap).

Pagination: page with the cursor/offset the API returns until total is covered. Guard: if total ≥ 9,900, the 10k search ceiling is in play — partition the query (e.g. by hs_lastmodifieddate ranges or per-lifecycle-bucket) and union the pages; log a warning and set sidecar status:partial if the cap is hit.

Caveat: industry is free-text and there is no email domain-suffix operator; "*.edu" over-matches (~67k) — use the bare "edu" token, then refine in Stage 2.

Stage 2: Refine + per-contact suppression

Drop a contact if any:

• Not truly HE: email does NOT contain .edu (incl. .edu.<cc>) AND industry has no education token. (Keep international .edu.au etc. Do NOT drop solely because industry is blank when the email is .edu.) Track the drop count; if > 25% of the pool is dropped here, set status:partial and log the delta vs the warehouse's ~5,312.
• AE-owned: hubspot_owner_id ∈ AE_SUPPRESS_SET = {Lex 82625923, Ron 423155215, Phil 190030668, Anthony 165812850, Anh 165815985} AND ownership/last-touch < 90 days. KEEP (dialable, do NOT drop): Robert 158872429 (works new leads), Tim 87486452 (BDR).
• Bad lead status: hs_lead_status ∈ {UNQUALIFIED, BAD_TIMING}.
• (Lifecycle skip/route-out already handled at the query layer; Nurture 970063423 if present → email-only.)
• Golden Rules (Customer-domain / Channel Partner / Device Owner): apply phone-verification-waterfall-skill/golden-rules-filter.md.
• Suppressed (Stage S gate): bdr_suppression_until IS SET AND > TODAY → drop. INCLUDE if not set or < TODAY (cooling expired). HubSpot filter: propertyName: "bdr_suppression_until", operator: "NOT_HAS_PROPERTY" OR operator: "LT", value: TODAY_ISO. Template: phone-verification-waterfall-skill "Stage S — Suppression Gate".

Stage 2b: Deal-association suppression (efficient set-based)

Build an EXCLUDE-by-deal contact-ID set, then remove those contacts from the pool:

1. Open deals: search_crm_objects objectType deals, filter dealstage NOT_IN [<closed stages>] (closed stages are pipeline-specific — fetch via get_properties on dealstage once; typically closedwon/closedlost). Collect associated contact IDs (deals→contacts associations).
2. Recently-closed deals: deals with closedate ≥ today − DEAL_COOLDOWN_MONTHS (default 12). Collect associated contact IDs.
3. Exclude any queue contact whose ID is in either set.

Rationale (per review): an associated open/recent deal is a more precise "hands-off" signal than owner alone. Set-based (query deals once) avoids per-contact lookups. num_associated_deals from Stage 1 is a cheap pre-filter (only contacts with >0 need checking).

Stage 3: Tier classification

Buckets (config): HOT = {opportunity, salesqualifiedlead, 970124947, marketingqualifiedlead}; WARM = {1217869329, lead, 1028712882}.

recently_touched = notes_last_contacted set AND (today - notes_last_contacted) <= TIER1_RECENCY_DAYS   # default 30
IF bucket == HOT AND recently_touched:  tier = 1   # ~150-200
ELIF bucket == HOT:                     tier = 2   # ~1,000-1,100 (incl. never-touched)
ELSE (bucket == WARM):                  tier = 3   # ~2,300

notes_last_contacted is a proxy (may be sequence-stamped, not a human dial). Default window 30d limits inflation. Prefer a true-dial signal (Nooks/Clari) if available.

Stage 4: Order, assign SDRs, daily cut

1. Sort: tier ASC, then notes_last_contacted DESC within tier, then hs_lastmodifieddate DESC.
2. Daily target: default 240 (4 SDRs × 60), configurable. Top target rows = today's dials; rest stay queued for fall-through.
3. SDR assignment — HARD PRECONDITION: round-robin the top target across the confirmed pod roster. ⚠️ Roster names are not yet confirmed (placeholder SDR-1..4). Do NOT emit a production sheet with placeholder owners — block the full run until the roster is set (dry-run/sample is fine with placeholders).

Stage 5: Emit the Nooks dial-queue Google Sheet

MCP tool: mcp__claude_ai_Google_Drive__create_file.

1. Build CSV text: header contact_id,name,phone,email,tier,last_touch_date,lifecycle_stage,sdr_assignment + one row per queued contact, ordered Tier 1→2→3. phone prefers phone, falls back to mobilephone. lifecycle_stage = human label. Escape commas/quotes per CSV.
2. Call: title="HE Dial Queue — <YYYY-MM-DD>", textContent=<CSV>, contentMimeType="text/csv". Per the tool, text/csv auto-converts to a Google Sheet (application/vnd.google-apps.spreadsheet) — do NOT set disableConversionToGoogleType. Optionally set parentId to the pod's Drive folder if known.
3. Capture the returned File object's id/URL → use in the sidecar and the Slack message. Nooks imports this Sheet (or a HubSpot view).

Stage 6 (GATED): Stamp HubSpot tier property

Only if he_dial_tier (1|2|3) + he_dial_date contact properties EXIST (MCP cannot create them). manage_crm_objects batch-update the top target contacts. If absent: skip silently, log "tier-stamp skipped (property not provisioned)".

Stage 7: Inventory monitor (greenfield top-up gate)

tier12 = count(tier in {1,2})
floor  = TIER12_DAILY_DIALS * 3        # default 67*3 ≈ 200 (Tier 1+2 consumption rate, NOT the 240 total target)
IF tier12 < floor:
    top_up_needed = true
    slack_send_message: "⚠️ HE Tier 1+2 low: <tier12> (< <floor>). Triggering capped Apollo top-up (≤50/day)."
ELSE:
    top_up_needed = false              # zero credits

Slack via slack_send_message (pod channel / Tim DM U0AAJUZH2PK). The ≤50/day Apollo top-up (no Clay waterfall) runs as a guarded sub-step or the inventory-monitor skill.

Stage 8: Emit outcome sidecar

Write ~/.claude/skill-analytics/last-outcome-he-dial-queue.json:

{"ts":"[UTC ISO8601]","skill":"he-dial-queue","version":"1.1.0","variant":"default",
 "status":"[success|partial|error]","runtime_ms":[est ms],
 "metrics":{"pool_dialable":[n],"tier1":[n],"tier2":[n],"tier3":[n],"suppressed_ae":[n],"suppressed_deal":[n],
            "daily_target":[n],"tier12":[n],"floor":[n],"top_up_flagged":[bool],"sheet_url":"[url]","credits_used":0},
 "error":null,"session_id":"[YYYY-MM-DD]"}

Use "partial" if the 10k cap was hit, the HE delta exceeded threshold, or a stage degraded; "error" only if no queue was produced.

Validation protocol (first runs)

1. Read-only dry run: Stages 1–4 only. Print tier counts + 10-row sample. Reconcile total vs the warehouse's 5,312; sanity-check (T1 ~150-200, T2 ~1,000+, T3 ~2,300).
2. Small write test: emit the Sheet for the top 25 rows; eyeball columns + Tier ordering + the returned URL.
3. Full run: full queue → Sheet; Stage 6 only if the property exists; only after the SDR roster is confirmed. Never run Stage 7's top-up unless the floor is genuinely breached.

Hard preconditions (block a full production run)

• SDR roster (4 names) for sdr_assignment — SDR-1 = Edgar Marroquin 93367782 (recorded). ⚠️ HOLD: do NOT build/assign a list for Edgar yet (per Tim 2026-06-09). SDR-2..4 TBD.
• AE_SUPPRESS_SET — CONFIRMED: suppress Lex/Ron/Phil/Anthony/Anh; KEEP (dial) Robert 158872429 + Tim 87486452. (No longer a blocker.)
• he_dial_tier/he_dial_date HubSpot properties (only if Stage 6 stamping wanted).
• Nooks↔HubSpot activity logging (for the downstream disposition loop).

Known gaps / scoping-call items

• Closed dealstage values are pipeline-specific — fetch once via get_properties on dealstage.
• DEAL_COOLDOWN_MONTHS (6–12), TIER1_RECENCY_DAYS (≤90), daily target (200–240), TIER12_DAILY_DIALS — confirm at scoping.
• True-dial signal (vs sequence-stamped notes_last_contacted) — needs Nooks/Clari engagement read (disposition loop, separate component).

Skill metadata

Version: 1.1 · Status: v1 (Higher-Ed, follow-up-first) · Author: Tim Kipper Integration: HubSpot (21530819) + Google Drive + Slack; Apollo on top-up only; Nooks dialer (HubSpot/Sheet-native) Tier: P1 (Core BDR Automation) · Triggers: Scheduled (M-F 6:45 AM ET) + Manual Feeds: morning-brief (7:30 AM), the disposition feedback loop, and the daily pod digest.