leadership-report: split PQL/PEL into OSS / API / Studio buckets and add Enterprise Dashboard link

[Daniel Beer (DanielB945)]

Summary

Restructures the leadership weekly Slack post so each lead is counted exactly once across the three sections, and surfaces the Hex Enterprise Dashboard as a deep-link from the Studio Enterprise section.

Bucketing rule (locked):

Section	PQLs	PELs
Open Source	(none — no PQL flow)	OSS bucket only (no product flag)
API	API + Both buckets (sum)	API + Both buckets (sum)
Studio Enterprise	Studio bucket only	Studio bucket only

Sample reconciliation on the week of Apr 27 → May 3, 2026:

• OSS PEL = 388 + API PEL (API+Both) = 51 + Studio PEL = 294 → 733 total = lead-level table count (matches).

What changed

• queries/weekly_report_combined.sql — new marketing_pels_by_product CTE on marketing.final_sf_lead_opportunity (lead-level, deduped via qualify) classifies PELs as API / Studio / Both / OSS via ltx_api / ltx_studio flags. Surfaced as J_studio_summary.wow_trends.pels_by_product. Dry-run cost: 545 MB (was 537 MB).
• n8n/render_body.js — drives bucketing off pqls_by_product and pels_by_product via two new helpers (findByProduct, sumByProduct). Block G (G_pql_pel) is still emitted by the SQL but no longer rendered so all three sections share one source of truth. Adds a STUDIO_DASHBOARDS constant containing the Enterprise Dashboard URL — the line is omitted when the array is empty.
• references/report-template.md — refreshed sample output, added the bucketing rules table, documented the STUDIO_DASHBOARDS knob.
• SKILL.md — updated section descriptions, added two new knobs (Lead bucketing, Studio dashboards), added new source table, added drill-through link, bumped scan budget.
• n8n/tldr_system_prompt.md — extends the PEL-skip rule to all three PEL lines (OSS especially noisy from product-flag tagging gaps), so the LLM never headlines PEL swings. Reference TL;DR untouched (it doesn't reference lead counts).

Verified rendered output

```
Open Source
• LTX 2 family: 699,156 downloads (↓59.7% WoW)
• LTX 2.3 official repo: 402,147 downloads (↓7.4%) · +39 likes (↓22%)
• Community feedback categories: not available this week
• Leads: 388 PELs (↑7660%)

API
• Total revenue: $55,565 (↑7.4% WoW)
Resellers: $40,094 (↑21.3%) · Contracts: $10,679 (↓7.9%) · Self-serve: $4,792 (↓32.4%)
• Top request type: Image-to-Video (131K requests, 67.8% of volume)
• Self-serve: 3,389 new signups (↓11.8%) · new paying users ↓24.4% · visitors ↓9.6%
• Leads: 44 PQLs (↓35.3%) · 51 PELs (↓3.8%)
…

Studio Enterprise
• Active orgs: 15/18 (Bosch, Indegene, Meta inactive) · Newell signed as new contract
• Total tokens: 7.97M (↑4.5% WoW)
…

Pilots (10 active)
…

• Leads: 13 PQLs (↓43.5%) · 294 PELs (↓4.5%)
• Dashboards: <https://app.hex.tech/…/Enterprise-Dashboard-…|Enterprise Dashboard>
```

Test plan

• bq --dry_run on the n8n SQL (passes, 545 MB)
• Full BigQuery run returns all 10 sections with pels_by_product populated (4 buckets: API / Studio / Both / OSS)
• Local Node harness runs parse_sections.js → render_body.js against real BigQuery output and produces the expected Slack body
• PEL totals reconcile to lead-level table count (388 + 51 + 294 = 733)
• Manual Monday run via n8n once merged

Notes for the reviewer

• Stacked on tldr prompt: ban active-orgs ratio in weekly report TL;DR #58 (aheden/tldr-ban-active-orgs-ratio). Once that merges, GitHub will auto-retarget this PR to main.
• Block G (G_pql_pel) is intentionally still emitted by the SQL but no longer rendered — see the comment in render_body.js. Easy to remove in a follow-up if we want a smaller SQL surface.
• The OSS PEL number for this sample week (388, +7660% WoW) is real but suspect: 306 of those landed on a single day (2026-04-29) all with leadsource = 'Inbound - PQL' and no product flag set. Likely a tagging gap or bulk import worth flagging to marketing separately. The TL;DR prompt already treats all PEL moves as noise so it won't be headlined.

Made with Cursor