Cross-Source Analysis

Most non-trivial analytics questions touch more than one source: signups in the warehouse, deals in HubSpot, tickets in a support tool, pageviews in first-party analytics_events. The failure modes are joining on the wrong key, double counting the same entity, and presenting a blended number with no traceability. This skill is the recipe for doing it safely.

Account Deep-Dive Recipe (Default Order)

For any question about a specific customer account, follow this source order:

1. CRM (HubSpot) — company identity, deal stage, contacts, ARR, owner
2. Support history (Pylon/Zendesk) — open/resolved tickets, recurring issues
3. Community (Common Room) — engagement, feature requests, sentiment signals
4. Call transcripts (Gong) — customer voice, objections, next steps, risks
5. Product usage (BigQuery warehouse) — event counts, active users, feature adoption
6. Service health (Prometheus/Grafana) — error rates, latency, incidents
7. Tickets / Engineering (Jira) — bugs or feature requests tied to the account