Kavean Gobal — Backfilling HubSpot Data

Using Claude to backfill HubSpot data. (Backfills come up constantly — whenever a process changes or a new one launches. Instead of manually filtering lists/views, authorizing via a Private App and giving Claude context kills the manual work and produces documentation the Architects can review and pass to the client.)

How this works

1. Watch the loom below (2x speed is fine).
2. Comment at the bottom of this page — what landed, what you'd steal, what you'd do differently.
3. If this is the use case you're trying this week, record your own loom of it and drop it in the channel.

Facilitator note

Backfills hit everyone, and Kavean's is a masterclass in doing one carefully. On Harbinger, a subchannel remap (e.g. Google → Google-PaidSearch) wasn't a simple find-and-replace — it was multi-step across the Latest / Original / MQL channel fields, with automations firing off those values. Using a HubSpot Private App token (HubSpot ↔ VS Code), Claude ran it as a dry run → confirm → batch update → summary doc flow: 1,587 contacts, 3,000+ field updates. The parts to highlight: it skipped and flagged unmapped values (some "Latest subchannel = Google" rows were really email/referral) for architect review instead of guessing; it surfaced a downstream issue (a channel workflow that needed updating); and it produced an MD summary — portal IDs, mapping, filters, results — that drops straight onto the ticket. The data change plus the audit trail, in one pass.