Zapier → n8n migration checklist — audit, port, cut over

The savings are real: at 30,000 tasks/month, Zapier's Team plan runs ~US$100+/month versus under US$10/month self-hosted on n8n — and the n8n version gets error handling and logging Zapier never gave you. Here's the checklist we run on every migration.

Phase 1 · Inventory

• ☐ Export every Zap (Zapier's "Manage → export" or screenshot each).
• ☐ For each: record trigger app, action apps, monthly task volume, and the human owner.
• ☐ Tag obvious zombies (zero runs in 90 days) for deletion, not migration.

Phase 2 · Score & pick the top 20%

• ☐ Rank by task volume × business value.
• ☐ Identify the top ~20% — they usually carry ~80% of total tasks. Port these first.
• ☐ Decide the long-tail policy: leave low-volume Zaps on Zapier's cheapest tier, or delete.

Phase 3 · Rebuild in n8n

• ☐ Recreate each high-value Zap as an n8n workflow (cloud or self-hosted).
• ☐ Add what Zapier lacked: retries, idempotency, a dead-letter path, and audit logging (see our error-handling playbook).
• ☐ Pin credentials in n8n; never hard-code keys.

Phase 4 · Shadow mode

• ☐ Run the n8n workflow beside the live Zap on the same triggers.
• ☐ Compare outputs until variance is acceptable (we target <0.5%).
• ☐ Log every run for the proof-of-match record.

Phase 5 · Cut over

• ☐ Switch the live trigger to n8n.
• ☐ Pause (don't delete) the Zap — it's your one-week rollback.
• ☐ Watch error rates and the audit log for the first 48 hours.

Phase 6 · Decommission & the four traps

• ☐ Delete migrated Zaps; downgrade or cancel the Zapier plan once task volume drops.
• ☐ Trap 1 — webhook URLs. Anything pointing at the old Zap webhook must be repointed before cutover.
• ☐ Trap 2 — silent rate limits. n8n can fire faster than Zapier throttled; respect downstream API limits.
• ☐ Trap 3 — built-in formatters. Zapier's Formatter steps have no 1:1 node; rebuild with Set/Code nodes and test edge cases.
• ☐ Trap 4 — auth scopes. Re-grant OAuth scopes in n8n; Zapier's connections don't transfer.