Equivalent workflow evaluation worksheet — prepared 28 September 2026

This is a test protocol, not a new benchmark. Every run is marked not_run.
Keep the existing September 6 local n8n result separate:
https://www.marsentech.com/examples/zapier-to-n8n/results.json
No equivalent Zapier execution or billing result has been measured by this release.

1. Use fictional records, one explicit event ID and one controlled destination.
2. Hold the input corpus, field mapping, acceptance criteria and retry policy
   constant. Record meaningful platform differences rather than hiding them.
3. A cloud Zap cannot reach the existing lab's 127.0.0.1 endpoint. Configure an
   isolated, authenticated test destination reachable by both environments.
   Do not expose a production CRM or the unauthenticated local lab to the internet.
4. Confirm the available connectors, plan and task allowance before a test.
5. Run normal, replay, temporary failure, invalid input, missing permission,
   permanent failure and restore/replay checks. Preserve redacted execution
   evidence and the resulting destination record for each scenario.
6. Read actual usage from each account's usage report; keep trials, testing,
   production executions, replays and retries identifiable. Never substitute
   canvas step count or this worksheet's expected result for measured usage.
7. Repeat under comparable conditions before claiming timing or reliability.
8. Record implementation time, operator recovery time, maintenance assumptions,
   hosting and third-party fees separately from billed platform units.

The worksheet starts blank for both platforms. Historical n8n observations do
not count as a matched new run. A missing result is unmeasured, never zero.
No cloud endpoints, credentials, customer data or active workflows are included.