If Zapier, Make, n8n, Google Sheets, CSV, or a webhook workflow is broken, do not start with the whole system. Start with one bad input, one expected output, the actual log or error, and one replay step.
I scanned public automation repair jobs today. The recurring failure was missing evidence: no stable trigger sample, no expected state, no replayable proof. That turns a small repair into guesswork.
This is the intake shape I would ask for before accepting a Mikael 48h Fix job. Use it even if you never hire me.
01Pick one booking, form submit, webhook payload, CSV row, or sheet row. Not the whole workflow.
02Example: “this booking ID must update exactly one existing row,” or “this CSV must produce 24 valid records.”
03A fix is not complete until the bad case is replayed and the expected output is visible.