Why this happens
Credits are counted per step, not per run. A step that fails, or that has zero input and output rows, does not itself consume a credit — but every other step in that same run that completed successfully still counts normally, whether it ran before or after the failing step. So if a five-step flow fails on step 4, steps 1–3 already ran and consumed credits before the failure happened. The run shows a credit charge even though it didn’t complete end to end. See Understanding and Managing Credit Usage in Flows for how each step type is priced.The step that actually errored does not consume a credit for that run. The charge reflects the other steps in the run that executed successfully.
How to check what happened
- Open Run History on the flow and expand the run in question.
- Confirm which step failed and read the error message.
- Everything upstream showing a completed status is what consumed credits on that run.
If you think credits were consumed unnecessarily
- Debugging a failing step: re-running while you iterate on a fix consumes credits each time, same as any run. Test with a smaller sample while you sort out the failing step, then run the full dataset once it’s fixed.
- A verified Parabola platform incident caused the failure: contact help@parabola.io with your flow run IDs and the affected date/time range. Refund eligibility is reviewed case by case once the incident is confirmed internally — see Billing and pricing.