I run a 209-node automation pipeline on n8n. I modeled what it would cost on per-task billing.

The thing that separates n8n, Zapier, and Make isn't features or polish — it's the billing unit. So I wrote 60 lines of Python to find out where the unit actually bites. I run the entire content pipeline for a small publication on self-hosted n8n: 10 production workflows, around 209 nodes, on a single cloud box. The most-asked question I get about that setup isn't "why n8n" — it's "wouldn't Zapier have been easier?" Easier, yes. But the reason I didn't reach for a hosted per-task tool has nothing to do with features and everything to do with one number that nobody puts on the comparison page: the billing unit. So I modeled it. Here's the short version, then the code. The three tools bill on three different units n8n bills per execution — one run of a whole workflow counts as 1, no matter how many nodes it has. Zapier bills per task — every action that runs is 1. Make bills per operation — every module run is 1 (now metered as credits). That sounds like a