Which automation platform is cheapest for my workload?

How this is calculated

The unit count comes first, because it is where the platforms actually differ.

Zapier bills per task, roughly one step execution. A six-step Zap running 8,000 times is about 48,000 tasks.

Make bills per operation, with a similar shape - and iteration is where it bites. A scenario looping over 500 records runs the steps inside that loop 500 times, so a single run can consume 500-plus operations.

n8n Cloud bills per execution, so that same loop is one. Steps and loop size do not change the count at all, which is why n8n's advantage grows with workflow complexity rather than with volume alone.

Self-hosted n8n has no per-run charge, in exchange for a server and operational attention. The server is cheap. The attention is not, and it is the line most comparisons leave at zero - which is how self-hosting appears free and then is not.

The loop share input

Iteration rarely wraps the whole workflow. A scenario that fetches records, loops over each one for three steps, then sends one summary has some steps running once and some running per record. That split is what the loop share sets, and on Make it is frequently the difference between a comfortable plan and a surprising invoice.

What this deliberately leaves out

Connector coverage, which is Zapier's real advantage and is not a cost question. If the integration you need exists on one platform only, the comparison is over before the arithmetic starts.

Build time. A workflow that takes two days on one platform and half a day on another has a cost difference that dwarfs several months of subscription.

Plan tiers. This models one plan per platform. Real pricing steps rather than scaling smoothly, so treat a near-tie as a tie and check the tier boundary.

For which platform suits which job - organised by what you are trying to do rather than by feature - see Zapier, Make or n8n: a decision table. On what self-hosting actually asks of you, see self-hosted n8n.