I have shipped real workflows on all three. Here is what I actually pick, and when.
We've shipped automations on all three platforms over the last two years. Not tutorials—production workflows for clients. This isn't a feature matrix. It's how we actually decide.
Zapier wins on breadth of integrations and speed to build. If a tool has a Zapier integration, it works. The interface is the most approachable for non-technical users.
The problems: pricing scales aggressively with task volume, and the logic layer is limited. Multi-step conditional workflows get ugly fast. The 'Paths' feature helps but isn't designed for complex branching.
Pick Zapier when: the client will maintain it themselves, the workflow is simple (trigger → 2–3 actions), and volume is low enough that per-task pricing is acceptable.
Make's visual canvas is genuinely good. Complex workflows are easier to reason about than a linear Zapier chain. The data manipulation is more powerful—array functions, iterators, aggregators.
The learning curve is real. Make uses concepts (scenarios, bundles, modules) that aren't intuitive. Debugging a failed scenario requires understanding its internal data model.
Pick Make when: the workflow involves transforming or aggregating data, the client has some technical literacy, and you need visual documentation of the logic.
n8n is self-hostable, open source, and has a code node that accepts JavaScript. For technically complex workflows, nothing touches it.
The downside: you own the infrastructure. A self-hosted n8n instance needs monitoring, updates, and backup. The cloud version exists but partially defeats the cost argument.
Pick n8n when: data sovereignty matters (GDPR, legal, healthcare), the workflow is complex enough to warrant code, or you're building something you'll maintain yourself.
“The right tool is the one you'll still understand when you open it at 11pm because something broke.”