I quit Zapier three weeks ago. Here's what actually happened.

The bill was the problem. Not the tool. Zapier is excellent — clean interface, 8,000 integrations, you can build a functional zap in under 15 minutes without reading any documentation. But I was on the Professional plan at $49/month, and my automations were getting more complex. Every branch, every loop, every additional action step is a task. The math starts to go sideways the moment your workflows grow up.

n8n charges per workflow execution, not per step. One run of a 25-step workflow costs the same as one run of a 3-step workflow. That's a different product, not a feature upgrade. I switched three weeks ago. I'm not going back. Here's the honest version.

Why I Left Zapier

I built a lead enrichment workflow. Someone fills out a form, the workflow hits Clearbit for company data, scores the lead, creates a contact in Airtable, sends a personalized Slack notification to my team, and logs the full interaction to Google Sheets. Eight action steps.

On Zapier Professional, that's 8 tasks per form submission. At 200 leads a month, that's 1,600 tasks for one workflow — before any other automation ran. I burned through my monthly cap two weeks into March. Zapier's next option: upgrade to the Team plan at $103/month.

I opened n8n on a Wednesday. I didn't go back to Zapier after that.

The First Week Is Hard — No, Really

n8n's interface is visual like Zapier's. Nodes connected by lines, trigger on the left, output on the right. The resemblance ends there.

Where Zapier puts pre-built action templates in front of you, n8n surfaces the underlying structure. You'll write JSON paths by hand. You'll configure HTTP headers manually. The if/else node works, but you have to understand what it's doing -- not just click 'add condition.'

I rebuilt that lead enrichment workflow in n8n in about six hours. It took 90 minutes originally in Zapier. That gap is real, and I won't minimize it.

What helped: n8n's execution preview. When you test a node, the output appears immediately beside the settings panel — you see exactly what came in and what passed forward. It's a better debugger than Zapier's. I just didn't appreciate it until I needed it.

What didn't help: the documentation assumes a baseline of technical comfort that not everyone has. It's accurate, thorough, and not a tutorial.

By end of week two I'd rebuilt seven workflows. By week three I was building things Zapier couldn't have done at my task budget.

What I Miss

The integrations are the expected answer, and they're real — 400 native connectors against Zapier's 8,000 is a gap you'll feel the moment you try to automate something in the long tail. Two of my workflows needed apps n8n didn't natively support. I built them via the HTTP request node, which works fine. It added an hour I wasn't planning for.

The unexpected miss: Zapier's decade of documented failures. I still open Zapier's help pages when I'm reverse-engineering an unfamiliar API structure. The answers there aren't about Zapier — they're about how these third-party APIs actually behave in the wild: the rate limits documentation doesn't mention, the authentication edge cases, the field that always comes back null on the first call. That accumulated knowledge doesn't migrate. It just stays over there. n8n's forum is 40,000+ members and genuinely useful, but ten years of answered questions is ten years of answered questions.

What I'm Not Going Back To

Counting tasks.

On Zapier, every automation design had a shadow cost calculation running in the background. I avoided loops because each loop iteration multiplied task consumption geometrically. I simplified branches to stay under the monthly ceiling. The tool was shaping my architecture in ways I didn't notice until they stopped.

Last month I added a loop to a workflow that runs 45 iterations. It cost one execution. I didn't think about it at all. That's the difference.

At scale: n8n Pro cloud is €60/month for 10,000 executions. Self-hosted is free and unlimited — running on a €7/month VPS, no ceiling. Compare that to Zapier Team at $103/month for 2,000 tasks.

Should You Switch?

If you're under 1,000 tasks per month and your zaps are simple: no. Zapier is faster to build in, friendlier to non-technical users, and the integration gap matters more when you're in the shallow end.

If you're regularly hitting Zapier's task ceiling, building multi-branch workflows, or running loops: yes. Budget one weekend for the first five rebuilds. It's slower upfront. You'll stop paying for complexity within the first billing cycle.

The switch is worth making. If JSON and API calls are genuinely unfamiliar, budget more than one weekend — the community will get you there, just not instantly. But if you're paying $49–$103 to Zapier every month and counting tasks like a grocery budget, you're paying for comfort, not capability.

Do the math. I waited a year longer than I should have.

What Should I Review Next?

I'm building the Tool Tuesday queue. Drop a comment with the tool you're either genuinely curious about or quietly frustrated with — both make better reviews than neutral opinions.

👇 Leave a comment below — your vote shapes the list.