n8n, Make or Zapier? Automation tools for SMEs compared

The short answer: Zapier for individuals with a few simple flows, Make for marketing and sales teams connecting many tools, n8n as soon as data sovereignty, cost at high volume or AI agents come into play. The decisive difference, though, is not the headline price but the billing model — and that is exactly where the surprises on the invoice come from.

Last checked: October 2026. Prices verified against the vendors’ own pages on the day of publication.

Disclosure: I build predominantly with n8n myself and have written several articles about it. Which is why the cases where Zapier or Make are the better choice come first in this article — and they are more common than I, as an n8n user, would like.

The short version

• The billing model is the core: n8n counts complete runs, Make and Zapier count individual steps.
• A 20-step flow, run 1,000 times a month = 1,000 units on n8n, but 20,000 on Make and Zapier.
• Zapier is the simplest and the most expensive per step. Free tier: 100 tasks.
• Make is the cheapest entry (Core $9) and strong when connecting many tools.
• n8n is free self-hosted — at the price of taking on operations yourself.

The point everything turns on

Before we talk prices, we have to talk units. The three vendors count different things, and that difference is larger than any price difference.

What that means in practice takes one calculation to show. Take a typical flow: an enquiry arrives via the contact form, gets validated, written to the CRM, assigned to an employee; a confirmation email goes out, an internal notification, a calendar entry — with a few checks and branches in between. Say 20 steps. The whole thing runs 1,000 times a month.

• n8n: 1,000 executions. Fits inside the Starter plan at €20 (2,500 executions).
• Make: 20,000 credits. The Core plan at $9 includes 10,000 — so you need more.
• Zapier: 20,000 tasks. Well above the entry tiers.

This is why automation projects blow their cost estimates. You budget for “1,000 enquiries a month” and overlook that this can become 20,000 billed units. Conversely: if your flows are short — three or four steps — the effect almost disappears and the models converge.

The rule of thumb: the longer and more branched your flows, the more strongly the billing model argues for n8n. The shorter and simpler, the less it matters — and then usability decides.

The plans

Two notes. First: with Make, scenarios stop when credits run out — you can buy more or enable auto-purchase, but the standstill comes first. Anyone putting business-critical flows on it should know that. Second: the jump from n8n Pro (€50) to Business (€667) is substantial, with nothing in between. High volume on a limited budget therefore quickly leads to self-hosting.

When Zapier wins

When you want to start quickly without technical knowledge. Zapier is the most accessible of the three. If you need a connection between two tools this afternoon and have no intention of learning a subject, this is the place.

When you have few, short flows. With two- or three-step automations the disadvantage of per-step counting barely registers — and then usability counts for more than the billing model.

When you need to connect an exotic tool. Zapier has traditionally offered the largest catalogue of ready-made integrations. If your industry tool is supported anywhere, it is most likely here.

When Make wins

When you want to work visually and connect many tools. The graphical builder is Make’s strength — you see the flow rather than reading it. For marketing and sales teams where several people work on the same flows, that is a genuine argument.

When the budget is tight and flows are medium length. Core at nine dollars with 10,000 credits is a very cheap entry — as long as your flows do not run to twenty steps.

When you do not want to run a server. Like Zapier, Make is a pure service. No operations, no updates, no responsibility for availability.

When n8n wins

On long or branched flows. See above — here the difference is not gradual but an order of magnitude.

When you need data sovereignty. Self-hosted means the data does not leave your environment. For businesses with sensitive data or strict requirements that is often the only viable route — and the Community Edition is free for it.

When AI agents come into play. n8n has developed furthest here; I have covered this in several articles — for instance on building n8n workflows with Claude and MCP and on agentic RAG in n8n.

The price for it: self-hosting means you own availability, updates and backups. That is not a footnote. If nobody at your place can or wants to look after a server, the free option is not free but merely unpaid — the work happens regardless. In that case the cloud plan is the more honest choice.

A realistic selection route

1. Write down three planned flows — with the approximate number of steps and the expected frequency per month.
2. Run both models: runs (n8n) versus runs × steps (Make, Zapier). That single calculation answers the cost question completely.
3. Check the data question: may this data go to an external service? If not, only self-hosting remains.
4. Check who is going to look after it. Without a named person, every automation becomes a liability after the first failure.
5. Start on the free plan and build exactly one real flow. That answers more questions than any comparison.

For getting started with concrete examples: Five practical automations for SMEs.

Frequently asked questions

Is n8n harder than Make or Zapier?

Somewhat, yes. The gap has narrowed, but n8n demands more willingness to engage with data structures. If nobody at your place enjoys that, Make is the more realistic choice — a tool nobody operates saves no money.

Can we switch later?

Only with effort — flows cannot be transferred between platforms; they have to be rebuilt. That is an argument for genuinely doing the calculation in point 2 above beforehand, rather than having the invoice explain it to you after a year.

What does self-hosting n8n really cost?

The software, nothing. A small server sits in the low double digits per month. The real item is care: updates, backups, troubleshooting when a flow stalls overnight. Budget a few hours a month — and answer the question of who does it during holidays.

Do we need automation at all if we have AI assistants?

They are different things. An assistant helps one person with one task. An automation runs without a person, identically every time. For anything that happens regularly and identically, an automation is the right tool — including, and especially, when an AI step sits inside it.

What about the automations the vendors ship themselves?

If your flows happen entirely within one platform, you often need none of these three — Microsoft, Google and most CRM systems have their own automation on board. These three tools pay off where you automate between systems.