n8n vs Make vs Zapier: which one fits a Danish SMB best?

Pick Zapier if the flow has few steps and has to touch an app nobody else supports. Pick Make if the flow branches and you want the most logic for the money. Pick n8n if data has to stay on your own server, or if volume is high.

That is the short answer to n8n vs Make vs Zapier, and it holds in most cases. But the choice between the three rarely decides whether the automation is still running in a year. Error handling and maintenance do.

We build this kind of solution for Danish SMBs, so we have put all three platforms into production for customers. This guide is based on what actually went wrong afterwards, not on feature lists.

n8n vs Make vs Zapier: the short answer

On paper, all three do the same thing: they listen for an event in one system, fetch data, and write it on to another. For a simple flow, say a new form submission that should become a case in your CRM, you will not see any difference in the result.

The differences come down to four things: what you are billed per, whether you can host it yourself, where the data lives, and how steep the learning curve is.

Note what is not in the table: the number of integrations. Zapier has the most, and it is still the best argument for Zapier. But in practice most SMBs use the same fifteen systems, and all three support them.

Integrations are not where projects stall either. That happens when a system can only be reached through its own API, and someone has to write the call by hand. n8n does that without fuss, Make can do it with some effort, and at Zapier the option sits in the more expensive plans.

The prices are on Zapier's pricing page, Make's pricing page, and n8n's pricing page. We do not repeat them here, because they change several times a year and because the numbers cannot be compared directly anyway. That is what the next section is about.

Tasks, operations, and executions: the difference that moves the price

Zapier counts tasks. Every action step that runs costs one task. Filters, paths, and Formatter are free, but everything else counts. A flow with four action steps therefore costs up to four tasks every time it runs.

Make counts operations, sometimes called credits. Every module call counts, so it is essentially the same logic as Zapier. The price per unit is simply lower, and that is the whole reason Make typically ends up cheaper than Zapier on the same flow.

n8n counts executions. The whole workflow is one run, whether it has three nodes or forty. That changes the maths completely for complex flows.

A calculation that shows the difference

Take a flow that creates a case, looks up the customer, sends a confirmation, and writes a line in a spreadsheet. That is four action steps. If it runs 2,000 times a month, you have used 8,000 tasks at Zapier, 8,000 operations at Make, and 2,000 executions at n8n.

The order flips if the flow has only one step. Then it is 2,000 units on all three, n8n's advantage disappears, and Zapier's free filters suddenly start to matter.

That is why comparing starting prices is pointless. Count the steps in your three most important flows, multiply by the volume you expect, and then do the maths on the three pricing pages. It takes twenty minutes and moves the decision more often than feature lists do.

That gives two rules of thumb. If you have a few flows with many steps and moderate volume, Zapier is the most expensive and n8n the cheapest. If you have many flows with two steps and very high volume, n8n cloud can end up the most expensive, because every small run counts as a full execution.

Self-hosted n8n has no meter on runs. That sounds like free, and it is not: you pay in server and uptime. Patching, backups, upgrades, and monitoring are work that someone has to do every month. We have worked through what AI automation costs for both models, and the licence is almost always the smallest line.

Where may your data live? GDPR and EU zones

This is why the list looks different for a Danish company than for an American one.

Make lets you choose a data region at organisation level: US1, US2, EU1, or EU2. But Make itself says that choosing EU1 or EU2 does not automatically mean that everything is processed and stays in the EU, and that some features go through the USA regardless of zone, see Make's documentation on data zones. If you add your own AI provider, the guarantee lapses entirely.

Zapier processes data in the USA by default. That is not illegal, and they have a data processing agreement and standard contractual clauses in place. But it means a transfer assessment, and it means you must be able to explain it if a customer asks.

Self-hosted n8n is the only one of the three where the server location is your choice alone. If it runs with a Danish or European provider, customer data never leaves the EU. For a law firm, a clinic, or a company with public-sector customers, that is often decisive, and it is frequently why n8n wins despite the steeper learning curve.

This kind of clarification is also why the smallest companies come last. According to Statistics Denmark, use of artificial intelligence in Danish companies with more than ten employees rose from 15% in 2023 to 28% in 2024, while Dansk Erhverv has measured that 96% of companies with more than 250 employees are on board, against 57% of those with fewer than ten. These are Danish figures. The gap is rarely the technology. It is that nobody has had time to take a position.

Whichever tool you choose, check two things:

• Data processing agreement. It is typically only available on paid plans. A free account with real customer data in it is a problem you built yourself.
• What the flow actually passes on. Most transfers do not happen in the platform, but in the AI models and services the flow calls. Look at the whole chain, not just the first link.

Who maintains the flow in six months?

The section no comparison writes is the one that decides the outcome. Automations do not break because the platform is bad. They break because a system changes its API, an employee renames a field, or a customer writes something the flow was not built for.

So ask three questions before you choose:

1. What happens when the flow fails? A flow without error handling passes on half the data instead of stopping. All three platforms can handle it, but none of them does it by itself.
2. Who gets notified? If the answer is "we find out when the customer calls", the flow is not in production. It is an experiment.
3. Who can fix it? If one person in the company built it, and that person is on holiday, you have no cover. Documentation is the difference.

AI nodes make error handling more important

All three have added AI nodes, and they make it easy to put a language model into a flow. They also make error handling more important, because a model can answer wrongly without failing. A flow that writes a model's answer straight into the customer's case needs a human in the chain.

n8n goes furthest, with its own agent nodes and memory between runs. Make and Zapier cover the most common cases well enough. The difference matters little if you are automating invoicing, and a lot if you are building customer service on top of a model.

Here Zapier is genuinely the best for a company without a technical owner. The error log is readable, and you can often fix a flow yourself without help. Make takes more practice, but shows the branching visually. Self-hosted n8n has the best options for monitoring and the highest risk if nobody looks after the server.

This is also where the real calculation sits. On average we find 15 hours a week that can be automated away at a customer, and we have built 53 automations. Those are not the hours that are hard to find. The hard part is the hours that are still saved a year later. We have described how we measure the ROI of AI automation, because a flow without a measurement is impossible to defend after the first six months.

A concrete example of the difference: Markeity saves 20,000 DKK a month on systems we built once. That number holds because someone took responsibility for operations, not because the tool was the right one.

How to choose between n8n vs Make vs Zapier

Make the decisions in this order. The first one that gives a clear answer settles the choice.

1. Does data need to stay in the EU?

If the answer is an unconditional yes, because you work with health data, client cases, or public-sector customers, then self-hosted n8n is your starting point. If the answer is "preferably", Make with an EU zone can be enough, provided you read the caveats and document them.

2. Do you have a technical owner in-house?

If you have someone who can read an error log and fix a node, all three are on the table. If you do not, choose the easiest tool or buy the maintenance along with it. Self-hosted n8n without a technical owner is the classic expensive mistake: cheap to start, dead after the first upgrade.

3. How many steps do your flows have?

Few steps and few runs: Zapier is up and running fastest and the price never becomes a problem. Many steps: Make or n8n. Many steps and many runs: n8n, either cloud or self-hosted.

When the tool is the wrong question

Sometimes the answer is none of them. If your systems do not talk to each other without a manual go-between, it is system integration you are missing, not an automation tool on top of the mess.

And if the process is not written down, you are automating a workflow nobody agrees on. Map it first: see automating workflows for the sequence we use.

It can make sense to bring in a specialist when three things apply at the same time: the process touches real customer data, more than one system has to be updated, and nobody internally can take over the flow afterwards. Then the tool is a detail, and the deliverable is what matters: a solution in production, the documentation for it, and the ownership of both.

One last Danish factor that changes the maths: SMV:Digital covers 50% of the advisory bill, up to 150,000 DKK, according to Virksomhedsguiden. That makes the gap between building it yourself and having it built smaller than most people think.

FAQ

Is n8n better than Zapier?

No, they solve different problems. n8n is better if you want to host it yourself, have unlimited runs, or write your own code in the flow. Zapier is better if you want to be live today without a technical owner. For an SMB without a developer, Zapier is usually the right first step.

Is n8n free?

The Community Edition costs nothing in licence fees, but it is source-available under the Sustainable Use License, not open source in the OSI sense. Use is limited to internal business purposes, see n8n's licence terms. SSO, Git versioning, and several security features require a paid key. And the server costs money, whatever the licence says.

Is there a Danish alternative to Zapier?

Not one that can compete on integrations. The real Danish option is self-hosted n8n with a European provider, where you control data residency yourself. It is not a different product, it is a different operating model.

Can I switch tools later?

Yes, but you rebuild the flows from scratch. There is no usable export between the three. That is why data residency and ownership are worth settling before the first flow, not after the twentieth. If you build from a documented process, the cost of a switch is working hours, not lost knowledge.

Choosing between n8n vs Make vs Zapier takes an hour, and that hour is not what decides whether the automation is still running next year.