Zapier vs Make vs n8n: which one should you choose? a recommendation for three common situations

01

Zapier, Make and n8n on the same axes

Comparisons usually line up features. The decision turns on seven things you only see on the invoice, in an audit and in the implementation calendar. Below are all three tools on the same axes.

The n8n column describes the variant people compare most often, the install on your own machine. There is also the vendor's cloud, and then most of the differences in that column disappear, because upkeep goes back to the provider.

Both cloud vendors move their plan thresholds, so take them from the vendor price list with the date you checked, not from an article. Catalogue sizes and billing units come from the Zapier and Make price lists, as of September 2026.

02

How does billing actually differ between the three?

This is the difference that catches people out on the first invoice. All three platforms sell usage, they just each count something different, and the name of the unit gives nothing away.

Zapier counts tasks, meaning actions that ran. A trigger is not a task, and neither is a filter that stops the flow. Make counts operations, meaning every module call: the trigger, a read and a write all cost the same. Self-hosted n8n counts nothing, because you pay for the machine, not for the steps.

That difference decides which tool gets cheaper as complexity grows. Zapier and Make stay predictable on short flows that run rarely. n8n becomes predictable once you have many flows with many steps in them.

• Zapier: a task is an action that ran; the trigger and a run stopped by a filter stay off the counter
• Make: an operation is one module call, including the trigger and every data read
• Self-hosted n8n: the bill is the machine plus the hours of whoever watches it, whatever the node count
• n8n on the vendor cloud: the unit is one full flow execution, not a step inside it
• AI model usage is billed separately in all three: EUR 0.005 to 0.07 per query, paid to the model provider directly

03

Where does your data sit and who keeps the engine running?

Zapier and Make are vendor clouds. Data from the flow travels through their infrastructure and no configuration gets around that. It is not a flaw, it is the model: that is exactly why you have no machine to maintain and no Sunday call-out when something breaks.

Self-hosted n8n flips the arrangement. The instance runs on your machine, your IT team grants the access, and data leaves only where you send it: to the API of the system you connect to, or to a language model provider if the flow calls one. We cover the install options on the n8n automation page.

The consequence is administrative. A platform that carries customer personal data belongs in your record of processing activities and needs a data processing agreement under the GDPR. With Zapier and Make that is a box to tick with your legal team. With n8n the processing stays with you, but you take on work that never shows up on an invoice.

04

Which one fits your situation?

Below are the situations companies come to us with, and the recommendation for each one. The right-hand column leads to the page where we go through that platform in detail, with install options and the work involved.

If two rows fit at once, the higher one wins: a requirement about where data may sit is harder than convenience while building, and having nobody to run a server ends the n8n conversation regardless of everything else.

The first two rows share one condition: somebody on your side owns the instance by name. Without that person the recommendation changes to Make or to n8n on the vendor cloud, and we say so on the consultation, before any quote exists.

05

When does the tool you picked stop being enough?

None of these platforms is a forever choice, and it helps to know that early. Each has a threshold past which keeping the flow alive costs more than moving it somewhere else. Those thresholds show up in numbers, not in feelings.

Moving between platforms is not a file import. The flow gets rebuilt, because modules and nodes map data differently. What does carry over is the expensive part: the field mapping and the list of exceptions proven in production.

• Zapier stops being enough at the third path in a process: branching sits on paid plans and the flow splits into several separate Zaps
• Zapier and Make stop being enough on lists: a loop over 100 records means 100 calls of every step behind it
• Make stops being enough when data arrives in the flow that cannot travel through the vendor's infrastructure
• n8n stops being enough when the person watching the instance leaves and nobody picks up updates and backups
• All three stop being enough when the process is a mess: automation locks in steps that should not exist

06

How do you check this recommendation against your own bill?

The recommendation in the table is a starting point, not a verdict. Checking it against your process takes fifteen minutes and needs four numbers, three of which are already in your systems. The selection method that does not depend on tool names is in our guide on how to choose an automation platform.

1. 01

   Take the run count from the system

   How many times the process fired in the last full month: as many as there were tickets, orders or closed deals. A number from the team's memory usually comes in low.

2. 02

   Count steps and records separately

   Walk the process step by step and mark the read steps: in Make they cost an operation, in Zapier they cost nothing. If there is a list in the middle, count the records, because those are what multiply the bill.

3. 03

   Convert into tasks and into operations

   Tasks: number of write actions times number of runs. Operations: number of module calls times number of runs, trigger included. Put both numbers next to the current vendor price list.

4. 04

   Price the third option properly

   On the n8n side count the machine at your hosting provider, the hours for updates and backups, and the response to a failure outside working hours. Only then do the three amounts compare honestly.