Make, Zapier or n8n: which fits your process, and when to write code

Make, Zapier and n8n compared at a glance

The rest of this post explains each row and what it means for one real workflow.

How do Make, Zapier and n8n charge?

Each tool counts something different, so the same workflow costs a different number of units in each.

Zapier counts tasks. Zapier's help centre says "a task is any successful action that runs in Zapier." The trigger, filters, paths and Formatter steps never count.

Failed steps don't count either. On plans with pay-per-task billing on, extra tasks are charged up to three times the plan's limit. On other plans, Zaps stop running until the next billing cycle.

Make counts credits. Make's pricing page says credits are now the billing unit. "Every time a module in a scenario performs an action, it consumes credits," and most actions use one.

A module is one box on Make's canvas. Reading data from an app or a webhook counts, but the router that splits a flow does not. A webhook is the web address your form sends each new inquiry to. When credits run out, scenarios stop until you add more.

n8n counts executions. n8n's pricing page charges per workflow execution "regardless of complexity". An execution is one full run of a workflow, however many steps it has.

A worked example shows the difference. Say a new web inquiry should create a contact in your CRM, open a deal, post a message to your sales channel and send the person a confirmation email. That is one trigger and four actions, and you get 400 inquiries a month.

The units are not prices. On 14 October 2026, Zapier's Professional plan cost $19.99 a month billed annually for 750 tasks ($29.99 billed monthly), and its Team plan $69 billed annually. The example's 1,600 tasks need a larger tier. n8n Cloud's Starter plan was €20 a month, billed annually, for 2,500 executions.

Make's page sells plans by the number of credits you buy. Prices change and depend on volume, so price your own workflow on each vendor's page before you choose.

The pattern holds as workflows grow. A workflow with many steps costs more in Zapier and Make with every step you add. In n8n it costs the same per run, so long workflows are relatively cheaper there.

Where is your data stored, and does it matter under GDPR?

It matters, because these tools process personal data on your behalf. Every inquiry, invoice or support email that passes through them is that data.

Under GDPR, the tool is your processor: a company that handles personal data on your instructions. Article 28 requires a written contract with it, usually called a data processing agreement (DPA). Chapter V of GDPR adds rules for sending personal data outside the European Economic Area. Each tool answers these differently.

• Zapier says it "hosts data in AWS servers located in the United States." It is certified under the EU-US Data Privacy Framework, the scheme that allows transfers to certified US companies.
• Zapier's DPA includes the EU's standard contractual clauses. These are contract terms the European Commission approved for sending personal data outside the EEA. Zapier also says it can't sign other companies' DPAs, so you accept its own.
• Make lets you choose the data centre location for each organization, in the EU or the US. Its help centre is blunt: "You cannot change the location of the data center after you create the organization." An EU zone also doesn't cover everything. Make says a few of its AI features always send requests to the US, whatever zone you chose.
• n8n says its hosted plans store data "within the EU, on servers located in Frankfurt, Germany." If you host n8n yourself, the data stays wherever you run it.

US hosting is not automatically unlawful. It does mean one more thing to document, and some customers or sectors will ask you about it. If your process touches health, finance or HR data, ask where the data sits before you build, not after.

One practical step costs nothing. If you choose Make, pick the EU zone when you create the organization, because you can't move it later.

Is n8n free to self-host?

You can run n8n's Community Edition on your own server without a subscription. You pay for the server and for the hours to keep it running: updates, backups, monitoring and fixing it when a run fails at night.

The licence has a limit worth knowing. n8n is published under the Sustainable Use License, which allows use "only for your own internal business purposes or for non-commercial or personal use." Running it for your own company is fine. Selling access to it as a service to others is not.

Self-hosting also leaves the connectors to you. n8n has fewer ready-made app connections than Zapier, which advertises more than 9,000 apps. That is why we tested a self-hosted n8n and kept Make: we would have had to build too many app connections ourselves.

Who on your team can maintain each one?

Whoever builds the workflow becomes the person who fixes it when it breaks. Choose with that person in mind.

• Zapier is a linear list of steps. Someone in operations can build and repair a simple Zap without technical help.
• Make shows the whole flow on a canvas, with routes, loops and error handlers. It suits someone who thinks in data structures, and it rewards a little training.
• n8n suits a developer or a technical team. It runs JavaScript and Python inside workflows, and self-hosting means someone owns a server.

Whichever you pick, put the account in your company's name, not a freelancer's or an agency's. Then nothing has to be handed back if the person who built it leaves.

A short runbook helps too: a page that says what each workflow does and who fixes it. Our post on monitoring and maintaining automations shows what goes in one.

When is custom code the better route?

Custom code means software written for your process and run on infrastructure in your name. It is the better route when the workflow has stopped being glue between apps and has become part of how your company works. Five signs:

1. The step count is high and volume keeps growing. In Zapier and Make, every added step and every extra run costs more units.
2. The connection you need doesn't exist. You end up calling a raw API from inside the tool. An API is the interface one program uses to talk to another. Calling it this way is code without code's tools.
3. Mistakes are expensive. Code can be versioned, tested and reviewed before it reaches a customer. A changed module on a canvas often can't.
4. The data has to stay in your infrastructure. Zapier and Make can't meet that. Self-hosted n8n can, if you run the server.
5. The logic is your advantage. If the workflow encodes how you price, route or decide, you want to own it outright.

Custom code is not free either. Someone has to write, host and maintain it. For a two-step workflow that runs a few hundred times a month, a no-code tool is the cheaper and saner choice.

This is the direction we're taking ourselves. On Provap's site, which we built, quote requests go to a Make.com webhook, and that is still the right tool for that job. For our own operations, we're slowly moving work from Make to fully custom software, managed by our AI agent team.

How to choose in practice

Start from one process, not from a tool. Write down its trigger, its steps, what data passes through, and what happens when a step fails. Then decide in this order:

1. Data first. If the data must stay in the EU, rule out what can't guarantee it, and choose the zone before you build.
2. Then the person. Pick the tool the person who will maintain it can actually maintain.
3. Then the bill. Count units for your real volume on each vendor's pricing page.
4. Then the exit. Expect to rebuild, not copy, a workflow if you switch tools later.

Some processes shouldn't be automated yet at all. If the steps change every week or live in one person's head, read when not to automate a business process first. If you are connecting a CRM, connecting AI to your existing Pipedrive CRM walks through the routes, including no-code connectors.

What we do, and what we don't

Our AI automation work starts with an audit of one workflow, not with a tool. We map where the hours go, then build on Make or in custom code, whichever fits the process, and we say so if neither is worth it. Accounts and repositories are in your company's name from day one, not ours.

If you have one process in mind, tell us the workflow. A person replies within one working day. The first call maps that one workflow, with no pitch.